Discover more from From Data to Product
Why Is Treating Data Like a Product So Hard?
If you’ve read my writing for any period of time, first of all, kudos to you for your resilience. Second, you’ve probably wondered, why write about data as a product if what we’re really talking about is using product management principles to create value from data? As someone said to me, “data product management is product management in a different domain”. While their intent was to tell me why this wasn’t worth the time, I actually thought of it as a gift - specifically, exploring what value is there in applying product management principles to data? And what keeps us from doing that? Here’s what I mean:
Since I started paying attention to data science and related topics in the early 2010s, data has had this “magic” to it. From the early explosion of big data, to data science, to AI, ML and deep learning, for a decade we’ve lived in a data world with some “mystique in the black box”. For years now, companies and people have invested in the idea of the magic in the black box as much as the value in the black box. It is hard for me to explain, but when I heard an executive say in 2016 “we just need to trust that there is value in the data haystack”, I knew we were dealing with something different.
My takeaway from the last few years is that there seems to be a different set of rules and standards for evaluating the value of data. We get away with a lot of investment and ideas because of the allure of “magic” being embedded in the data. I’m not saying there isn’t magic to be found, but the idea of magic being there enables a lot of bad ideas and investments. That’s where I think the key value proposition of data as a product comes in - we can apply clear standards and identify strategy and value propositions to data in the same way we do to other products we build.
If data as a product is a way to create accountability and standards for data investment and value creation, a critical question remains: why is it so hard to do? I’ll be completely transparent here: I don’t have a definitive answer. But I have some hypotheses and hope that you’ll join the conversation. Here are 5 ideas about what “blocks” us from effectively treating data like a product:
The mystique of data itself creating value. For years now, we’ve been told that we need to invest in data. “You’ll be left behind” if you don’t has been a rallying cry for tons of data investments. I agree that companies are left behind if they don’t invest in data, but the problem is that we’ve let the idea that simply investing in data initiatives will product value for a company. It won’t. When companies fail to see the value created that they expected, the failure is ascribed to the model, the quality of the data or the “complexity of the problem”. We often don’t think enough about why we believe data will actually create value for us and in what way. It’s not automatic.
Data is a team sport - data science alone won’t let you create great products. Creating value from data requires data science, but data science alone doesn’t create sustainable value. Think of the dashboard graveyards you have in your company, or the models that no longer update or work. Those were great ideas and products in the moment, but there was no one thinking about the long term value they wanted to create. There was no one to manage that long term value creation. We can’t expect data science teams to be scientists, product managers, analysts and strategists all at the same time. But we do it nonetheless.
Product management energy is directed toward external products, not internal ones. If you go to any product company, they likely have teams of energetic and thoughtful product managers creating the next big thing for consumers or businesses externally. This makes sense, the revenue upside to building transformational external products is huge. But the energy given to external products doesn’t ofen translate to internal products. Maybe it is the lack of direct revenue attribution, or it is harder to find PMs skilled and motivated to manage internal products. In any case, I’ve learned that many companies just aren’t as interested in investing in internal products (like data) as they are external ones.
Quantifying value from data is hard. For many products, metrics like churn, revenue growth, retention and engagement provide a fairly good idea of how much value is being created for customers/users. But when it comes to investing in data, a lot of the value created doesn’t show up in direct revenue attribution. In many cases, data enables others to mitigate risk, to better understand how the business is operating and to have increased confidence in product or operational decisions. Product managers, and those focused on growth, want attribution. I’m a firm believer that if we want to treat data like a product, there needs to be agreement from leadership that the value may not show up in the ways we are used to capturing value.
We need talented people to be data product managers. Put simply, finding people with the skill sets to manage data products is much easier said than done. It isn’t very easy to find a data scientist these days, much less a product minded data scientist who is technical enough to converse with engineers but also fluent enough in strategy to set a roadmap for an entire product. Ultimately, the 4 points I mentioned above aren’t that important if we can’t find people who can function in this space. A lot of companies seem to want these people to magically come into their hiring pipelines. I don’t think that will ever happen. If we want this skillset, we need to be intentional about developing it, rather than waiting to find a unicorn.
To recap, my writing this week was about two things: 1) exploring what value there is in applying product management principles to data and 2) what keeps us from doing that? I hope you found this useful as a starting point for more conversations!
I want to end with a plug for an awesome (and free) upcoming conference called Future Data. If you’re interested in data as a product to any degree, I highly suggest you take the time to attend or pick a session or two that seems compelling. The CEOs of companies like Sisu, dbt Labs, Mode, Databricks and others will talk about the future of data and the data landscape. If you’re interested, I’ll be giving more detail into the topic I just discussed above - why data as a product is hard to implement.
As always, I’d appreciate it if you’d subscribe or share with others who might be interested in these topics and conversations.
Have a great week!