Business Analytics driven management accounting

(1)

Business Analytics driven management accounting

A sensemaking perspective

University of Groningen Faculty of Business and Economics

Nettelbosje 2 9747 AE Groningen Year 2017-2018 25-06-2018 Jan Norden (S3256251) Supervisor: A. Bellisario Co-assessor: D. Swagerman Master thesis A&C: Controlling track

Word count: 12.500

___________________________________________________________________________ Abstract

This study investigates the use of business analytics (BA) within a controlling department at Dutch postal company PostNL. BA use was found to be understudied on the individual level and the same goes for the application of sensemaking theory in management accounting literature. The BA driven sensemaking has been studied based on the paradigm model of Tillmann and Goddard (2008) through a case study at the PostNL head office in The Hague. This resulted in a sensemaking model which consists of three main processes: information provision to dashboards, analysing and interpreting in cooperation with the business and reporting to senior management and adjustment initiatives. The study shows that cognitive constraints are overcome by BA-driven sensemaking and that BA shifts human attention to the social processes within sensemaking. Also this study firstly covers strategical adjusting activities from a sensemaking perspective and it shows that BA supports the contextual roles of strategy implementation and data projects. Overall, the study provides more understanding about planning and control within BA-driven management accounting and shows commonalities between planning and control and organisational sensemaking.

(2)

2

1. Introduction

Through modern day technological advancement data is available to everybody and in enormous amounts. This is partly because companies like Google and Microsoft invested in huge data centres throughout the world, as Google recently opened one in the northern part of the Netherlands in the ‘Eemshaven’ (Vroegop, 2016). The availability of data has led to the emergence of ‘Big Data’. According to Laney (2001) Big Data is about large volumes of extensively varied data and capturing, generating and processing of it at high velocity. Big data holds great opportunities in increasing efficiency, effectivity, price setting, selection of people for jobs, quality improvement and customer relationship improvement (Chen et al., 2012; Davenport, 2006; McAfee and Brynjolfsson, 2012).

Since the availability of huge amounts of structured and unstructured data the challenge for companies is how to create value through it (Ylijoki and Porras, 2016). According to Silvi al et al (2010) this challenge also holds for management control since there is a lacking connection between data and managerial actions. The solution to this problem lies in the opportunities that business analytics (BA) provide us. Maisel and Cokins (2015) state that analytics simplifies the huge amounts of data in order to answer complex questions like: why does decision X lead to outcome Y? So this makes us question what the implications of BA on managerial control are and how these implications lead to a better connection between the data and the managerial actions?

Current management accounting research and practice has already tried to deal with connections between data and managerial actions by identifying non-financial performance driving information in techniques like the balanced scorecard (Kaplan & Norton, 1992) and activity based costing (Cooper & Kaplan, 1988, 1991). More recently Appelbaum et al. (2017) have already taken this to the next level and incorporated BA into the balanced scorecard, furthermore Silvi et al. (2010) and Raffoni et al. (2018) have made contributions by applying BA on other performance measurement systems. Unfortunately research on the behavioural implications of BA into managerial control are currently lacking. Brand and Holzblatt (2015) predict that BA will transform the way in which controllers are analysing and interpreting data, furthermore they warn controllers to “adapt or die” (p. 12). Nevertheless this still lets us wonder how controllers are currently using BA. Therefore the main research question of this this research is:

(4)

4 “How are controllers making sense of strategy through business analytics driven

management accounting?”

This research question will be addressed through a case study in which controllers are being observed while making sense of strategy and in-depth interviews are going to be held with these controllers. The foundation of the case study will be an application of ‘the paradigm model of sense-making in a strategic context’ developed by Tillman and Goddard (2008).

The academic relevance of the research questions lies in the further understanding of the individual implications of BA on management accounting since in current research this is understudied relative to the organizational implications (Appelbaum et al, 2017; Silvi et al, 2010; Raffoni et al, 2018; Brand and Holtzblatt, 2015). This research aims to follow up on Tillman and Goddard’s (2008) call for further development of the grounded theory of strategic management accounting. Amongst others they propose investigating management accounting and sensemaking in different contexts. Furthermore, this research aims to show how sensemaking is relevant to (research on) management accounting practices.

This research is also relevant for practitioners since it is addressing the research-practice gap which has been outlined by Tucker and Lowe (2014). This gap is addressed as this research explains how BA is deployed in the case companies.

The next section of the paper presents the theoretical background of the research. In here the theoretical foundation for the research question and choice of the theoretical framework will be addressed. Next the methodology section will be presented in which the set-up of the case study will be uncovered. Following up on the methodology section the findings will be presented and this thesis will end-up with the answering of the research question(s) in the discussion section.

(5)

5

2. Theoretical background

2.1 Organizational sensemaking

Sensemaking is a concept which has been studied in organizational literature for the last 20 to 25 years. We can define sensemaking as “the process through which people attempt to understand issues or events that are somehow surprising or confusing” (Maitlis, 2005, p. 21). Now let us get closer to the core of sensemaking and focus on the two main questions involved in sensemaking: “How does something come to be an event for organizational members?” and “What does an event mean?” (Weick, Sutcliffe & Obstfeld, 2005, p. 410). These questions show us that sensemaking is about constructing meanings and that these meanings are created through interpretations of events, issues and interactions which are either surprising, complex or confusing (Cornelissen, 2012; Isabella, 1990; Sackman, 1991; Sandelands & Stablein, 1987; Starbuck & Milliken, 1998; Weick & Roberts, 1993). So sensemaking is about constructing meaning through interpretation, but this meanings-construction process needs further unravelling to fully understand it.

