Building management skills to benefit from Data Analytics

How bol.com builds a 100% fact-based organization

 

Bol.com pioneered in shaping the Dutch online market and has become one of the strongest retail brands. Bol.com’s leading position has been fueled by the powerful capabilities of the Data Analytics teams, built up by Jens Waaijers and developed in partnership with MIcompany. Over the last years, new analytics capabilities have resulted in the ability to better service both the customer and the business teams. However, the strong growth in categories and category teams has led to fragmentation and different ways of working, as bol.com customer interaction is managed with some degree of freedom for the teams themselves. To address this challenge, Jens developed a vision on becoming 100% fact-based within the entire organization. In order to support bol.com in reaching this goal, MIcompany was asked to tailor its’ existing commercial analytics foundation program to the bol.com needs. The resulting Bol.com Analytical Masterclass (BAM) was rolled out to all category teams after a successful pilot to create more impact with Data Analytics. In this article we describe the approach, program set up, results and our reflections.

 

The challenge – how to develop data driven business teams

Bol.com is the leading Dutch online shop for books, toys, but also categories as Health & Wellness and much more, selling over 13 million product. According to research published in 2015, bol.com is the strongest retail brand with the highest expected growth in the Netherlands1. As a pure online player, bol.com has a strong Data Analytics focus and has spent much of its’ early efforts on strengthening its’ Data Analytics capabilities and tooling. Together with MIcompany, the analysts have been developed in the 3-year MIacademy talent program, resulting in a very strong analytical and technical pool of people equipped to deliver extraordinary results.

With the ambition to make all decisions within the company based on facts by 2018 however, bol.com faced the challenge of teaching its’ business teams how to navigate the Data Analytics road. The business teams are growing rapidly in size and number, and many of bol.com new recruits are from different backgrounds. With only two years left to reach the ambitious goal of being one hundred percent fact-based, bol.com’s business teams were facing four key challenges:

  • Lack of a common problem-solving approach. There was no uniform problem-solving approach within bol.com. Proposed initiatives by the business teams often didn’t involve a proper problem-solving loop. As a result, the teams weren’t asking the right questions to be solved. Which resulted in ineffective use of the analyst teams.
  • Difficulty in priority setting. The fast-growing online businesses of bol.com and expanding product ranges and categories offer an enormous array of opportunities for the business teams. The possibilities for new initiatives seemed endless, making it extremely difficult to choose where to focus efforts. Also, evaluations of initiatives were not done, or not done correctly, making it nearly impossible to benefit from a learning loop to improve actions. Business teams needed to master the ability of being able to determine which levers drive their category growth. Not by ‘feeling’ that this is the case, but through the use of data-driven hypotheses.
  • Business plans with insufficient fact-based foundation. The bol.com category business plans varied strongly in quality and generally were insufficiently driven by facts. In part this resulted in inaccurate forecasting of category KPI’s in the planning process. Category key priorities and supporting activities and measures were not driven by fact-based decision making.
  • Varying quality of data driven communication. Business teams spent quite some time preparing presentations for internal usage, but also for meetings with suppliers and other partners. Most of this time was generally spent on visualization instead of presentation structure and message, resulting in a strong variance in quality of output. Moreover, visualization standards were lacking, resulting in sometimes incomprehensible data driven graphics or no graphics at all.

In essence, business teams were missing a uniform, effective, data-driven approach to creating and maintaining value for bol.com and its’ customers.

 

The approach – creating a Data Analytics training for the bol.com business teams

To address these challenges, bol.com and MIcompany partnered up and developed a training program for the bol.com business teams to teach them how to become better problem solvers and work more fact-based. Central in the program is the focus on impact using the problem-solving cycle. Teams are taught how to work in an hypothesis driven way. Simply put: to observe, gather facts, formulate a hypothesis, analyze the data, test the hypothesis, improve if need be, and then embed the improvements. In all steps of the process, facts, data, and insights are used to make decisions. For this, the teams need to master several analytical topics important within the cycle. These topics range from data usage and data structures, to basic analytics and statistics, customer value management, return on investment and forecasting.

