How you can build a TRIPPLE*S Data Organization - Sustainable, Scalable and Successful

The professional usage of data has been extremely accelerated in 2020. The forced disruption towards “no-touch” digital services demands a more holistic interpretation of the customers’ online and offline behavior. Many experts predict that this changed customer behavior will continue even after the unprecedented time we currently still in.

From an organizational point of view, I have experienced that businesses of all sizes acknowledge the value and importance of their data. However, they are mostly still at the very beginning to harvest the “digital gold” for tapping into new business models or to simple increase operational efficiency.

My passion is DATA and my purpose is to make organizations each day a bit more data driven.

I would like to share hands-on concepts and my personal experience how to build what I call a TRPPLE*S Data Organization. Ultimately, an ecosystem of technological, personal and organizational capabilities needs to be created. Only if all three above are tackled simultaneously data can be used at scale to contribute with a significant positive business impact.

Before I open the floor for the 6 DATA TEASERS below I want to offer a summary in form of a cooking recipe. This should help you to get started. I don’t want to undermine the complexity of building a TRIPPLE*S Data Organization. The good news is that in most cases with rather small initial actions you can already prove the value and thereby get the buy-in and investment for the longer journey.

TEAMING - Get a group of motivated data experts together and give them a clear use case.

VALUE - Find a suitable use case – it is reasonable to start with a reporting use case, not ML or AI.

TIME - Limit the time to three months, then check and adapt.

FUNDS - No or only low invest required.

TECH - Use an existing or new cloud platform and onboard the data. If you have to start from scratch plan enough time for your companies' cloud security checks.

KPI - Define your measures of success and provide the current baseline numbers you want to compare with.

SUPPORT - Pitch this to one senior leader (ideally non-IT) and get his buy-in.

If I got your attention, you should browse through the details below. Before you do so, let us define the crucial term DATA PRODUCT for a joint understanding. It is build with DATA, TECH, ADVANCED ANALYTICS and BUSINESS PROCESS. All combined a Data Product provides monetary or non-monetary value either for an internal process or for an external customer.

DATA TEASER ONE - One delivery model – Business Unit(s) and IT melt together:

To successfully deliver a data product, organizational boundaries need to be torn down.

Many discussions I experienced start at the point where Data Scientist and Data Engineer should be organizationally located. The answer is quite simple because it doesn’t really matter. What matters most is to successfully build a dedicated product team. This team should be fully empowered by giving them the responsibility for the business outcome. A crucial role is the Product Owner (PO). He manages the stakeholder expectation and designs the product. He is also in charge of the product backlog (EPICS and users stories). By guiding the development team through every product increment he is also on top of the features as well as the value provided. This role and skillset needs to acquired in most organizations. Don’t start without it. It would be starting a football game without the goal keeper.

DATA TEASER TWO – FOCUS on user/customer value:

A TrippleS Data Organization purely focuses on outcome in contrast a project driven organization that creates output. This means that a projects deliverables are scope, time and budget. Whereas a data product maximizes the value for user & customer.

The product team mentioned in the first teaser has the ambition to conquer new markets or to significantly improve the customer experience.

DATA TEASER THREE - Agility is the only way to be successful:

The incremental development, demoing, testing and adaptation is a crucial success factor. In an ideal TrippleS Data Organization new features are available every months and at least bring a small value. The development team shows the features to the product owner and he needs to accept them based on the Definition of Done. After that the features a rolled out to the end customer. Monitoring and measuring are essential tools for the PO and his team to understand the impact. I experienced that the design of the features into well-sized user stories is often a challenge. I advise that you rather start smaller.

DATA TEASER FOUR - Develop the required skillsets, because a data scientist is NOT a data engineer:

All required roles and skills for the data organization are in high demand at the moment and not easy to find. If you are in the comfortable situation to have them already on board – good. If not, don’t worry you will get there. I am recommending to start with Data Product that are aiming at improving the efficiency of internal processes. Regarding the staffing a business analyst, a business intelligence developer, a data engineer and a cloud solution expert are required. The next maturity level are data products using advanced analytics. Here a data scientist and data engineer are a must have. Keep in mind that every person can acquired these skills. From my experience the most complex role is the data engineer due to the technical knowledge, paired with data and programming skills.

DATA TEASER FIVE - Reliable Tech Platform – don’t worry about speed and scalability:

All big cloud providers have offerings that are suitable for the TrippleS Data Organization. I would recommend that you set a standard in terms of the development language used. I additional a joint source code repository and a high degree of automation right from the beginning is advisable. Costs usually are of lower importance at the beginning. However, you need to watch out how they increase if your data products scale.

?DATA TEASER SIX - Measure form the beginning:

There are several types of Key Performance Indicators (KPIs) that should be taken into account. Business value KPIs (e.g. sales increase), Productivity KPIs (e.g. velocity) and Technical KPIs (e.g. APIs calls). The topics is not new. I recommend that before a data product is initiated the target KPIs are defined as well the current baseline. ?On an organizational level the introduction of OKRs are helpful. This is only needed of several data products are managed and developed at the same time.