In general, sensemaking can be divided into three processes: scanning, interpreting and responding (Daft & Weick, 1984; Thomas et al., 1993). In order to initiate the scanning, one has to collect information, which can be both external (Daft & Weick, 1984) and internal (Cowan, 1986). Whether it is internal of external depends on the organizational role of the sensemaking individual, for instance top management members would be making use of both internal and external information in order to formulate the appropriate strategy (Kraus & Strömsten, 2012), whereas for instance the Sales Director would probably use only external information. Unfortunately organizational members are limited in their capacity to process the information, which forces them to be selective in the information they use (Huber & Daft, 1987). The scanning process has also been referred to as a process of bracketing and labelling (Weick, Sutcliffe and Obstfeld, 2005). In the latter description people bracket or put emphasis on surprising or confusing events which are called ‘cues’. After the bracketing individuals start to label points of reference to be used in the next stage of sensemaking. In the next stage of sensemaking individuals try to attach meaning to the scanned, bracketed and labelled events. In order to do this individuals must firstly apply some structure to the information (Gioia, 1986; Taylor & Crocker, 1981) and after that they need to put meaningful labels to the information (Cantor & Mischel, 1979; Smart & Vertinsky, 1984; Taylor & Fiske, 1978). Cornelissen and Werner (2014) describe the aforementioned as searching for appropriate frames to interpret the

(6)

6 bracketed cues. They argue that putting a bracketed cue inside a frame helps to better understand

the cue, or as Weick (1995) puts it: “a cue in a frame is what makes sense, not the cue alone or the frame alone”. According to Cornelissen and Werner (2014) a frame is defined as a ‘knowledge structure’. These knowledge structures could be things like organizational plans, expectations or assumptions. The ‘framing’ of cues can be done through data analysis, but it is often done through conversation with others. This basically makes sensemaking also a social process (Weick, Sutcliffe & Obstfeld, 2005). This leaves us with the third and final stage of sensemaking, namely responding. In this final stage a third sensemaking question is central: “what do I do next?” (Weick, Sutcliffe & Obstfeld, 2005, p. 410). On the basis of the earlier constructed meanings an appropriate course of action can be chosen. This last stage also marks a cyclical relationship between sensemaking and organizational decision making.

In fact, sensemaking is believed to both precede and follow decision-making (Maitlis, 2005). Managers firstly engage in a sensemaking process before the decision is being made. After a decision has been made and executed managers engage again into a sensemaking process to make sense of the consequences of the aforementioned decisions. Since sensemaking is explicitly invoked when the current state of the world is complex and different from the expected state of the world (Weick, Sutcliffe & Obstfeld, 2005) it is a vital process in the current dynamic and constantly changing world. For instance Balogun, Bartunek and Do (2015) found in there longitudinal case-study at a fast-moving consumer goods corporation that sensemaking activities were key in implementing strategic changes imposed on by the head office on the local branches. Senior management were firstly to make sense of the changes imposed by the head office and secondly engaged in another sensemaking process when trying to implement changes in the local branches. The interesting part of their research is that the latter sensemaking was affected by the relational context in the local branches. The link between the interpretative part and the responding or action part of sensemaking has also been statistically tested by Thomas, Clark and Gioia (1993). In their research they studied the linkages between sensemaking activities on the basis of questionnaire data retrieved from public-access hospital CEO’s. The results show emerging positive relations between scanning, interpretation and action activities. Next to the aforementioned activities they also added performance to measure the outcomes of the action activities. So sensemaking seems to be valuable for decision-making in organizations. In the following section more attention will be paid to influencing and/or facilitating decision-making from another distinct field or research, namely Management Accounting.

(7)

7 2.2 Sensemaking in Management Accounting

According to Malmi and Brown (2008) management accounting (MA) is supporting decision-making at any organisational level by decision-making use of accounting systems that are specifically designed for this purpose. For instance, MA could entail the use of a budget in order to asses if a production process achieves a standard product cost. So, most often MA compares budgets to actual outcomes and on the basis of that corrective action is taken or budgets are changed. This is done through a single loop feedback (Chenhall and Moers, 2015). In this explanation of MA we can see similarities to sensemaking. The budget versus actual comparison refers to interpretation activities and the corrective actions refer to the responding activities in sensemaking.

Through the years innovation and digitisation also had its impact on MA. During the nineties Kaplan and Norton (1992) came up with the balanced scorecard (BSC), which introduced a more prospective viewpoint next to the existing retrospective viewpoint of traditional MA practices. They did so through connecting non-financial performance drivers (leading indicators) with traditional financial measures (lagging indicators). This called for more non-financial data on operational performances. These innovations were further accelerated and perfectioned through the rise of ERP systems and ‘Big Data’. Appelbaum et al. (2017) picked up on this in their development of the Managerial Accounting Data Analytics (MADA) framework. They used the concept of BA as a way to analyse data from both internal and external information systems in order to construct and measure the aforementioned connections between leading and lagging indicators. According to Davenport and Harris (2007) BA is about the use of data, IT, statistical analysis, quantitative methods and mathematical or computer based models to assist managers in gaining insights in their operations and enable them to make better and fact-based decisions. Appelbaum et al (2017) distinguish three types of BA: descriptive, predictive and prescriptive. Descriptive BA is about answering ‘what happened?’ questions, whereas predictive BA focuses on ‘what could happen?’ questions. Finally prescriptive BA answers the ‘what should be done?’ questions based on first two types of BA. In their paper Appelbaum et al (2017) essentially implemented the BSC methodology into performance measurement and planning and control decisions making aspects of MA in order to incorporate BA in the related process. From a sensemaking perspective these advancements seem to enhance the scanning activities by making it possible to scan more information by making use of Analytics software than a human could on the basis of its above mentioned

(8)

8 cognitive restrictions. Furthermore the Analytics software is capable of better and faster

interpretation and framing and lets humans focus more on the social aspects of sensemaking. Despite the various developments in MA like the ones mentioned earlier on, up until now research from a sensemaking point-of-view is lacking in MA literature. Fortunately Tillmann and Goddard (2008) did make a contribution of this kind in MA literature. They conducted a case study at German manufacturer of consumer and industrial chemicals in which they investigated Strategic Management Accounting (SMA). The research resulted in a ‘paradigm model of sensemaking in a strategic context’ (figure 1). The model consists of several activities. They found that at its core sensemaking is about increasing understanding of strategic situations and improving organisational transparency. This was done through understanding cause-effect relationships and linkages and ‘grasping the wholeness of situations’. Constant sensemaking activities were found in MA practices like reporting and cost analysis. To further explain sensemaking Tillman and Goddard (2008) found that it consists of three activities: structuring & harmonizing, bridging & contextualising and compromising & balancing.

