Maximize the impact of data products

As a scientist, we often are so inspired by the beauty of mathematics that our work might be perceived as misaligned with companies goals. TL;DR "geeks are doing useless slow geeky things". Along the projects I coached as a Data Product Architect advisor, I gathered a few concrete tips in order to maximize the impact of data products in companies, while keeping engineers' motivation high for squeezing their best performance.

• Reframe with the WHY with the stakeholders and design relevant target KPIs

Regardless of the business' state of a client: growing, stabilizing or even sinking, it's business. The headline therefore in the project should directly refer the company objectives. For achieving coherence between technical projects and company goals, KPI definition is the key bridge between stakeholders' needs and the company vision. I particularly enjoy working with the OKR (objective key results) frame that offers a comprehensive structure.

For instance, a CEO might be overwhelmed with articles about big data and AI and wants to "jump on the train before it's too late". The role of the Data Product Architect will be to gather the relevant strategic and operational brains around the table and define together WHY this tool is the right one for the company to put their efforts and resources in at the moment.

The WHY helps you translate a wish like "I want AI to handle the onboarding" into tangible value perceptible by the end user, such as "with this AI component, we aim at reducing the onboarding time from 10 minutes to less than 1 minute, leading to an increase in acquisition volume of 30% over the next quarter".

👉 Read more about how to build efficient KPIs

• Data scientists work hand-in-hand with the product team

Allow data-* (analyst, scientists, engineers, etc.) to wear relevant hats during the relevant phases of a data product development. Those hats might be worn by the same person however require various skills, engage different stakeholders and require adapted technical / non technical communication.

Typically those phases are - yes it sounds like product management! Because your data product is no different than a technical application: it has users, user stories,

1. Discovery: analysts/scientists are involved even before the project kickoff in order to determine the project's feasibility and potential impact. We don't crave data science for the beauty of algorithms, but because it's the most powerful way to work with information nowadays. However, machines and resources require a certain budget, and this budget should be in adequation with the added-value (monetized, or perceived) that it brings to your product.

2. Implementation: engineers get involved to source the relevant data, not just for the proof-of-concept phase but throughout the lifetime of your product. Meaning they might rightfully ask for DevOps to be involved, a test environment to be created, an alert/notification system to be put in place to ensure a continuous quality delivery of information and feedback. Cost monitoring will be essential, and regular assessment of the target KPIs chosen the leading piece of information towards better iterations.

3. Impact measurement and iterations: once a POC is developed, throwing a data-science algorithm unsupervised (pun intended;)) in the wild is a guarantee for failure. First because a data-science algorithm output is often non-deterministic, i.e. you won't get the same results if you ask the same question twice. This is why output should be scoped together with the stakeholders in order to get close to the expected output as often as possible. Secondly because machine learning is all about learning, i.e. without feedback on the results, no possibility for improvement. This emphasizes again the close relationship that the data scientist will develop with the product management team, in order to get the product features to facilitate abundant, precise and reliable feedback.

   
   

With all this in mind, it appears clearly that the Data Analyst, Scientist and Engineer need a Data Product Manager to align their technical efforts within the company's daily development rhythm.

As a freelancer working for smaller scale-ups that have no data teams yet, I often have to wear all those different hats at once, and calling out those roles in meetings help developers, product managers and stakeholders understand my needs to take the data product further.

---

Do reach out if you are under the impression that you have access to skilled technical resources, and yet struggle to demonstrate the value of data products being developed. I offer mentorship for data teams to reshape their efforts for better impact measurement and ultimately alignment with business objectives