Thematic Product Insights For Making Better Strategies
By Claudia Natasia
Co-Founder & Chief Executive Officer
I was chatting with a product manager who is working on a feature with the intent of improving user retention. To help formulate a direction, the product manager scoured user research notes, product analytics dashboards, and Zendesk tickets, only to end up with 25 possible directions for the feature without a clear way to discern which of the directions would likely yield better results.
This was a problem I’ve personally faced, and after probing deeper, I realized that the root cause remains the same: user data exists in disparate forms, making it difficult for anyone to combine and distill them into meaningful insights for decision-making. More often than not, your user analytics exist in Postgres, research notes are stored on Google Drive, and customer feedback is recorded on Salesforce, Zendesk, or Gong calls.
The key is to build a thematic product insights practice. This helps you inject organizational discipline into how you collect and analyze data so you can always create strategies with more confidence that you will be influencing the most crucial business outcomes.
Here are four steps for how you can start a thematic product insights practice:
1. Select and center your practice on key goals.
Identify key goals that you hope to drive with your product (e.g., Adoption, Retention) and their relevant metrics (e.g., MAU, Recurring Usage). Ensure these goals map toward your company’s high-level business objectives. Here is a talk I did on how to create goals that align with business objectives (talk starts at 1:39:18).
Additional Tip:
Be brutal in selecting these goals. Less is more; otherwise, you end up with 30 and analysis paralysis. When selecting goals, you should feel confident that if you were to focus everything you are building in the next year on improving these goals, it would be the right decision.
2. Cluster existing insights around the goal.
Review existing data and cluster the ones that are relevant to a particular goal. This creates an evergreen body of knowledge you can pull from that is centered on a higher-level business objective and is project-agnostic.
Additional Tip:
Before you start clustering existing data to relevant business goals, set clear parameters on what could fall under a particular goal. This helps avoid inconsistent classifications.
3. Cluster new insights around the goals.
Now that you already have bodies of data clustered toward relevant business goals, evaluate the new data that you need for your upcoming products. Remember to always incorporate new data with these existing goals.
Additional Tip:
Be extremely disciplined with the research and product analytics projects you start so that you can always tie back the data you collect to core business goals. If not, you may end up in a state of analysis paralysis trying to wrangle noise that is irrelevant to your end goal.
4. Statistically predict the likelihood of goal achievement using your data
To better forecast the success of a strategy, model your existing data to predict a particular business objective. For example, if you’re trying to drive recurring usage (user retention), run a regression model against user retention data to see if there are specific user cohorts that are more likely and least likely to return to the product. Then, aggregate any qualitative feedback from these user cohorts and run a topic model to figure out strong themes behind why they are returning/not returning.
Additional Tips:
Other common models that you can use to predict the likelihood of achieving certain goals through a series of events or insights include:
Markov Analysis: particularly relevant for product telemetry
K-Means Clustering: particularly relevant for figuring out user cohorts/personas that are the drivers of certain goals
AFINN Lexicon: particularly relevant for calculating the average sentiment behind qualitative data
From my experience building a Product Insights practice firsthand, I know that it is extremely expensive, as it takes deep data science investments and a large amount of time. That’s why we decided to build Riley AI to help make data-driven strategies accessible to every single organization and automate every single step in just mere minutes.
We’re here to make data-driven decisions extremely easy for you!
If you’d like to be part of a community of product people testing and refining Riley AI, you can sign up here, and I’ll be in touch.
Claudia is the CEO & Co-Founder of Riley AI. Prior to founding Riley AI, Claudia led product, strategy, and data science teams across the Enterprise and Financial Technology space. Her product strategies led to a $5B total valuation, a successful international acquisition, and scaled organizations to multi-million dollars in revenue. Claudia is passionate about making data-driven strategies collaborative and accessible to every single organization.
Claudia completed her MBA and Bachelor degrees at the University of California, Berkeley.
LinkedIn
