How can you set up a data-driven product development process?

There are so many metrics to choose from, some more important than others depending on where in your journey you are at. For example, your sales rep capacity is less important at the start than understanding your customer lifetime value (LTV), or Customer Acquisition Cost (CAC). In my experience, I find that aligning with a framework provides the best structure and metric selection criteria. Start with your revenue funnel. From there, work out what metrics are of value at the Awareness stage for example. E.g. Conversion goals completed rather than Page Views. For an established product, and assuming that you have product market fit, metrics focused on capabilities that your customers use are more important to track.

I am a big fan of establishing a dedicated data infrastructure capability. Something that you can secure, govern and centralize the capture of your data. A solution like Amazon's AWS S3, Lake Formation, and Glue products is ideal for this. They will allow you to master your data and grow, in a way that a spreadsheet will struggle. In addition they are agnostic to data source, and are flexible enough to support any type of data that your customers and your product will create. Further, they will set you up well to analyze the data, using either a spreadsheet, reporting tool like Tableau or Machine Learning / AI

Managing metrics at the company is great, but this approach will not tell you 1) where your revenue is coming from 2) differences in customer behaviour Your customers will naturally migrate into different clusters or groups. Each one will have different attributes that define them. I.e. forward payers, monthly payers, heavy users, light users, open to upgrade, like where they are, for example. In your analysis, segment early. Machine Learning like K-Means can help you here. Now analyze each group to find true insight, rather than generic overview.

Two approaches, AB Testing and Multivariate testing. The former focuses on a single attribute, while the latter tackles many. For example, a multivariate test on a web site, the test would change the product picture size, font size, color of the Call-to-Action (CTA). The process takes a couple of weeks. Establish a base line. Collect data on what your customers are doing now. Make your changes. Collect data on the new layout. Compare. Keep in mind the importance of segmentation, so you may want to limit your test to a specific segment.

The process takes time. Gather the data and compare. If the output in inconclusive, then consider changing one attribute or adding a completely new attribute to the test. As you look at the results, even if you are targeting a single segment, look to see if a new sub group has emerged. Look at the outliers. They could be the start of a new untapped type of customer or something that you need to ignore.

Documentation and Knowledge Sharing: Document your processes, findings, and insights to create a knowledge base. This aids in preserving institutional knowledge and sharing learning across the organization. Adaptation and Evolution: Be ready to adapt and evolve the process continuously. Embrace new technologies, methodologies, and best practices to stay relevant and improve the effectiveness of your data-driven approach over time.

In my experience, to set up a data-driven product development process, integrate continuous user feedback mechanisms directly into your product. Utilize A/B testing to make informed decisions about features and design changes. Foster a culture where decisions are made based on empirical evidence and encourage cross-departmental collaboration to ensure diverse perspectives and data sources contribute to the product's evolution. Invest in technology that can predict trends and user behaviors, thereby allowing you to proactively adapt and innovate your product. Prioritize agility and flexibility in your development cycle to quickly respond to the data-driven insights you gather.