How can companies build a strong foundation for innovation through data?
According to DieProduktMacher CEO and Head of Data Science Fabian Dill, it all comes down to taking a holistic approach that combines relevant, thoughtfully-sourced data with a deep understanding of users.
While there’s no magic formula for unlocking the power of data, a 360-degree approach can help you establish a company culture that prioritizes data in decision-making—helping you answer the ‘how’, ‘what’, ‘who’, and ever-elusive ‘why’ behind your product.
Below are my rules for data-driven product development, gleaned from years of experience building dozens of digital products for clients at DieProduktMacher.
‘What' you collect matters
First things first: make sure you’re bringing in the data you need. While this probably seems obvious, it’s so important to take time to ask people about their daily JTBD (jobs to be done), and what information they need to make decisions.
Once you have the numbers, get to know them intimately. Build a complete understanding of what data gets generated when people use your product. While an eCommerce business might track shopping cart size and value, an ad-based site will focus on impressions and unique users. Acquisition teams will look at traffic inflow, click-through-rate and the like; retention teams care about visits per user and open rates on email campaigns; and others are interested in feature adoption, depth and frequency of usage.
The individual properties and breadth of data here will vary greatly by industry and vertical, but the important thing is that you know how user information flows to your team.
With your data house in order, make it widely accessible by designing focused and relevant dashboards. The best dashboards include:
- Absolute and relative numbers (e.g. conversion and number of transactions)
- Context (e.g. a timeline)
- Anomalies (e.g. alerts for unusual activity)
So does ‘how' you do it
When it comes to determining your methodology, the approach will change depending on whether you’re working with an existing product, or building a new one. I like to structure my thinking like this:
- Existing product → optimization → data analytics
- New product → machine learning → data science
Existing products will already have historical usage data available, and that data should always be used as the starting point. New products, on the other hand, don’t have the benefit of historical usage data, so collection needs to be implemented for the purposes of optimizing later.
You might also opt to leverage the power of machine learning for your product. But remember that a sound understanding of any existing data is necessary to guide you through the process of adapting and optimizing a machine learning model to meet your needs. That’s why data analytics should always come first.
Some outside-the-box thinking here can propel your product forward. Think about the different data sources available—and the ways new data could be generated—as part of your product definition process.
Could there be valuable data in a previous version of your product, or a totally different product? Along with your company’s own data, what publicly available data exists? Are there other products within your company’s portfolio that could inform how you collect information? Google is the master of giving away free products in order to collect data for other (paid) products.
At DieProduktMacher, we encourage a practice of gathering insights, information, and data from as many sources as possible: market trends, benchmarks, usability tests, A/B tests, analytics data, user interviews, and more.
Know ‘who' you’re innovating for
While numbers and anonymous market research can help, the only way to truly understand your users’ needs is to talk to them. Go beyond the bubble you typically operate in and connect with both users and non-users of your product. Ask them why they are or aren’t using it. Use this information as the basis for reading all future data you collect.
Combining an understanding of your users’ needs with the “what” in the first section of this article is crucial for optimizing your product. Ask yourself: “what data do I have to help me address my users’ needs, and what patterns could be used to deliver user value?”. Taken together, these two insights can begin to reinforce each other.
My mantra when it comes to user needs:
Prioritize simple solutions over complex ones.
Many of today’s products are designed with AI capabilities in mind, but the best ones always place the user at the center. The thinking shouldn’t go: “GAN (generative adversarial network) technology is cool. Now that it’s possible for machines to generate photographs that look authentic to the human eye, let’s build a feature that lets us apply it.” Rather, it should start with: “What are my user needs? Is there anything I could address with GAN technology?”
Getting to ‘why’
Remember, quantitative data can only provide us with one perspective—it helps us understand the ‘how’, ‘what’, and ‘who” questions, but never the ‘why’. This is where qualitative data comes in. Qualitative data is much better suited to answer those burning ‘why’ questions and inspiring new, innovative approaches towards data analytics.
For example, let’s say you release a feature with personalized content aggregations that can provide recommendations to users. You might find that this feature is heavily used in the beginning, but over time, usage declines. A detailed cohort analysis tells you that users who engaged with the feature over a longer period of time were more likely to churn. Interviews with some of those users reveal that, in their opinion, the quality of the recommendations decreased over time and the content was no longer interesting.
Numbers may tell you that a particular product feature isn’t being used, or that user retention is declining. But only by talking to users will you be able to uncover why they’re ignoring it, or why they aren’t returning to your product. Taken holistically, these insights will help you formulate a hypothesis, implement it, and begin the iterative process of gathering data once again.
At the end of the day, no one thing in isolation can provide you with the truth about your product, or the next steps you should reasonably take. But by continuously striving to understand data from multiple angles, companies can begin to paint a clear picture of their users, and—most importantly—decisions at every level can become truly data-informed.