How do companies develop a data strategy?

Companies and entrepreneurs are facing more challenges than ever before. Driven by growing competitive pressure and declining customer loyalty, new, data-driven business models are often being set up. From a purely technical perspective, data has never been as easy to collect and process as it is today. As a result, more and more internal and external company data is being generated. Companies can use this data to identify undesirable developments earlier than before or to adapt corporate planning to market changes more quickly. The results of methodical data analyses thus become an instrument for achieving business goals. And they make it possible to objectively determine business success.

However, even if data collection and processing make decisions without bias possible, they do not relieve decision-makers of responsibility. On the contrary: only access to a broad, differentiated and consistent database makes data-based decisions possible for which responsibility can be taken.

In order to gain these advantages from the data that is already available or still to be developed, companies must develop a data strategy. Such a strategy should implement the goals of data governance:

• Determine which data is tapped and used and how
• Determine who is responsible for the consistency of data and metrics in which areas.

However, experience shows that the transformation into a data-driven company is full of pitfalls and most data strategies fail. The approach, goals and implementation should therefore be precisely defined and planned.

In the following, I outline seven steps to a successful data strategy.

1. Determine the organization's maturity level in dealing with business intelligence.

To develop a tailor-made data strategy, the actual state in the company is first determined. To increase assertiveness, it is advisable to set up a steering committee at this point. The committee should represent the circle of clients at the highest level and subsequently act as a decision-making body. It should be able to meet regularly at relevant milestones and on an ad hoc basis to resolve escalating conflicts.

The client group names the key stakeholders for each business area - preferably divided into management (as candidates for data ownership) and specialists (for the subsequent assumption of data stewardship). In various workshops and interviews with the named stakeholders, the situation is examined from the following perspectives to record the current situation:

• Organization (Corporate-strategy, Processes),
• data (availability, management),
• Control (Governance, Ownership),
• as well as IT (tools, architecture, reporting).

The collected results are consolidated and presented to the client group. The target state is determined for each perspective based on the current state. The measures to achieve the target state in the context of the data strategy are described and further prioritized by the client group.

2. Analyze the information requirements.

The next step in developing a data strategy is to analyze the information requirements: For the most important measures from the maturity level measurement, the associated information requirements are determined and described. Here, it has proven useful - also in terms of an agile understanding of delivery - to translate the needs or requirements into products to be delivered.

Common delivery product categories are:

• Reporting (reports, dashboards)
• Planning (Processes, Logics)
• Data (source systems, data transformations, Key figures)
• Infrastructure (BI tools, databases).

In practice, most measures will trigger activities from several delivery product categories that need to be processed and managed together. For example, a required dashboard that is based on sources to be developed for the first time will require activities from the Data product area before the new dashboard can be created and delivered.

An important measure within the information requirements analysis is also the creation of a dimensional, logical information model. This involves grouping the information requirements

• according to original data or data sources (e.g. Turnover from financial accounting, number of active customers from CRM system),
• according to derived, calculated data (e.g. average turnover per active customer),
• by evaluation dimensions with hierarchies (e.g. month → quarter → year → customer). Month → quarter → year, customer → customer subgroup → customer supergroup, product → product group → product family, etc.).

Original and derived data share some analysis dimensions so that they can be displayed together in a meaningful way.

3. Design an improved BI organization.

In many cases, the design of an improved BI organization is also part of the development of a data strategy. This is because companies often have silo-like distributed responsibilities for decision-relevant information. It is not uncommon for top management to lack confidence in the information. The origin and validity of the information is questioned instead of making informed decisions based on generally accepted data.

A Business Intelligence Competency Center (BICC) is a common organizational structure for dealing with business intelligence processes. Ideally staffed with interdisciplinary specialist and IT experts, this is where the company's information requirements are met. Depending on the company's level of maturity and BI strategy, the BICC can be a virtual or permanent organizational unit. In the target picture, a fixed core of BICC personnel is recommended, which is then supplemented by temporary resources when processing information requirements. Naturally, company resources for processing information requirements are limited. In particular, staffing is usually a bottleneck factor. Therefore, suitable procedures and processes are also required to concentrate valuable resources on the most important requirements.

In practice, it has proven to be a good idea to set up a transparent demand and delivery management process. The demand-and-delivery management process enables the structured recording and implementation of BI requirements from the specialist departments.

The demand-and-delivery process

This process essentially consists of five steps:

1. BI requirements are reported by the specialist departments. Initially, only the most essential details are to be described in a less formalized manner.
2. In a pre-check, BICC examines the basic feasibility, the implementation effort is determined in the form of a size class (e.g. T-shirt sizes), and the implementation costs are calculated.
3. The demand management committee prioritizes the BI requirements to be processed. This committee is made up of interdisciplinary representatives of the specialist departments and BICC representatives who are both knowledgeable and authorized to make decisions. It meets regularly, e.g. every two weeks, and in the case of agile processing of BI requirements also in the specified Scrum sprint rhythm.
4. From the list of prioritized BI requirements, the top requirement is processed further. The requirement must now be described in such technical detail that a technical specification and implementation can take place after further checks for compliance and data protection.
5. The deliverable is created on the basis of the specification and made available to the requester for testing and acceptance.

4. Establish data governance as a regulatory framework for the use of information.

The next step in developing a data strategy is to establish a regulatory framework for the use of company data and information. This framework defines the responsibilities for handling decision-relevant information. Such data governance must not be missing when introducing a suitable process and organizational structure. In business practice, there are many examples of uncontrolled growth of unstructured, competing and costly BI island solutions. These often contain contradictory information on the same facts. This undermines trust in the data and makes it difficult for top management to make data-driven decisions.