We tailored our existing Data Analytics foundation program for the bol.com business teams in such a way that it would teach them the above. It was therefore set up around the bol.com problem solving loop. The program consisted of the 4 parts described below and shown schematically in figure 1:

  1. Driving hypotheses through structured opportunity finding. In order to determine which growth levers are most important, possibilities (hypotheses) are formulated and sized using facts in a structured process. Also, the customer value metric is introduced as well as the new insights this metric can deliver in addition to sales and margin metrics. Hence the impact of steering on customer value is shown.
  2. Understanding required analyses and modeling. The specific hypotheses which seem to have the most potential can then be tested by analyzing the data. The steps in the analysis process are taught and applied using basic statistics and modeling to interpret results. The resulting insights are then used by the business teams to develop initiatives, test them, evaluate them and improve them accordingly.
  3. Mastering applied methodologies of forecasting and ROI. Two important methodologies are mastered in the program. First, forecasting techniques are shown and applied to year-on-year (budgeting) forecasting, and more advanced driver-based forecasting cases are covered. Secondly, a fundamental methodology to determine return of investment (ROI) of different activities is taught and used in the case examples of the business teams.
  4. Presenting with impact. The business teams are trained to develop a structured storyline and communication using the pyramid principle. The ability to distill the main message and supporting fact-based arguments is put into practice extensively. Guidelines, best practices and do and don’ts of data driven visualization are covered thoroughly in order to ensure impactful presentation of insights, hypotheses, results and conclusions.

 

 

Figure 1. Bol.com program set up following the problem-solving loop.

 

 

In essence, the program teaches business teams how to ask the right question and how to use and present the resulting answer. Right questions being those that are based on facts and geared towards the biggest impact. And based on an underlying hypothesis which is tested with data. Essential in asking the right questions is having a basic understanding of Data Analytics, and by doing so, decreasing the gap between the analyst and the business team. But how do we best realize real behavioral change within the business teams, while they differ significantly in experience, back-ground and problem-solving approach?

Bol.com and MIcompany together developed six bol.com guidelines for the setup of the program. The guidelines are described below and the resulting program set up is depicted in figure 2.

  1. Test and roll-out. As is common practice in managing continuous improvement we first tested the program for one business team in 2015, further improved the program based on learnings and applied it for 100+ team members from all 18 business teams and all supporting specialist teams in 2016.
  2. Apply learnings to develop tangible outcome. We set up a program consisting of five modules of one and a halve day (see figure 2), that led the teams through the problem-solving loop. In order to link the applied learnings to tangible output for the teams themselves, the outcome of the problem-solving loop was a fact-based presentation of the business team category plan attended by bol.com senior and executive managers.
  3. Link program to key business topics. In our joint view, impact of the program is further increased when the problem-solving loop steps and analytical methodologies are linked to key business topics. We therefore linked the different modules to the key topics of budget planning, customer flows, traffic, promotions and merchandising and sourcing respectively, see figure 2.
  4. Use own cases prepared with own data. During the modules, many data cases were integrated in the program to master the content. As relevance improves when teams can relate to their own business and as the business teams differ largely in lifecycle and business dynamics, we tailored these cases for all teams. As a result, category specific data was used for all cases and outcomes could be applied to their own businesses directly.
  5. Leverage bol.com methodologies and tooling. We wanted to align the learnings as much as possible to the bol.com way of working and available business information, as this would allow the teams to better anchor the fact-based way of working. To this end, the program incorporates the bol.com methodology of continuous improvement through structured learning and uses the performance indicator dashboards provided for by the retail service teams. We invited bol.com specialists to teach the teams about the performance dashboards and models used.
  6. Bring in specialist team expertise. In the pilot we learned that bringing in specialist roles together with business teams helped to create better mutual understanding. We therefore invited supply chain, online marketing, digital flow, Plaza specialists and controllers to join the BAM program.

 

 

Figure 2. The key players for Big Data success

 

 

The result – becoming more effective business teams.

