Sharpening your competitive edge with a data strategy
TriFinance Data Dinners: Antwerp
As a rule, data strategies are developed with specific business objectives in mind. Some organizations just want their data analytics to support traditional functions like marketing, to for instance create better buyer personas, or HR, by quantifying workforce performance. Others have a more general aim: they want their data to drive business or inform strategic and operational decisions.
Any which way, a data strategy is a necessary component of an organization’s digital transformation. ‘To sharpen their competitive edge,’ Toon Borré says, ‘organizations really need to develop - as part of their business strategy or next to it - a clear and articulate view on how they want to collect, handle and use data in all its dimensions.’
September 5, Toon Borré hosted the first of a series of TriFinance Data Dinners in Antwerp. During these dinners, Toon presents a conceptual framework for senior executives to discuss, adddressing strategic issues. Participants are also able to exchange views and experiences from their respective industries. The focus of this first Data Dinner was the alignment of business and data strategies. Taking part in the round-table discussion were senior executives from a broad range of industries, varying from services and retail to graphics and petrochemical industries.
Article written by Dirk van Bastelaere
Data Maturity Pyramid
To develop a data strategy, organizations first need to determine the maturity level of their data practice by plotting it (or some of its divisions) on the Data Maturity Pyramid. The figure basically shows different levels of an organization’s data use. It gives organizations the opportunity to pinpoint the As-Is as well as the To-Be situation, facilitating the development of their data strategy. Not every organization should strive for the predictive or prescriptive levels, because their As-Is position might not allow for such an ambitious leap.
Probably the most crucial level to start with data strategy is the data bedrock. To enhance the multifunctional use of data, it should be made available across organizational silos, all the while guarding and optimizing its quality. Next to data governance and data quality, the bedrock phase should include GDPR compliance.
Garbage in, garbage out: on data quality
Data quality, however, depends on the people inputting, collecting and handling data. That proved to be an area of concern for several Data Dinner guests. A big retail company actually tries to create an acute awareness in its employees that data is essential to the company. 70 percent of its 30.000 staff are collecting and processing data on a daily basis. Employees in logistics are for instance made aware that their role - and the quality of their work - are quintessential to the rest of the chain.
An organization specialized in time registration emphasized that employees should be convinced that correct use of data is in everybody’s interest, especially if the company aims for the predictive level. ‘The best data governance is useless,’ Toon Borré says, ‘if people handling the data aren’t involved. People should be shown what the impact of their work is. ‘Garbage in, garbage out’ especially applies to data. So, it is crucial that people involved in day-to-day operations also be involved in the data strategy preparation. Knowing the daily practice, they will not only contribute to solutions, but also be the organization’s greatest ambassadors for data quality.’
The case was mentioned of a utility company that made annual bonuses dependent on data quality. In terms of safety, a field technician locating a gas pipe 10 centimeters or 1 meter below the surface makes a huge difference. Rewarding correct data input financially might be a factor of its success.
From basic to predictive analytics
In the basic analytics phase, organizations use standard reports generated by SAP, Oracle, Microsoft AX and similar environments. Starting organizations that don’t have the budget or the capacity for more complex analysis typically dwell here. During the advanced intelligence phase, BI solutions like Power BI, Qlik Sense, QlikView or Tableau are used to analyze data stored during the bedrock phase. Organizations try to understand the past by analyzing existing data on a more advanced level, checking last month’s financials or evaluating the decision to open a new subsidiary, though more innovative approaches like process mining also become possible. This is the level where management information rules.
On the real-time level, organizations are capable to take informed decisions without any delay, based on real-time data analysis. The success of predictive analysis depends on the quality of the data sets and insights that the statistical model is built on. Organizations that try to act prescriptively on their data, typically work on influencing customers. Past behavior inherently contains information about future actions. Companies acting on the prescriptive level try to send the right messages at the right time in the right way. That, however, is only possible if you know what a person would actually do or buy in certain circumstances. Since brand loyalty is notoriously hard to break, it’s essential that you know exactly why a certain person buys AUDI to make her switch to Mercedes. Data quality on this level should be impeccable. Data silos should be abolished. In the domain of prescriptive, Artificial Intelligence is king.