Through structuring & harmonizing complex information was made more understandable and non-comparable information was made more comparable. It also included presenting information in standardized templates in order to increase understandability. In conclusion, information went through a process of transformation and came out transparent and understandable. The bridging & contextualising activities attributed to ‘grasping the wholeness’ by assessment of the past through performance measurement and making projections of the future on the basis of past data. In order to understand information it has to be understood in the context of the company, which is done through comparisons with the same phenomena in the past and other departments. Finally compromising & balancing refers to being aware of in-availability of resources, information, knowledge and experience. This awareness effects the filtering and choosing of information to be analysed as well as the summarising, aggregating and reducing of information that is going to be communicated by the sense maker. Tillman and Goddard (2008) also found so-called contextual conditions which either caused phenomena to take place or formed the backdrop. Two of these contextual conditions are the internal and external context. The external context was found to consist of the general external environment, competitive and customer pressures and external stakeholder expectations. These pressures were perceived as opportunities or threats which had a major impact on strategic decision-making. The internal context was found to consist of organisational visions, strategies and goals, organisational structures, accounting rules, diversity in product groups and the way

(9)

9 people appear and think in the company. The combination of internal and external context

formed a contextual field that underpinned all strategic decisions. The last three contextual conditions are ‘feel for the game’, ‘professional know-how’ and ‘sets of information’. These are called intervening conditions because they either facilitate or constrain the interactional strategies. ‘Sets of information’ basically stands for any information that can reach a sense maker. These can be either internal as external information. The concept also entails the ‘purposeful gathering of information’ or information collection. In contrary to the latter, ‘feel for the game’ is a less straight-forward concept. It refers to assessing the future on the basis of ‘gut feeling’ or the general features possessed by people. Furthermore it is about interpersonal communication and co-operation and a common language. Lastly the concept of ‘professional know-how’ was found to entail knowledge of accounting techniques and rules and knowledge of other disciplinary fields like marketing and production. Also experience is found to be part of the ‘professional know-how’.

Since the framework of Tillman and Goddard (2008) is a unique application of sensemaking into the field of MA literature and it targeted the same organisational group of people as in this thesis, this model is also going to be used in this thesis research.

We discussed the importance of sensemaking to organisational decision-making and therefore its value to MA practices. Furthermore we have determined the lack of research from a sensemaking point-of-view in MA literature and especially in light of current digitisation development it is interesting to research the sensemaking activities in current MA practices. So, how does sensemaking takes place in MA practices? And, how does this sensemaking look like in BA-driven MA practices? These questions are to be answered by means of the main research question: How are controllers making sense of strategy through BA-driven management accounting?

(10)

10

(11)

11

3. Methodology

The used research approach to investigate the sensemaking activities is grounded theory-building based on interviews and observations (Glaser & Strauss, 1967). This strategy was deemed most suitable due to the lack of research on the observed phenomena from a sensemaking point-of-view as was pointed out in the theoretical background section of this thesis.

3.1 Research setting

The organization studied is the largest Dutch postal and logistic solutions provider PostNL. At this time they employ 44.263 people and are mostly active in the Netherlands, but there are also subsidiaries in Germany, Belgium and Asia. PostNL is divided into three segments: Mail in the Netherlands, Parcels and International (figure 2). The interviews and observations were conducted in The Hague at the controlling department of the Mail in the Netherlands segment. The choice for PostNL and the Mail in the Netherlands controlling department is based on two particular reasons. Firstly PostNL has strong digitisation and data capabilities (PostNL, 2017, p28), which enables the company to come up with data-driven decisions and solutions. This is also applicable to the researched controlling department. Secondly because the researcher is currently employed by PostNL he has better access to controlling personnel and he is better informed about the internal environment, which enables him to conduct a deeper and more thorough analysis of the observed phenomena. So there is a mixture of several sampling strategies involved, namely convenience sampling (because of the above mentioned accessibility of the researcher), emergent sampling (part of the interviewees were selected based on emerging findings) and critical case sampling (due to the emergent data-driven strategy of PostNL) (Cohen and Crabtree, 2006).

The Mail controlling department is divided into several sub-departments like Operational Control (focus on sorting centres, preparation sites and the mail deliverers), Commercial Control (focus on revenues and sales process), Cost Control (focus on cost registration in the accounting systems) and Planning & Control (focus on coordination and communication to the controlling management team and PostNL’s Board of Management). As a result of the constantly declining mail volumes which are caused by digitisation, these controlling departments are under some pressure to keep on track of cost savings and price increases to customers. The cost savings and price increases are needed to be able to deliver the sustainable cash flow as mentioned in the PostNL strategy (PostNL, 2017, p26).

(12)

12 Figure 2: PostNL’s organisational structure (simplified).

3.2 Data collection

Most of the data was collected during a two week period, in which the researcher was daily present at the Mail controlling department. Beforehand he conducted an unstructured pilot interview with a Senior Data Scientist of PostNL to get familiar with Analytics use and terms at PostNL. Also the first draft of the interview guideline was tested on colleagues of his own department at PostNL and his family. These steps were taken to ensure that the questions were addressing the right issues, invoking the wanted conversation and easily understandable for practitioners. The interview guideline was based on the ‘paradigm model of sensemaking in a strategic context’ from Tillmann and Goddard (2008). Firstly the theoretical concepts from the model had to be translated into practical terms. This was done by examining the coding analysis from the Tillmann and Goddard (2008) paper and the above mentioned testing on family and colleagues. The selection of informants went through consulting the managers of both the Operational Control and Commercial Control departments. As the interviews were progressing, the researcher also felt the need to add a business controller from the Operational Control department and he approached the controller on his own initiative. The latter refers to the earlier mentioned emergent sampling.

During the aforementioned two week period he sat between the operational controllers and commercial controllers in The Hague. The days consisted of conducting one or two interviews per day and observing and noticing the daily activities that went on. Furthermore he scheduled two sessions in which he was shown the various tools, models and reports which were used in the department, from these meetings he took notes and received several screenshots. In total he conducted 11 semi-structured interviews with various informants, of which four were via telephone and the rest face-to-face (see also table 1 for the specifics). On the basis of the information from the first interviews the interview guideline was altered to make the questions even more to-the-point and to get closer to the core. Because the researcher was also interviewing informants from two other departments he made more changes to the interview guideline as some questions were not applicable to them in the original form. During the interview with the data analyst of Operational Control he was also shown the various

(13)

