How Data Strategy Sprint helps you achieve your business goals

Artificial intelligence and machine learning have made enormous progress in the last decade. AlphaGo plays the abstract strategy game Go consistently better than any human, speech recognition can be used in various environments, and internet search engines and ad targeting are sometimes accurate to a frightening extent.

In spite of this ongoing development, AI and machine learning have had a major impact in only few companies.

Over the years, we’ve done numerous data and AI projects, ranging all the way from building modern data architectures to autonomous intelligent systems. Having seen dozens of applications, we’ve learned that there are a few main reasons why organizations are unable to grasp the full potential of data.

Sometimes, organizations might concentrate on building a fancy data infrastructure but barely use the data that it provides. Another typical issue that’s standing between organizations and their data is a siloed organization structure. Business drivers can simply be too far away from the implementing arm, or the two might even have contradicting goals.

Solving these types of dilemmas has led us to develop an approach that we’ve found very rewarding: it aligns the organization behind shared goals, makes data available throughout the organization, and ensures that the data projects are relevant for business. We call this approach the G-A-I-D method, short for Goals, Actions, Information, Data.

Here’s how it works in general, and as an example, for a video-on-demand service.

1. Goals and Actions

“Forget data-driven. Data is a way to achieve your goals, but it’s not a goal itself.”

Data in itself has no value. Nevertheless, it can be used to achieve goals – for instance to increase customer retention or to reduce churn, manufacturing costs or downtime – which is where the value comes from.

Once the goals have been set, the next step is to start thinking about concrete actions. They can be new features in services or products, customer-acquisition campaigns, or even completely new services or products.

As an example, consider a video-on-demand company with a flat-rate business model seeking to increase customer retention. By looking at user statistics, they’ve realized that customers who consume plenty of content, have a higher lifetime value than people who use the service occasionally. As a result, they come up with a hypothesis: Maybe their churning customers don’t find relevant content or there simply isn’t enough of it for them in the service. The company can start testing this by creating personal recommendations based on their users’ watching habits.

2. Information

“Information enables better decisions and even decision automation.”

Information makes it possible to know what content to promote to which customers, who would be interested in the service, or who are in risk of churn. Statistical methods and machine learning provide a theoretically grounded basis to get this information from data in an optimal manner.

In the case of the video service personalization, it would be important to know how likely each customer is to watch each video, if the service recommended them. Should this be known, the service could be designed to recommend each customer what they want to see.

3. Data

“Leverage data in a systematic manner.”

As industries and domains are different, there’s no silver bullet that would magically work in everything. The trick is to use data in everything the company does – and to experiment relentlessly.

As a tip, the following data sources could prove to be highly beneficial: user reading history, user demographics or service-usage patterns, levels of chemicals, pressure and temperature, quality of the end product, service history of your devices (e.g. replaced parts), measurements of the parts in action.

The video company could, for instance, dig out information on customer preferences and likings – the watching history of all customers being probably the best source of all. In addition, any content-describing data would be valuable, as it would make it possible to predict who’s likely to be interested in watching new, unwatched content.

Data Strategy Sprint can get you started

“Start small and experiment, break silos and iterate.”

How can the Goals-Actions-Information-Data approach be brought to life? Recently, we had the pleasure of helping out KSF Media, the owner of the biggest Swedish newspapers in Finland (such as Hufvudstadsbladet). The task was to lay out their strategy in order to quickly get value out of data and artificial intelligence. We did this with a five-day data strategy sprint, which was fuelled with these agile principles:

  • Iterate: get feedback and talk to people early on.
    As you go through the G-A-I-D steps, you most likely get new ideas: What kind of knowledge is needed, what data to gather and which actions should be taken next. The approach encourages you to iterate over solutions, as it gives you the best value in the long run.
  • Get value fast: start small and experiment.
    A concrete result, however small, is more valuable than perfect far-reaching plans that will never be executed. Concrete results give you the possibility to measure their impact. This will show your organization the value of what was done. In addition, by measuring, you can control the amount of resources that are used for reaching the goals.
  • Forget silos: Invite all relevant people.
    In many organizations, business and IT are separated. The G-A-I-D approach forces you to break these silos – and will give you a well-aligned organization for free.

With the words by Sami Kallinen, their Chief Digital Officer: “I find using data strategy sprints a good way to align my organization around common goals. Reaktor’s participation and their fresh GAID approach was a significant add-on in the process.”

Recommended content