10 dimensions of making data science work (part 1)
Expectation: Dimension #1
“What is data science according to you?”
I have asked this question multiple times to people from varying backgrounds and functions in the start-up ecosystem when I was at xto10x— founders, technology leaders, business executives, people heading operations and the list goes on. The responses were around providing insights from data, building automated intelligence and applying the latest trends in deep learning.
Through the multiple conversations we had, we arrived at the most simplistic definition of data science — Making data useful for business. This definition not only helped demystify the complexity associated with the term but also inherently made it everyone’s responsibility in the organisation to make it work. Then the next question naturally was
How to make data science work effectively?
The quest to answer this question is what lead us to the 10 dimensions of making data science work. Each dimension has stages of maturity in the evolution of a start-up from a young company that has just achieved product-market fit to a mature data driven organisation. This is targeted towards all the decision-makers in the start-up ecosystem including founders who play a role in making data science work for their organisation.
In this part 1 of the series, we will be discussing the very first and the most important dimension — Expectation.
Expectation : Make it or Break it
Having the right expectations out of any initiative is almost 80% work done in achieving effectiveness. Many would agree that a new function or initiative in an organisation especially data science often suffers or fails to provide the right results due to one of the following reasons — lack of clarity on what to expect or having wrong expectations from this highly invested elite function created called — data science. Here are 5 stages of how you can expect this whole journey of making data useful to evolve.
- Stage 1 : Building to fit — During this stage, a start-up is just iteratively building solutions to find the fit with the market. The only use of data here would be to validate the hypothesis and intuition with which products were built via experimentation.
- Stage 2 : Tracking the Business — This is the stage where the start-up is on a scaling journey and hence the most important role of data becomes answering the question of — “how is the business doing?” as fast as possible with minimum human effort. All you can expect in this stage is regular tracking of L0 and L1 metrics on automated dashboards. Often this is the right stage to invest in getting a blueprint of data strategy for the business chasing a goal 12–18 months away. This helps define roadmap for all the teams with a balanced long term vs short term bandwidth allocation practice.
- Stage 3 : Age of v0 — When a business is trackable, it obviously leads to questions of — “Why is this happening? How can we improve this?” Like they say, you can only improve what you can measure. This gives birth to the period of several analytical questions answered and basic models often called version zero (v0) and rule frameworks built that help people make decisions better as well as automate some part of making the decisions. The low hanging fruits achieved during this process set a baseline that needs to be beaten by further stages.
- Stage 4 : Platforms and Production — This is the stage where fundamental requirements to apply sophistication are worked on. Platforms like data platform, experimentation platform, feature stores and machine learning serving platforms are built without which only hiring data scientists who build models on their notebooks will either take months together or not make it to production at all. Collaboration across multiple teams solving the right problems and enabling platforms for faster time to production is key during this stage. This is actually when the non-linear impact of leveraging data the right way starts becoming visible. Imagine the level of patience and right expectation a founder and the leadership team need to have to reach this stage from the very first one.
- Stage 5: Rinse and Repeat — The final stage is that of maturity where data science function is expanded to newly formed business verticals. Mature data-driven ways of working now gets embedded into every business unit and the cycle starts again.
The key to having the right expectation and understanding of the stages is to be aware that the metabolic rate of data science is way different from business operations. This function needs to live through the gestation period of these stages and be nurtured to flourish.
Soon, we will cover the second dimension of Strategy in part 2 of this series.
Content jointly created with Anand Sharma and Maneesh Mishra while at xto10x