13 information systems that were used. Afterwards he received screenshots of this in order to be

able to triangulate the evidence from the interview. Data source and use

Informant Department Type of Data Duration Use in the Analysis Senior Data Scientist ACE1 _{pilot interview} _{58 min.} _{Familiarizing with BA}

context Senior Data Scientist ACE semi-structured

interview

44 min. Investigate sensemaking Commercial controller 1 Commercial

Control

semi-structured interview

Control

58 min. Investigate sensemaking Operational controller Operational

Control

59 min. Investigate sensemaking Data analist Operational

Control semi-structured interview 66 min. Investigate sensemaking Advisor Quantitative Support semi-structured interview 50 min. Investigate sensemaking Manager information provision

Control Mail semi-structured interview

35 min. Investigate sensemaking Controller information

provision

Control Mail semi-structured interview

Control

Field notes and screenshots

60 min. Triangulation of data Controller information Control Mail Field notes and

screenshots

90 min. Triangulation of data Data analist Operational

Control

screenshots - Triangulation of data Table 1: data sources and use

3.3 Data analysis

The interview data was coded by using ATLAS.ti version 8 and the observation notes were coded manually. In this first-order analysis a large number open codes emerged from the data (Strauss and Corbin, 1998). After that the acquired documentation and screenshots were analysed and if possible added to the aforementioned first-order codes. If the material was not applicable to existing codes, new codes were generated. As a result of the large number of codes the desired sense of ‘getting lost’ was achieved (Gioia, 2004). Through a process of ‘axial

(14)

14 coding’ (Strauss and Corbin, 1998) a more manageable number of codes emerged through a

process of categorization and comparison. At the following stage the researcher was looking for deeper structures between the emerged axial codes and came up with several theoretical themes to explain the observed phenomena. These themes were than compared to existing literature to interpret and find meaning for them. Based on the above a data structure was constructed which gives a clear overview of the analysis process of this thesis research. In the next section (Findings) further elaboration and foundation will be given on the emerged codes, themes and dimensions.

(15)

(16)

4. Findings

After having presented the theoretical and methodological parts of the thesis, in this section the emerging results will be presented and explained. The data structure (figure 3) from page 14 emerged from the conducted interviews and observations.

4.1 Strategy implementation

As strategy and budgets are a key input for the various dashboards and reporting models and therefore are supporting controllers to make sense of the results, it seems logical to also firstly unravel the process behind this. Within PostNL strategy implementation takes places through the so-called ‘Strategic Plan’. In the following quote the manager of the ‘Informatievoorziening’ department (manager IV) explains this process:

“This process sometimes happens a few times per year and sometimes once per year, depending on what the CFO wants. The CFO then indicates that we need a new Strategic Plan and that means that we will make an estimate for three years ahead of what the results will be and what the UCOI (Underlying Cash Operating Income) will be. That starts with: what are the volume decreases, which flows and which mix are you going to get, how are your sales going to be, what are you going to do on the costs side and what kind of result does that yield. That result is used by the CFO for the Outlook that we issue. Then there are on the cost side Master Plans processed and that is actually the input for the controller. There is usually a project manager on such a Master Plan, but it is possible that such a Master Plan will not start until next year or in two years and then they will not appoint a project manager immediately. In the end they always appoint a project manager and there is a controller involved to ensure that the calculations we have there are correct and that the savings will stay on track.”

So based on the above the controllers come into the picture when the strategy is translated into Master Plans and therefore becomes somewhat more tactical. Controllers need to make sense of the impact of the forecasted volumes on the cost levels and revenues. The questions about how this process is practically executed and how controllers are dealing with it still remain unanswered. The following quotes will shed some light on this:

“What happens in practice, for example, is that a consultancy firm like McKinsey is hired to think about what we should do differently and where the possibilities are. In this way, a project is defined by the movements that you have to make. Ultimately as you get closer, you will need to be more involved as controllers. So in the first instance when it is still far away and it is still three years ahead, the figures are reasonably 'high-level' and there is no period breakdown. The closer it gets, the more important that planning becomes and the more certainty you must have that the savings will be realised.” (manager IV)

“The Strategic Plan is created and your budget has to meet the Strategic Plan. We build budgets bottom-up together with account managers. We have one on one conversations with account managers. They have customer portfolios where we see X happening and where we have seen Y in the past and then we ask them to give a prediction for the rest of 2018. Then it is our role to 'challenge' them.” (Business controller commercial control)

(17)

17

“When I did the ‘coding rule’ project and the ‘combi bundle’ project I had to give an update every year for the Strategic Plan and also indicate what we expect for the coming years in terms of savings, investments, depreciations and initial costs. That then went to the management of Mail NL and they went to see what we could expect in the coming years.” (Business controller operational control)

As the manager IV explained controllers were involved in the process when the so-called Master Plans had to be developed in cooperation with the to the Master Plan appointed project leader. Together they had to make the business case in an Excel template and submit the files in the ‘CAPEX workflow system’ to get the approval of the all involved internal stakeholders (e.g. Group Reporting, Group Tax, Business Director). In this template controllers had to give a periodic break-down for the cost savings and investments (if applicable). During the observation session with the controller IV he told me about the ‘Project Application PostNL’ (PAP). PAP basically is an Excel database in which controllers were able to maintain their projects and in which they periodically had to put in the realized costs. In PAP the controllers had to create a new project and fill in the required master data regarding the Master Plan. In the case of a revision of an already created project the controllers could also adjust the master data accordingly. After the project has started controllers needed to fill in the actual costs every month in the PAP database. After that an export was uploaded into the ‘Masterplannen report’. This latter report basically gives a more concise confrontation of the budget versus actuals and in this report controllers’ had to explain the realized savings per line item. Two examples of the projects submitted by controllers are the ‘coding rule’ and the ‘combi bundle’ projects. The former entails an innovation of the sorting process where letters are coded by a new type of sorting machine which enables the letters being sorted not by postal codes, but by a new and more efficient coding system. The latter project entails an innovation in the delivery process where postmen were sent off in the streets with multiple formats of letters in one bag or bundle. As mentioned in the quote above the commercial controllers had to match their budgets with the forecasts in the Strategic Plan. They made their budgets through one-on-one sessions with account managers and tried to challenge them through predictions based on historical data. Often predictions from the account managers would not meet historical data and controllers challenged account managers to explain why their predictions were different from the data. One controller also referred to this process as a game:

“True, you always have a discussion in it. Account managers may be bonus driven and are therefore inclined to adjust volumes downwards, while we try to keep those volumes higher. It is a kind of game that we play with the account managers and sales directors.” (Business controller commercial control)

(18)

18 4.2 Use of BA within PostNL and data analytics projects

As previously discussed in the theoretical background BA entails use of data, IT, statistics, mathematics and computerized models in order to gain more insights in operations and enable fact-based decision making. Also there is a distinction between descriptive, predictive and prescriptive BA. During the pilot interview with the data scientist it was pointed out that the use of analytics could be plotted on an ‘analytics maturity curve’ (figure 4) which starts off with basic KPI reporting and Excel models, progresses to root-cause analysis and finally ends up at predictive modelling and simulations. This suits well with the theory as the bottom stands for descriptive BA and the top of the curve stands for predictive BA. According to the data scientist most PostNL departments are currently at the bottom of the curve and the ACE department is at the top. In the following sections we will get deeper into the BA use of the departments which were investigated for this thesis research.

