Making Data Actionable: The Immense Challenge of Good Data Products
“Data-driven” is such a passive phrase. Who the driver is matters a great deal.
When you think data science, the image of models, algorithms and statistical methods comes to mind. So, why, then, when you mention data science in many companies do you get the response “that’s all well and good, but I just need a chart to see some things”? I know it feels like that “devalues” data science in some way. “They just don’t appreciate what I can do” is what I hear from those who have experienced those interactions. I don’t think that is quite right.
When a stakeholder, business partner or colleague says they “need” something, they are telling you something valuable - they are telling you what they are trying to do and what would help them do it. Our partners tend to understand that data science is powerful, complex and useful. But what they are telling us when they say “I just need a chart to see some things” is that they need what we create to be actionable.
The problem is not that the methods, tools and outcomes are not useful in general. We can talk all day about the cool things we build. The problem is that usefulness is in the eye of the person consuming the results. If they don’t think the product of our data science work is actionable, no level of “cool” will matter to the business. So the problem I see is that data needs to be actionable. Not just plausibly actionable but easily actionable.
The good news is that data products help get us there. While making something actionable through a product is likely as diverse as the methods, processes and visuals we have for data, there are some places to get started when you think about actionability. Here are 5:
Focus on the consumer, not the producer. Data producers, product builders and data scientists are adept and knowing what they’d like to consume because they know what they are trying to do. But when it comes to business partners, stakeholders and (insert other buzzword here for those you work with), the key is to know what they are trying to do. The first step is to know what actionability means to the consumer of your data product.
Use the 10 second rule. If your data product is truly simple and focused, the consumer or user of your product (or insight, dashboard or metric) should be able to capture the main idea of it in 10 seconds or less. That means it should be crystal clear what matters and where to find it. If people spend time hunting for the message/idea, you’ll lose consumers fast.
Give time for design. I get that it is easy and tempting to toss together an MVP. For data scientists that talk to each other, a rough draft and ugly table or metric is probably sufficient to drive to actionability. Yet that simply is not good enough when your consumers are business partners across the company. Design matters. That doesn’t mean everything needs to “look pretty”. It means that you should give great care to how you want people to interact with the product you are creating.
Talk to the non-adopters. I *love* when people tell me something my team has built does incredible things for them. But that is like using sales to measure demand. What about the people who would have used it but were blocked for some reason? Were there any access issues? Were they confused at the takeaway message? Did they not know how to find what you created? If you don’t think about the potential audience and only curate for the existing audience, your product will never be as powerful as it could be.
Keep pace with consumer needs. Products built for actionability in one month might not support the right type of actionability 6 months later. It isn’t the product that caused issues. Instead, the consumer’s needs changed. Perhaps the business shifted. Potentially there’s a new focus or primary outcome. Creating an actionable product is not a moment in time endeavor. There is also a commitment to improvement that comes with it. And if we can’t make that commitment, it might be time to think about deprecation.
“Data-driven” is too passive because it does not identify who is doing the driving and what actions that driver is trying to create. Where are they going? Why are they going there? Are there alternate routes? To create great experiences with data, we need to think about being in the passenger seat with them, rather than sketching out a map of a journey from 1000 miles away.