6 Essential Lessons for Building Great Data Products
After over a decade in data and writing a book on the subject, I’ve distilled 7 essential lessons for creating data-powered products.
For much more on those lessons, check out the book on it: “The 7 Data Product Strategies” packed with lots of examples and hundreds lessons more.
Let’s get straight to it:
1. Collect Unique Data
According to Marc Andreessen, the big challenge of venture capital is not to tell good from bad investments but rather to tell good from great ones.
Because for better or for worse, the very best VCs collect fast amounts of companies, most of which will go absolutely bust. But there is a power law to their returns, meaning the Top 3 of their successes basically dwarf every other win, and then the top 3-10 again dwarf everything else.
So, what does a great VC really do? He collects a vast amount of companies, knowing almost all of them will go bust and not return a penny. He's willing to do so because he knows some will be a gold mine.
Adobe took this approach with data. They collected everything—small, seemingly insignificant pieces of user interaction data. Years later, those seemingly minor investments became the foundation of their AI kingdom. A kingdom that by now includes dozens of AI features no competitor can replicate because they’ll need to collect data for almost a decade to catch up.
Be like the VC: gather all the data you can because what looks trivial today might just be your billion-dollar exit tomorrow. Be a data hoarder (within legal limits), but focus on your unique treasures. Ignore 'dark data,' and you risk missing out on hidden diamonds that could define your market. Your unique data is your edge—don’t underestimate its potential.
2. Build Feedback Loops
Feedback loops feed growth like a hungry monster that gets stronger with every bite. Yes, they get hungrier, but they will also be super powerful! So, do you want to have such a monster inside your product? Well, of course!
Because, unlike the actual monsters, the data feeders we talk about aren't actually hurting anyone. They just want more and more data.
Google’s search success is a classic example: more searches lead to better results, attracting more users. Design loops that reinforce your product. Think of it as an ecosystem—self-sustaining, with every piece nourishing the whole and always growing stronger.
Of course Google’s search needs vast amounts of data to become better, but I think you’ll agree, owning 90% of a market is kind of worth it.
3. Start Smart, Refine Later
Don’t wait for perfect data. Start with what you have—an algorithm, user trust, or even a small dataset. Launch, then refine. I learned this the hard way—one project stalled for a year because we waited for the 'perfect' dataset. I should have started with what I had and iterated along the way. Google began with a solid algorithm, and Netflix started with DVD rental data. Focus on your strength, then iterate.
There's always a way to start, even without the tech, without the data, or without any customers.
4. Exploit Information Asymmetries
Fail to recognize an information imbalance, and you're leaving money on the table for someone else to grab. Glassdoor exploited a lack of salary transparency and filled the gap. Wherever one party has more information, there’s an opportunity to bridge that gap and create value. Information asymmetries are fertile ground for datapreneurs—don’t miss out.
5. Give to Get
If you want to build a data heavy product, well, you need data! How can you get if the key is to get unique data? One great way is to provide incentives!
Create incentives for users to share data. Glassdoor required users to contribute salary info to access the platform—a win-win that accelerated data collection. If you want your users’ data, give them something valuable in return. Design systems today that make them want to share—no trade, no data, no progress.
6. Watch for Behavior Shifts
Changing behaviors can make your data suddenly valuable. COVID-19 made Yoodli's video meeting feedback features a hit. Stay alert, adapt, and capitalize on these shifts. Watch for changes that might make irrelevant data critical tomorrow. The future belongs to those who anticipate where the puck is going.