A successful pilot was conducted in 2015 with the Health and Beauty business team. In a top down estimate, we quantified overall benefits to easily reach a tenfold return on investment of the BAM program. The key potential benefits identified were: an improved overall purchase margin due to either direct improvements on buying conditions, increased promotion effectiveness, and reduced stock supply costs by better forecasting of promotions and sales at a more granular level. As a result, the program was rolled-out this spring to all 18 categories. We started with 6 classes that each consisted of on average three business teams that were completed with specialists and the Belgium market managers. To manage scope, business team attendance was limited to the category manager, buyers and other senior roles when available. We planned for three to four weeks in between the modules to allow for application of the program content to daily business (through assignments), preparation of the modules, and for the final presentation of the category plan. The question of course is now: to what extend did the business teams and specialists that followed the BAM program improve their fact-based way of working? And how did the BAM program address the challenges for bol.com mentioned in this article?

After the successful pilot of the Health & Beauty team we saw strong improvements in many areas that were sustained and strengthened in the year thereafter and that drove the quantified benefits mentioned above. For the other business teams we have seen promising first improvements during the program. The five key areas of improvement that we have seen, and the teams have observed themselves, are described below.

 

 

“Together with MIcompany we have set up an impactful program that changes business teams’ behaviors. The roll-out has led the business teams to better leverage the work done by our analysts and retail services teams. Furthermore it has led to better use of the bol.com tools and available insights.”

Jens Leendert Waaijers, Director Shopping Experience bol.com

 

 

First, and arguably most important, the teams have adopted a common way of working that is applied to many aspects of their own work and in the interaction with other teams at bol.com. By working more fact-based, teams have started to ask other (and better) questions to the specialist teams and analysts. Often these are related to tracking customer behavior in time and at a more granular level. Other questions are related to better understanding the underlying drivers of customer behavior. This year, bol.com introduced the continuous learning loop through structured learning as the bol.com problem solving way of working. The methodology was practiced during the BAM program and will help to further anchor fact-based problem solving.

Secondly, the business teams show a better understanding of their own business and use that to better target their customers. More specifically, teams better understand who their customers are, what underlying customer flows are important to their category, and which customers drive category growth. As a result, business teams often refocused their efforts towards attracting existing bol.com customers already buying in other (related) categories. Recommendation bundles have become more relevant and are based on buying patterns of specific groups. Also, email campaigns are better tailored to address the relevant customer groups as a result of better alignment with marketing.

Third, teams have become more effective in setting priorities. Creating a structured overview of possible initiatives and applying a first, fact-based, top-down sizing to prioritize them has shown to be very effective in many ways and has made the teams more efficient and effective overall. The pilot team applied the sizing learning in the scrum process for business prioritization and became much more effective in selecting the right promotions first, before involving marketing. During the roll-out, teams were able to better select what content should be added with the highest priority, based on outcomes of BAM cases.

Fourth, the business teams have become more self-sufficient. Supported by new bol.com KPI dashboards and improved problem-solving capabilities, the teams can now obtain first answers to their questions themselves. This was illustrated by strongly increased usage of the dashboards compared to the team members who are not yet trained. Also, teams are asking the right questions to the supporting teams, thereby avoiding unnecessary workload in, for example, analyses not used or that need to be redone, freeing up analyst time for more complicated analyses. A good example in the pilot phase was the hugely improved forecast of Christmas perfume sales based on close cooperation between buyers and supply chain specialists in the business teams themselves.

 

 

“Building a common fact-based way of working and involving all category and specialist teams really realizes impact at scale. I have experienced the teams stepping up and becoming more effective in priority setting and decision making. I am looking forward to further improve problem solving and drive impact with the teams.”

Oscar Hundman, Unit manager Lifestyle bol.com

 

 

And finally, communication and presentations improved significantly. Teams have learnt to “better bring across the right message”, both internally to stakeholders and externally. For instance, category plans have improved in structure, message and visualization. But also presentations for quarterly meetings with suppliers have become more effective and better visualized. Resulting in meetings and presentations with more impact and better results. Using the data driven story lining and visualization guidelines have also been highly appreciated by the teams themselves and have been applied extensively in their daily business.

 

 

Figure 3. Bol.com Analytical Master class program set up.

