Using Data to Find Your Customers

Finding the right set of customers to pitch is always not easy for PMs who start building a brand new product. At the beginning, you have some interesting ideas, but you don't know who to validate with or sometimes the one you talk to should not be the right target audience. In addition to traditional approaches like customer interview or user study, data is your new tool.

In our journey of building a script library at Microsoft, we started with a rough idea that referencing others' scripts can help data engineers and scientists write their own queries. But the idea is too general which cannot guide us landing a useful product. Many questions come up - In what scenario do they need to search for others' scripts? What types of keywords do they input? What do they do when they find the search result useful? At that time, we collected search keywords from existing tools that offer similar capabilities. An interesting finding after analyzing the data is - one of the top search patterns is the query language grammar. This insight helped us refine our target scenario significantly - people need to learn the query language from real examples with a simple copy & paste. We use this statement to pitch our potential customers in different channels and the usage grew quite well after we launched the product.

Below are 3 questions to consider when you need to use this data-driven approach:

Where to find the data?

Start with your competitors (if you can get the data) or products which offer similar capabilities in other domains (e.g. when building script search, I can reference products that offer report, data, or even more generic search capability). From competitors, you can learn what they are doing well and what are the potential opportunities. From other related products, you can open up your mind for additional ideas to pitch your product and even find effective channels when you need to promote your product. The only thing you need to be careful is the data size. Make sure the product you collect data from has enough usage so you don't draw a biased conclusion.

How to analyze the data?

Aggregation is definitely the first approach you should use to analyze the data. By running basic statistics like sum, count, percentile, etc., you can quickly find some common patterns. However, based on my own experience, these "common patterns" are usually already addressed by existing products. The most fun part happens when you have enough patience to read through the data at more granular level and identify the gaps to full-fill customer needs. For example, when we checked the top n keywords searched in our intranet, most of them has nothing to do with work, but when looking at these long tail keywords, you can easily summarize the pattern that many people are trying to search language grammars. Using this pattern (e.g. a regex) to re-analyze the data, we found a new channel to promote our script library.

How to validate the insight?

Build your product quickly and launch it with the right channel to reach your potential customers. Log as much telemetry as possible so that you can have your own data to validate whether you are heading to the right direction:)