Whose Job Is It To Produce Actionable Insights?
Playing the blame game.
I read a recent article from fellow Medium writer Tim Wilson where he did a survey to a community of data analytics / data science practitioners on the pain-points they experience while working with their stakeholders. One of the pain-points that emerged was on the topic of actionable insights. Stakeholders sometimes (and perhaps even oftentimes) complain that the data-driven work that the analytics practitioners engage in lack actionable insights.
But what do the stakeholders mean by “actionable insights”? Do they have the same converged understanding of the term? And is it really the job of the data analyst / data scientist to produce actionable insights? Is the complaint even valid?
And so I dedicate my 56th article to this topic of who owns actionable insights.
(I write a weekly series of articles where I call out bad thinking and bad practices in data analytics / data science which you can find here.)
WTF is Actionable Insights
What exactly is “actionable insight”? There are several common interpretations of the word. Strictly from a knowledge management perspective, the word “insight” refers to information that is used to augment or improve the current state of decision-making. There’s actually a lot happening in that definition. For starters, the word “information” refers to interpreted data, which implies that it’s probabilistic in nature rather than deterministic (i.e. the same data can have more than one interpretation). Secondly, to augment or improve the current state of decision-making implies that the maker of the insight is deeply aware of the collection of information used in the current state of decision-making. Finally, to augment or improve the current state of decision-making also means that the maker of the insight is aware of the sub-optimality of the decision outcome(s). Based on this definition, you can immediately appreciate that there is no need to use the term “actionable insight” because there is no such thing as a non-actionable insight.
From a business stakeholder perspective, the word “insight” can refer to:
- A finding that indicates a potential opportunity or emerging risk.
- An explanation as to why my business isn’t working.
- A recommendation on solving my business challenge.
Business stakeholders love to tag on the word “actionable” onto insights to emphasise that they want to be able to immediately act on the data-driven outputs, rather than to think about the possibility of actions from the outputs as a logical next step. That is, the action should be self-evident or directly recommended.
Ownership of Actionable Insights
This range of definitions of “insight” or “actionable insight” obviously creates frustration for the analytics practitioner. In many instances, they may feel that the responsibility of creating actionable insights is not within their realm of responsibility. I would argue that in a fair number of cases, the responsibility of action rests with the business stakeholder.
It all comes down to how the data analytics / data science function is set up in an organisation. Broadly speaking, the analytics function can assume an order-taking role, an influence role, or a business driver role. The choice of role is very much influenced by the business stakeholders, despite how the analytics function sees itself. So if the business stakeholders view the analytics function as an order-taker, then it should not complain that the said function does not produce (sufficient) actionable insights since they are precluded from the top-of-the-funnel discussions and exposure to the business contexts.
“Action” implies decision-making. The ability to take decisions from a data-driven output is completely contingent on knowledge about business model context, resource availability, and customer and competitor responses. While every analytics function should strive to be more knowledgeable and helpful in these domains, they are generally not part of the analytics function primary competencies.
Aligning Expectations
It is the obligation of both parties (business stakeholders and analytics practitioners) to align on the output expectations for any piece of data-driven work. Given that the work requests generally originate from the business stakeholders’ end, they can often help their analytics partners achieve actionable insights by clarifying if the output falls into points (1), (2) or (3) as detailed above. If it’s point (1), the business stakeholder should provide information and context to the operating hypotheses around the request or problem statement. If it’s points (2) or (3), the business stakeholder should provide clarity around the assumptions inherent in their business model / business solution, allowing the analytics practitioner to quickly parse findings that are intuitive and obvious, versus those that are counter-intuitive and unique.
On the flip side, the analytics practitioner should be able to figure out what information is missing once the business stakeholder provides clarity on the output expectations. Are there simply pieces of information that do not exist and require experimentation to collect? Are there information about market or resource constraints that need to be factored into the analysis to ensure relevance?
Conclusion
Much of the problem with data-driven outputs being perceived as lacking actionable insights stems from poor requirements articulation, mis-aligned expectations and the inability to recognise missing information. This is why data literacy is so critical as a competency for the business stakeholder, and why data sensemaking is a paramount competency for the analytics practitioner.
To expect your analytics partner to operate independently and recognise the nuances in the data or analysis request, and ask the right questions that will lead to actionable insights, is not fair. Business stakeholders must take ownership for shaping the data-driven outputs they desire.