 

 

Our reflections on developing business teams

So, having significantly improved upon the bol.com challenge to make the business teams work more fact-based as part of the ambition to become a 100% fact-based company, what are our reflections? We conclude that a training program for business teams at large scale can be very effective to adopt fact-based problem solving within the whole organization. We see that successful use of Big Data within any organization asks not only for a strong analytical core and powerful tooling, but also for empowerment of the business teams within the organization. To get the maximum result out of your data, you need several key-players to adopt a similar way of working: the board, your business, and your analysts. Each of the three have their own role to play. The scope of the board should be such that they determine where the difference can be made and what should be the resulting growth. Business teams must be able to determine which levers will drive this growth. Not by ‘feeling’ that this is the case, but through the use of data-driven hypotheses. And finally, the analysts should be able to provide support on these hypotheses and can test them with data. (see figure three). Leading the business teams through the problem-solving loop by mastering their own cases with their own data and working on a tangible output for the categories can be a very powerful means to achieve the above.

 

 

“Bol.com is pushing the needle by building Data & Analytics capabilities in the entire organization. It’s great to be a partner in a journey that make all pieces fall together. Experiencing business teams and specialists becoming more effective is very rewarding”

Roland Tabor, Partner of MIcompany

 

 

Furthermore, it is important to closely monitor results and adapt when required. We carefully prepared for the roll-out program, team set up and planning. What is most important to achieve? In the roll-out it was shown to be crucial to balance the learning of new capabilities through teaching frameworks and by providing tooling that can be used the next day. During the program we adopted this to reach the best balance per module for the teams. Category leaders were involved to give practical input and further build engagement with the teams themselves. Is it important to note that people behaviors do not change overnight. Therefore, spreading the BAM program over a period of 4-6 months helped to digest on and practice with the new frameworks learned. In this setup, significant impact can be realized in a relatively short training program.

Third, learning can be very powerful when things come together. This is the case when the analyst and retail service teams are involved in preparations and training, when bol.com models and tooling are ready to be used during training and afterwards, and capacity is made available within all the teams involved. Under these conditions, learning is most powerful, as individuals can practice, make mistakes and improve in fact-based decision making using real data and real-life cases. In a roll-out of such a scale as the BAM program, this requires careful orchestration within bol.com and MIcompany. Therefore, organization wide development should be closely linked to the Data & Analytics road map.

Concluding, the BAM roll-out showed how through further standardization of tooling, deep-dive analyses and frameworks it is possible to support the business teams in becoming effective in fact-based working. This will also free up analyst time and will stimulate further innovation. In fact, fact-based business teams will drive the next wave of continuous improvements within bol.com. With this, bol.com once again shows itself to be a pioneer in building Data & Analytics capabilities within the entire organization.

 

Sources

  1. ‘Dit zijn de grote drie van de toekomst – Retail watching onderzoek 2015
Roland Tabor
Partner

MIcompany

Geertje Zeegers
Junior Partner

MIcompany

Marnix Bügel
Managing Partner

MIcompany

Jens Leendert Waaijers
Director Shopping Experience

bol.com

AI quick scan

Test how your company is performing on developing and implementing AI solutions and platform capabilities:

 

1.My organisation communicates a clear end-state vision on how AI can transform our business
2.AI initiatives are aligned with the strategic goals of my organisation
3.My organisation has a roadmap that focuses data analytics efforts on large scale business opportunities
4.My organisation succesfully builds high potential proof of concepts with AI
5.AI use cases are developed end-to-end from data to algorithm to an application, such that users can interact with the algorithm
6.AI use cases are automated and operationalized in a production environment
7.AI use cases are fully adopted by the business, without unintended retreat to own judgement
8.AI use cases change the business process fundamentally if adopted
9.AI use case MVP's are continuously being improved, leaving MVP state
10.My organisation succesfully scales impact of AI use cases
11.There is an infrastructure in place for users to approach (a selection of) data
12.There is a possibility to develop models, for example in sandbox like environments
13.Algorithms are operationalized and automatically predict on new data
14.Your organisation has deployed one ore more experiments with modern, for example cloud-based, infrastructure
15.AI platform standards are set and managed
16.Clarity exists on target infrastructure for new solutions
17.Existence of legacy is accepted as a given, but migrations and dependencies are managed with a impact driven mindset
18.AI solutions are in production on scalable modern infrastructure
19.There are standardized operational processes to develop, deploy and maintain models (model management) in modern infrastructure.
20.Users enjoy flexibility to access any data and use it to build solutions with diverse requirements, while still maintaining certain standards managed centrally