(19)

19 Controlling department

As mentioned the by the data scientist the use of BA is predominantly descriptive and consists of KPI reporting in Excel dashboards. However the current status of the IT infrastructure and the organization of information provision is of such a level that fact-based decision making is supported. For instance the existence of the data warehouse which is connected to the majority of the financial information systems provides a solid ground for fact-based decision making. According to the business controllers it is key to have the facts right, as it is their best weapon for persuading managers to adjust their forecasts. Also due to the work of the ‘Informatievoorziening’ department most of the descriptive BA is automated through the dashboards and quality of information provision. Nevertheless there are still steps to be taken according to the manager of the ‘Informatievoorziening’ department. He argued that controllers, despite the level of automation on descriptive BA, nowadays still are looking too much to the past and they need to advance to looking forward.

Data departments

The data departments ACE (Analytics Centre of Excellence) and KO (Kwanitatieve Ondersteuning/ Quantitative Support) are using more advanced BA methods like root-cause analyses and algorithms. Especially ACE is active in algorithm development, for instance last year they build an algorithm which predicts the timeslot in which customers could expect their parcel to be delivered. They named this solution the Time Slot Indication (TSI) algorithm. The KO department used a lot of applications for statistical modelling, but they also offered solutions for logistical problems like the placement of letterboxes. It is also because of the experience with advanced BA that a department as KO is involved in the development of the PostNL ‘Strategic Plan’.

Data analytics projects

The ACE and KO departments were occupied with projects within PostNL. One of the projects ACE did was the ‘cost of delivery’ project. This projected started because the business in cooperation with control had a need for more insights into the extra or less claimed time by postmen for their deliveries. These were important issues because the delivery process is very costly and the insights would help PostNL in determining the new working time arrangement which would be more realistic. Firstly the team started with identifying the scope of the problem during a scoping session and hypotheses during a hypothesis session with process managers and delivery team leaders. The data for this project was collected from several systems of which

(20)

20 the following two are the most important: Salesforce for data on why extra or less time was

claimed and OWS for the current working time arrangements and the time accounted for. The following quote regards the analysis of the data:

“In order to compare data we wrote a script which coupled different information sets to each other, next to that we had to devise some kind of ‘cookbook’ to run the scripts in the right order.” (senior data scientist ACE)

So in order to efficiently analyse the data scripts were developed by the ACE team and deployed in a structured sequence. This method led the team to some remarkable insights: the most frequently used reason for claiming extra time is ‘more mail’, which is remarkable since postal volumes have been decreasing for the last years.

As a result of the project a new working time arrangement has been established based on four key inputs: the old working time arrangement, the expected mail volumes, specific factors per delivery area and the structural difference between the old working time arrangement and the realized hours. The fourth input is the main contribution of the ACE project and is based on their analysis of the data from the Salesforce system.

The project had certain common ground with the controlling department. Firstly because the results of the project gave better insight into the costs of the delivery process and the newly established working time arrangement was an import input for the cost calculations of the controllers. Secondly the project team depended on input from control as they had a role in it as ‘business representatives’ and therefore controllers needed to provide answers to the data requests of the ACE team.

4.3 Controllers’ sensemaking with BA

The sensemaking process of controllers with BA is the primary interest of this thesis. The resulting model of this process will is presented at the end of this chapter (figure 6). Inside this department there were differences found between commercial control and operational control with regard to the information provision as a result of data quality issues at commercial control and differences in the way the departments are organized. These differences will be explained in the upcoming sections.

(21)

21 Phase 1: Information provision to dashboards.

The sensemaking starts off with the retrieving of both financial and non-financial data from various information systems. The financial data is stored and produced by a SAP ERP package, customer data is held by both Salesforce and SAP CRM packages and operational data (spend hours) from an Ortec planning software solution. However the majority of the data is not retrieved directly from the aforementioned systems, these are retrieved from a central data warehouse. The following quote of the manager IV sheds more light on his personal definition of information provision:

“I see information provision as the reports and models, but actually the data underneath it that we use which is neatly recorded with a HANA structure is also part of it. By having that structure is it easy to subtract data and you don’t have to worry about definitions.”

The reports and models to which the manager IV is referring are reports in Excel in standardized templates. Within the control Mail department they make use of a number of these templates or dashboards. These reports contain financials subtracted from the central data warehouse, which are compared to budgets and actuals of last quarter or last year. The reports are used to analyse the results in light of the budgets and to assess whether the company is on track financially. The analysis will be further explained in the next section of the sensemaking model. The ‘HANA structure’ the manager IV mentioned is a standard lay-out in which the financials are stored within the central data warehouse and because of this standardisation the financials are applicable for many reports and can be imported without much editing. The reports and dashboard are being maintained and developed by IV.

“Then we have the development role at IV. These are actually two things: many controllers have models that they use for analysis or to report or whatever. These are often things that they themselves have made and are inconvenient because they are collecting all kinds of data by themselves. We try to tackle those models and create more beautiful models that are automatically updated. If we determine that these are standard reports that have to be made every month for, for example, the next two years, we will go to the BICC2_{to make BI tooling for this.}

Then we ensure that data is automatically updated and that a kind of web report is created.” (manager IV)

The IV department is responsible for making the information provision more efficient and effective. They keep track of all the reports that are made individually and assess if a standardisation is necessary. If we zoom in on the developing aspect of IV in the following quote of the controller IV he will take us with him in the process of development.