In order to avoid island solutions with contradictory information as far as possible, "house rules" must be drawn up to regulate the development of data sources and the use of information. After all, the interaction between management, specialist departments and IT in the context of business intelligence can very quickly become very complex and thus just as quickly destroy the benefits of the collected data in decision-making.

The primary goal of data governance is to make transparent which data is accessed and used and how, and who is responsible for the consistency of the data and key figures in each area.

Setting up data governance

The fact that the departmental stakeholders play a key role in setting up data governance and therefore the data strategy has already been mentioned in step 1 above.

In contrast, the following steps should be dealt with in mixed working groups:

1. First, technical data domains are formed that thematically bundle key figures and analysis dimensions. Care must be taken here to ensure that IT specifics such as data sources or application systems are not considered.
2. Each data domain is then clearly assigned to a department. These departments appoint a business owner with decision-making authority (e.g. department head or "champion") who is responsible for the data domains assigned to their department.
3. Each business owner appoints ONE business steward and their representative for each data domain. The business steward is responsible for defining key figures and their evaluation dimensions in their domain.
4. A key figure can become obsolete or its calculation can change. The company can also reorganize and thus change the perspective in which the KPI is interpreted. And finally, the competitive environment can change and make new indicators necessary: In all these cases, the process just described would have to be re-initiated.

Ergo: Data governance is an ongoing process.

5. Implement the most important processes towards an improved BI organization.

There are three additional processes to initiate and monitor in order to improve the BI organization when implementing a new data strategy:

• Documentation of BI requirements from the business departments
• Budgeting of BICC-Activities
• Employee development

Documentation of BI requirements

The fact that the BICC supports the specialist departments in formulating their requirements, is crucial for the success of the data strategy after the initial introduction of the demand-and-delivery management process. BI requirements should be documented in two stages:

1. The "rough specification" stage initially only includes the mandatory components that are required for a pre-check.
2. Only when the pre-check has been successfully completed and the requirement has been prioritized is it specified in detail. Here, too, it is important that requesters and implementers regularly exchange information.

If no suitable project management standard has yet been established in the company, the BI demand and delivery management process would also be an opportunity to implement an agile approach (e.g. according to Scrum). .

Budgeting BICC activities

Another key process is the budgeting of BICC activities. After all, how quickly BI requirements are processed depends on what resources (people, material resources, money) are available for this. It has proven useful to inform the management or the client group regularly (e.g. quarterly) about the task backlog, the deliveries made, their benefits and the associated costs. These meetings should also be used to make important decisions. These include major investments, budget changes or even disagreements regarding the prioritization of tasks.

Employee development process

On the journey towards improving corporate management, new knowledge is required - on business contexts, new software and also on new forms of collaboration or project management. This is why it is also very important to keep an eye on employee development when improving the BI organization and implementing a data strategy. This applies to the actual employees of BICC, but also to the specialist colleagues working with them and to the users of the delivery items created.

Of course, this costs money. The effort for further education and training must be planned, both in terms of capacity and money.

6. Design a balanced KPI system.

Many companies have KPI systems with similar weaknesses:

1. On the one hand, they consist of many different control variables and key figures - so many that decision-makers lose track.
2. On the other hand, different areas of a company often report indicators on similar issues that have completely different metrics.
3. And finally, KPIs are often biased towards hard, but late factors such as turnover and contribution margins. However, these key figures usually only allow an ex-post view. So if the month with the poor turnover has already passed, nothing can be done about it. And then it is no longer of any use to the company if the reported deviation from plan is 100 percent correct.

In order to design a balanced KPI system, the key control variables of the respective data domains and specialist areas must be identified. Ideally, these should already result from the BI governance activities described above. In any case, it is important to develop a mixed KPI system that contains few hard and soft as well as early and late key figures. Proven approaches and methods such as going through a balanced scorecard cycle are suitable for this. In this way, suitable performance indicators can be derived from the business objectives and the corporate vision and, for example, the dimensions finances, customers, business process and learning/innovation can be analyzed.

These indicators can ultimately be linked together via a chain of causes and effects. The result is a balanced system of relevant metrics.

7. Design and implement a new planning and forecasting process.

Imagine for a moment that you drive your car by looking only in the rear-view mirror, with the windshield opaque. Can that work?

This is why companies strive for sophisticated planning processes. Sales, turnover and costs are planned in detail for the planning period in painstaking and time-consuming detail. Often, a counter-current procedure is used here, with which the top-down specification of the company management is to be confirmed by lengthy bottom-up planning by sales controllers and cost center managers.

Unfortunately, methods of this kind are in no way change-friendly:

• Gigantic Excel spreadsheets with multiple jump references open the door to input errors.
• If the market environment then requires a change in the product and/or customer structure, these tools prove to be completely unsuitable.
• In addition, these rigid tools make it impossible to carry out multiple planning cycles in line with market dynamics.

Processes and tools of this kind no longer fit into a fast-moving environment. In contrast, leaner processes can drastically change the benefits of a smart data strategy as well as the measurement and management of company performance. However, this requires the use of suitable planning tools.

Software with modern forecasting algorithms allows you to switch to rolling planning. With this type of planning, you can take the pressure out of annual planning. Operational planning, the mid-term plan and long-term planning can be mapped and simulated on data structures. Structural changes can also be modeled in this way and introduced as an option in the planning process.

Admittedly, you have to accept a loss of accuracy in return. However, this all too often proves to be a sham anyway.

With a clear data strategy, on the other hand, you have the chance of reliable data.