How to do customer discovery for a machine learning product
How to talk to people. How to spot patterns. How to quantify your findings.
Hey reader, welcome to the 💥 free edition 💥 of my weekly newsletter. I write about building machine learning products, working with people, and growing your career. You can learn more about me here. Feel free to send me your questions and I’m happy to offer my thoughts. Subscribe to this newsletter to receive it in your inbox every week.
Doing customer discovery is like trying to kindle a fire. When you're in the early stages of building a product, you have a hypothesis that you want to test. You have an idea of what people might need, but you need to validate if they will actually pay for it. How to discover what to build?
Here are 9 steps to do customer discovery for machine learning products:
1. Talk to people
One of the biggest pitfalls is starting to build the product without ever talking to potentials users/customers. Many people do it. Some people are not comfortable talking to people, so they get busy building. Only to realize that nobody wants it or cares about it.
Reach out to people. Talk to them over phone or zoom. Not email or text, but have a real-time conversation. Don't ask them if they're facing a problem.
Make them talk about status quo and how they perform a specific task.
You’ll gather a lot of insights during this process.
2. Gather all the facts
The next step is to meticulously gather all the facts. You need to have a clear understanding of reality.
All you need is a simple text editor to make notes. Don't automate anything. You don't need to be cute or clever.
3. Find out what's essential
Within those facts, something essential is hidden. Find out that essential thing. That's what matters to your customers.
In order to spot it, you need to have structured conversations. Get to the essence of it. Keep going until you can express it in plain english.
4. Spot the pattern
Based on that essential thing, you need to spot the pattern. In order to build a product, you need to spot the pattern.
The goal is to understand what is common between all the status quo descriptions.
What can be productized? What can make their lives 10X better? This way, the same product can be used by a large number of customers.
5. Quantify the pattern
Based on that pattern, identify the underlying function that can quantify that pattern. This underlying function is the way to build the product.
6. Build a product to automate that function
Make it 10X easier/better/faster than status quo. Customers use products to make their lives better. In machine learning, this usually comes in the form of automating a function that takes up their time/energy.
7. Put it in front of customers
Every time you do something, you need to put it in front of customers. Rapid iteration based on customer feedback is key.
Make sure to listen to everything, but use your judgment on what goes inside the product. You can't fit everything into your product. Your job is to spot patterns that can become reusable features.
This will also help you let go of customers who are not a good fit.
Usually because they need something different or that your product doesn't serve their needs in its current form.
8. Monitor product usage
It's critical to understand what's working. Spot the power users and see how your product is being used by them. Talk to them and understand what's working. They can really change the trajectory of your product if they're happy with it.
9. Go back to step 1 and keep iterating
Don't stagnate. Don't let your product becomes irrelevant. You need to keep being useful to your customers. This means constantly iterating on becoming better. Listen to your users/customers and see how you can embed your product even deeper into their daily life/work.