(22)

22

“It usually starts with a request from controllers and so a bit of interviewing precedes. This is to find out which data sources there are, what they want and which calculation rules they want to apply. From there we will decide how we will build it and where. I mainly work in Excel, but it may also be possible to look at another application. In that case it goes to a BICC solution where something can be made in SAP Business Objects (BI application). Such solutions then go via my manager to BICC and together they look to see if they can organize the reporting in this way. An example of this is Integral Steering Model (ISM), in which all kinds of SAP HANA data are extracted on the basis of which comparisons are made with regard to budgets and to which process managers are accountable.” (Controller IV)

During the interviews which the controller IV conducts before building a report he tries to capture the information that controllers seek in order to do their analysis. More specific he looks for the level of refinement that is needed, for instance two years ago a project module was released within the ERP package (SAP Business by Design) and controllers wanted to use this project structure to collect costs, so during that time controllers needed reports which could support this. Therefore he investigated the ways in which it was possible to subtract data from the ERP package and how he could turn this into a report in which controllers could ‘drill down’ on the projects for their analysis. The last quote also mentioned ISM as an example of BI tooling which is developed in cooperation with the Business Intelligence department (BICC). The choice for BI tooling is made if reports are needed frequently for a longer period of time and especially when it serves the needs of a large group of employees within the company. In the case of ISM there was a significant saving of time and resources realized throughout the company. The manager IV further expands on this and explains the tool further:

“We have about 20 delivery areas and 60 pre-sorting areas, then we have 70 process managers and 500 team leaders. The team leader reports daily to his process manager, weekly process manager reports to the regional manager and that also happens during the period end closing. Previously, controllers filled Excel models for this. Then you can imagine that if you have to do that for 70 managers, that is a huge amount of work. Now we have ISM for that, that means integral steering model and that means that the data is automatically uploaded through a tooling that we have made together with BICC. In the new situation a process manager logs in and as soon as he logs in, he gets his own report with his figures and he can put in his explanation for his figures. When his regional manager opens it he sees what that process manager has written to him. This means that this process does not consume any time anymore for controllers, the only thing controllers have to do is change the KPIs at the end of the year. We have not build ISM because of the complexity, but because it takes place on a large scale and it is structural, so we had to automate it. It also takes a lot of time to build such a thing, so that is always a consideration when choosing whether to automate it or not.”

(23)

23 As explained by the manager IV from now on managers had real-time insights into their results

and on explanation to it. In the past situation it would take weeks before the information was provided by the controllers and steering was difficult because the data was not timely enough. Nowadays it is completely actual and managers are supported to correct timely. The corrective actions will be further explained in the third phase of the sensemaking process, first the data retrieval processes will be further explained.

“I am currently doing everything with SAP by Design and SAP BW HANA. Furthermore, OVP (BPC tooling). What I do with it is 'emptying' SAP by Design using my model. In any case, I want to ensure that controllers can access the figures with as few actions as possible and that they are reliable. Controllers should be able to do so by the touch of a button. There is one model that you also know in Groningen3_{and that is the management report.}

Here I get all general ledger data from SAP by Design and 'throw a sauce over it'. Fictitiously I simulate a so-called distribution run into the model so that the controllers timely can see what the total costs per project are. This run deducts all costs from the projects to the cost centre’s.” (controller IV)

“It all goes into Access and that database is linked to that management report and at the moment that this report is opened they have the most up-to-date information from that Access database. The controllers will eventually work with that.” (controller IV)

“At the moment, however, the budgets can only be extracted on general ledger accounts and with Mail NL they budget on projects. So the budget that I use is supplied manually to the Access model. In principle, it is the same template that is uploaded into SAP by Design, but it is enriched with project tasks. That way we have budgets available at project level.”(controller IV)

The retrieval process starts with updating several reports (see figure 5, pink boxes) by making use of an Excel add-in option which establishes a real-time connection between the SAP by Design package and the Excel worksheet. The worksheets need to be updated manually clicking on the ‘refresh’ button. These worksheets are stored at a central network location and are through that location connected to an Access database. The Access database on its turn collects all these new data and uses several standard tables (see figure 5, grey boxes) to put all the costs on the right place. In the final stage the key reports (sees figure 5, green boxes) can be refreshed as they have a direct link to the Acces database. Furthermore some reports in the pink boxes are not connected to the Acces database, this is because these are retrieved from the SAP HANA Business Warehouse (BW) which already has a clear and usable structure. Operational control uses a lot of reports based on these SAP HANA BW data and therefore they are able subtract the data even more efficient. Their retrieval process is done by the aforementioned BICC and uploaded into the ISM. Only the data from the sorting processes is still not that well-structured,

(24)

24 so that means that those reports and dashboards are less easy to use. More about that in phase

2 of the sensemaking process.

Figure 5: visual representation of retrieval process and relations between reports and systems.

The controller IV also mentioned that budgeting data was not available on level of projects as is wanted by controllers. Therefore this data also had to be enriched separately to be able to compare actuals to budgets on projects level. This is done through a separate worksheet which divides the budgeting data over the appropriate project lines.

During this research it was made clear that there were differences in the information provision between commercial control on the one hand and operational and cost control on the other hand. Commercial control were found to have trouble providing the right figures to analyse upon in contrast to the other departments where they could analyse with more ease. What came across during the interview were things like: “the basics are not in order” or “we are devoting too less time on analysing because we are too busy with producing the right figures”. The reasons for this were found in insufficient recording of revenues and sales volumes and a transition phase. The latter refers to the instalment of new cloud-based applications for CRM data and revenue data. Because the revenue accounting process is very important and extensive the company choose to transfer the data bit by bit instead of doing a ‘big bang’ like approach in which all the data is transferred at once and the old systems are shut down immediately. In order to deal with the problems commercial control hired a new employee to help them sorting out the data issues

(25)

25 by making use of an Standard Query Language (SQL) database. This database made it possible

to solve differences in CRM data hierarchies by writing queries.

To summarize the first phase of the sensemaking model (figure 6): this phase starts with the data warehouse, Access and SQL databases to retrieve the data needed to analyse upon. The next step is to let the data flow into the several Excel reports and dashboards. This step differs across the control department as some data can be retrieved directly from SAP HANA BW and other data needs further adjustments along the way. Furthermore the IV department plays an important role in the maintenance and further improvement of the information provision. As introduced before the use of BA is primarily descriptive. This descriptive BA use is portrayed by the Excel dashboards and also the ISM BI-tooling. An important condition is the availability of a proper data infrastructure to support the descriptive BA. At commercial control the BA is less efficient due to data infrastructure issues as mentioned in the former paragraph. If the infrastructure is well-designed the BA can be utilized more effective. In the next paragraph the analysis of controllers will be dealt with.