Defining the As-Is
Most organizations, however, will position themselves at several levels of the Data Maturity pyramid, with different functions, departments or divisions to be situated at different maturity stages. An organization that is about to define its position should ask standard questions like: ‘How good is the quality of our data?’ ‘Have we made the best use of our advanced environments
?’ ‘Are we capable of building a ticker
to inform every manager with real time analytics? Developing a data strategy, organizations should inventorize the different maturity levels
present in their organizations for the As-Is, all the while defining their strategic goals
, keeping in mind that different divisions or product groups might need different strategic journeys. The only mistake organizations should not make, is immediately aim for a predictive or prescriptive approach, because nearly perfect data quality is a necessary condition to embark on that journey.
3 data strategy components
Starting point for every data strategy should be the organization’s business strategy. A data strategy should answer the question how data, in all its dimensions, can help realize business goals. Only then, business cases can be made to e.g. build siloless, prestigious repositories like data lakes to manage big data. Simple conceptual work won’t ever make it as a business case. Data projects that don’t fit your business strategy are useless.
To answer strategic questions, organizations should also start to inventorize their data. When asked to list the data sets they have available, it’s fascinating to see how short the list is. Not much seems to exist beyond ERP or CRM. Most organizations, however, live on top of a data goldmine. Fuel card data, telephone data, login data, gps tracking data, time registration data can all be added to the inventory. By also taking freely available open data in consideration like weather data, employment data, economic data or even library data or cost of living data, organizations can jumpstart their data strategy development. These data sets are actually the blocks organizations need to build their data practice.
A third element to inform the data strategy is knowledge. The higher an organization is plotted onto the Data Maturity Pyramid; the more expert knowledge is needed. A media company can know exactly what subjects you are interested in by analyzing your reading behavior. A retailer can send you custom coupons by analyzing your buying behavior, using the loyalty card. So, knowledge is needed about people’s behavior, but it also implies knowing what tools your organization is using.
Only the combination of business strategy, data sets and knowledge will be a solid enough basis to develop a data strategy and the starting point for identifying the organization’s strategic data opportunities.
Being a notoriously difficult process, the translation of the strategy into strategic data opportunities is essentially a creative process, where different stakeholders like marketing, sales, business people or even support people can help define data opportunities.
In one of Belgium’s largest retail groups, a so-called ‘business analyst’ function was created because a longlist of strategic data opportunities was made up. As a strong believer in data, the CEO created a role that was given carte blanche to look for added-value in the cross section of the data strategy components. The new role proved to be a success. Almost on a daily basis, the business analyst, who is now working on real estate, finds new business opportunities, creating that added value looked for by the CEO.
An organization offering travel assistance started to deploy data analysis to explain certain phenomena to higher management. By analyzing data logged by intervention teams, the data analysts were actually able to prove data analytics’ added value. It proved, for instance, that a pool of replacement cars in certain periods was insufficient to meet the clients’ needs. The problem was solved by sourcing externally, enlarging the car park in critical periods. In what was a typical advanced intelligence phase, management started to see more and more diverse possibilities for data analysis, actually contributing to the list of strategic data opportunities. The question was asked which factors were internal and could be controlled by the organization itself, and which were external and beyond the control of the organization. With weather conditions proving to have a huge impact on the number of cars having technical problems, the organization soon started to acquire data sources like weather data to improve their data analysis, moving up one level in the Data Maturity pyramid towards the predictive.
You need pressure to create diamonds
Moving from idea to execution can be difficult. But it needn’t be. Using a subset of data, an organization can set up a data lab to find out in ten, twenty days if there is e.g. a link between employee turnover and commuting time. ‘No need to develop complex processes or structures,’ Toon Borré says. ‘Just opt for a fail fast approach. If it works, you can proceed with the development. If you run out of budget or if it does not work: just kill the project. If it works, you have a proof of concept and are ready to develop your business case, after acceptance of which, the model should be made more mature and deployed.’
On the Data Dinners: Each of TriFinance’s Data Dinners has a different theme relating to data and data analytics. The dinners are organized throughout the country, offering senior executives the possibility to exchange views and experiences from their respective domains and organizations. Host of the Data Dinners is Toon Borré. Leading Data to Insight, TriFinance’s data expert practice, Toon assists organizations in becoming (more) data driven, focusing on data strategy, data governance and data analysis.