Phase 2: Analysis and interpretation in cooperation with business.

When all dashboards and Excel models are ‘refreshed’ and prepared, the business controllers would start their first analysis of the results. Also in this process there are differences between commercial control and operational control due to the earlier mentioned issues at commercial control. For instance one of the information systems (OVP; Omzet en Volume Planning/ Revenue and Volume Planning) commercial control used caused issues because of differences in the hierarchical structure of CRM data. The commercial controllers as well as the manager from the ‘Informatievoorziening’ department pointed out that commercial control had to devote too much time on sorting out all the data, which left them with too little time to conduct a thorough analysis on the periodic results. At times they were even glad to be able to present some figures at all.

“I noticed that the basics are not in order. We should devote 80% of our time to discussing the results with managers, but instead we are more occupied with linking tables to each other.” (Controller commercial control)

The ‘linking of tables to each other’ refers to a way to cope with differences in hierarchical structures of customer data in the two CRM packages commercial control were currently using in the transition phase between the old system and the new system.

(26)

26 In their analysis the business controllers were comparing actual periodic results to budgeted

results and results from last year. This was done by examining the Excel models and dashboards in which these numbers were all presented to them. However, the controllers could not do the interpretations all on their own. In order to fully understand the figures they needed the help of the ‘business’ (e.g. sales account managers for commercial control and process managers for operational control). The commercial controller explained it as follows:

“Each period we have a confrontation between what we were expecting (budget) and what the real numbers were (actuals). If there are deviations that we try to explain, then we don’t do this on our own but together with Account management. If we see that it concerns a specific customer, then it is wise to discuss what is going on over there with the account manager assigned to that customer.” (commercial controller)

“There used to be a method (which is currently no longer working) that account management received their overviews in the period end closing. This way they could see exactly what their customers had done. In this overview they had to explain major deviations before Tuesday 5 o'clock. They could then fill it in on a SharePoint site and then we could unlock that data. We no longer have that tooling and we do it manually. We do have to restore this method, so that we can have those statements on Wednesday again.” (commercial controller)

During the observations the Excel templates on which the analyses were done were shown and discussed with the researcher. In these template there was a comparison on customer level between the actual sales in that period, the budgeted sales in that period and the actual sales that period the year before. Not all deviations were analysed upon, the controllers focused on largest deviations and needed the explanation of account managers to know why these were happening. The most occurring reasons for the deviations were things like: digitization, phasing or a move to a competing postal company like Sandd. The choice of what to analyse and explain is supported by the following quote (please note that ‘GO’ means ‘gemiddelde opbrengst’ or average turnover):

“You are not going to explain every ten thousand euros and you look at a higher level. In the period everything can happen, but you also have to look 'year-to-date' so that you discover certain phasing and the deviation turns out to be not too bad. You are really looking at big deviations; where is the volume disappointing and where is the GO disappointing. You are not going to explain every deviation.” (commercial controller)

For operational control the analysis consisted of viewing changes in mail volumes and occupancy rates, whereas commercial control focused on sales volumes. The occupancy rates relate to the deployment of staff. A term that comes across in their dashboard is ‘AC’, which stands for ‘arbeidscapaciteit’ or labour capacity. It concerns the number of hours that have been deployed to keep the preparation, sorting and delivery processes running. These had to be compared to mail volumes because if volumes are decreasing, than so should the AC rates.

(27)

27 Furthermore operational control also had to report on strategic cost saving projects on the basis

of the results shown in the dashboards. Due to the advancement in the information provision business controllers did not need to alter or enrich the numbers in order to start analysing most of the time. However, not all the information provision within the operational control team was of the same level.

“For sorting it is something else. Preparing and delivering was always equal to each other and sorting was another branch of sport. The provision of information with sorting is also different from preparation and delivery. This means that we do not have a dashboard at sorting, like the one at preparation and delivery. We do have a dashboard, but that is not as extensive as the one for preparing and delivering”. (Controller operational control)

The latter quote states that there still are improvements to be made for operational control. The controllers in charge of sorting processes were found to be struggling with their dashboards and this caused them to have analyses of inferior quality in relation to preparation and delivering. The altering or enrichment of numbers was only needed in case of special information needs specific to the business area to be controlled, for instance an operational controller also wanted to benchmark an efficiency indicator to last year in order to look at the results from more angles.

“I always calculate some efficiency KPI’s through which I make comparisons with last year. There are different angles: you can compare to budget, you can compare with last year and you can look at trends. If you only focus on budget it does not have to say that things are going the right way. There are more causes for good or bad results, so therefore you need more angles to be able to make a good judgement.” (Controller operational control)

After having conducted the first analysis the controllers highlighted specific deviations to be deeper analysed on the basis of materiality, year-to-date analysis (in order to determine if it is due to phasing) and their knowledge of critical business processes. Operational control would then firstly discuss these deviations with a regional manager in order to prepare themselves for the weekly conference call with the process managers. For this preparation the controllers use the dashboard to be able to look deeper into the deviations and particularities. Also the controller and the regional manager would have a look at the ISM model to see what explanations the process managers already had filled in. This way the controller and regional manager could choose explanations which were unclear or questionable and if they needed to ask more about them. The choice for which deviations to zoom in on is explained by the following quote:

“You look at the extent of the deviation, the biggest deviations are in the process. If you look at preparation you see that the biggest deviations are in the preparation of orders and pre-sorting. Then you then zoom in on it to see what is going on. Sometimes they are incidents and have to do with supply, poor production or extra hours. But

(28)

28

there is also that there is a certain structure in the extra deployment of staffing and then you have to wonder whether the process on the work floor is well organized. Do the employees do what they need to do and do the team coaches support that? You look at things like that.” (Controller operational control)

So to summarize the second phase in the sensemaking model (figure 6) this starts with the controllers examining the dashboards and choosing the deviations they want to know more about. After having done that they would get more information about the deviations by asking account managers, regional managers and process managers about these particularities. The BA use provides the controllers with more time to analyse and interpret together with the business instead of sorting out the figure first. Although there are still differences which made the interpretation more difficult for commercial control then for operational control.

Phase 3: Reporting to senior management and adjustment initiatives

After having further analysed the highlighted deviations the business controllers had to present and discuss their findings with the senior management. In the case of operational control this happened through a weekly conference calls in which the controller would participate together with his regional manager and a few process managers. The following quotes will explain these conference calls further:

“Such a conference call takes half an hour and there are about eight managers participating. We mostly discuss three main items: we look at the costs we have, we discuss quality and we discuss complaints from customers. It depends on what happened in that week. If the costs are in line with the budget and there are no crazy things involved, but we have a number of misses in the delivery period, we will focus on that. Or if the number of complaints is very high, we will focus on that. It is therefore also just what we see in the results of that week. So in the preparation discussion (which I have with the regional manager beforehand) we determine what we are talking about and then, for example, we see something odd at a number of preparation locations and then we also ask some additional questions about this. It can be overall that the quality is poor and then you zoom in and ask everyone what the possible actions are and then it may be that a single location stands out in terms of costs and then you can still zoom in. It is not exactly the same every week. It depends on what is in the results and where is the reason to pay attention to at that moment.” (controller operational control)

“During the call we log on into ISM and then we can select our region and preparation cluster and look at the figures as well as the textual justification of the managers.” (controller operational control)

So there are subjects to be discussed but it depends on the week on which of the subject is put the most emphasis on. Also nowadays ISM is found to be a helpful guiding instrument during these calls. The systems gives the controller and regional manager input for their talking points in their preparation phase (see also phase 2 of the sensemaking model) and stimulates to ask the right questions to get deeper to the core problem. However the controller also pointed out a

(29)

29 disadvantage of ISM in comparison to the old method by which controllers had to maintain a

so-called SV (Sturen en Verantwoorden/ Steering and Justification) format Excel report:

“Well, ISM gives less space to fill in explanatory notes. In any case, a manager must briefly and concisely fill in this tooling and then it mostly happens that you have to ask some more explicit questions.” (controller operational control)

“…they could actually do more textually there. If you talk about the work package, you could break it down into the main process preparation, premiums and things like that. Now it only says work package together with the deviation and you have to write your explanation. You really have to clearly state where the deviations are.” (controller operational control)

The main disadvantage of ISM are the constraints with regard to what managers could fill in to explain the deviations. This frustrated the analysis because controllers had to ask more questions in order to understand were the roots of the problems were. Next to the discussion about the reasons behind deviations the controller would also discuss improvements and adjustments in the processes to tackle the deviations. To control and monitor these they would discuss the actions list:

“Certainly there is an action list on which the action points are named and put on. These points are named in such a way that it is actually an action. That you do not put in for example discuss blablabla, if you do that it is not concrete enough. Then we make it: feedback result talk blablabla, then you are much more concrete and clearer about what you expect from each other. This action list is updated every week in the call. You try to close that control loop as much as possible.” (controller operational control)

Sometimes this action list would contain long-time projects to achieve certain results. In that case the management team would check the financial outcomes of the project with the forecasted result. Once every month the weekly conference call would also be used to discuss the previous month or period.

“So there is a weekly call, but once a month that weekly call is used to discuss the periodic figures. If you are going to report on the last week of the period, you also look at the entire period, because in that period there are also a few KPI's that are not in the weekly KPI's.” (controller operational control)

“Especially in the field of personnel, inflow or outflow that kind of things. You look at it once a month and you get that input once a month. Every week you talk mostly about AC and volumes that are deployed and at the end of the month you have a few KPI’s in which you report extra about, including intake and outflow. But you also look back over the month and about the total that happens in those weeks.” (controller operational control)

“Every week in ISM you have a weekly dashboard and once per month there is a periodic dashboard in which you have to justify the entire period.” (controller operational control)

(30)

30 The discussions and presentation with senior management of commercial control would unfold

similarly. Firstly the controller would talk with sales directors individually:

“What I always do when I speak with sales directors during the period (because they are always interested, one more than the other): we would examine the customer package and then he or she also sees immediately the real-time situation of how they performed and on what customers. We usually do this on Monday or Tuesday when the figures are known. They then would call me to make an appointment to examine the figures. Then we look at it for half an hour to an hour to see the big lines, customer package x does so, where are we now and are we still a bit in line with the goals. That actually happens without any slides or explanations to see where they stand. We usually log into the system together and look at an Excel tool I made which is connected to the order-to-cash database.” (controller commercial control)

Next to the individual session the controller would participate in the monthly work meeting of the sales directors. According to the commercial controller these are informal sessions:

“You should not think that it is very formal. I then come in and we sit down. Often I sent the presentation beforehand by email, so if the subject is mentioned on the agenda, the person who is logged in opens the slides and then I walk them through it. I explain to them again what I have already said in the slides, maybe I will tell them a bit more, but then we just stop and think about the movements. What we do then is to look at the period and also try to look more and more at 'year-to-date'. So from looking at only the period, which is quite static, we try to ‘watch the entire film’ and how is it 'year-to-date'. By doing so, we immediately remove all phasing effects. These are then less relevant.” (controller commercial control)

According to the commercial controller the purpose of these meetings are to inform the management, but he also experienced that these session would result in more background information about specific customers like contract negotiations. The presentation which the controller presented consist of waterfall chart and the textual explanation about it.

Next to the reporting to sales directors or regional managers the controller also had to report to the Planning and Control (P&C) department. The difference with the aforementioned reporting to the business is the level of detail. For the latter reporting information had to be consolidated and on a higher hierarchical level. The following quote explains more about the reporting structure which goes eventually up to the Board of Management.

“P&C makes the formal story towards Group Control and they ultimately make a set for the Executive Board and the Executive Committee. They have a call on Monday and later in the week there is the 'Monthly Finance Meeting' and P&C provides input. For on the one hand Group Control and on the other hand the Monthly Finance Meeting in which all financial directors meet. The figures are also shared there.” (controller commercial control)

(31)

31 So this last phase of the sensemaking model (figure 6) entails the discussing and presenting of

periodic results with senior management. During this phase further interpretation on the figures are given through a collective sensemaking. The BA use one the hand provides for conversation material through pointing out where more interpretation is needed, but on the other hand parts of the BA are actually making the sensemaking more complicated due to restrictions of the system. However overall the BA is beneficial and time saving and therefore cost saving for the controllers’ sensemaking.

Business Analytics driven management accounting