Signal Mining: The fast way to find product/ market fit.

How AI can help you find "bite your hand off" customers in minutes.

Here’s a PayPal mafia story most people don’t know 👇🏼

In 1999, PayPal’s original use case was to beam money between Palm Pilots. 😂 

How did they get from that terrible idea in 1999 to an IPO in 2002?

It's easier to change your market than your product, so when Palm Pilots failed, they went looking for a new customer segment.

They studied payment messages in their system and noticed something peculiar: 43 users mentioned eBay auctions.

While 43 is a small number, they also noticed it was doubling every week. 👀

They wrote tools to message sellers, pretending to be a buyer, asking “do you accept PayPal for this auction?” And they built a feature to let sellers insert PayPal buttons into their eBay listings.

Off to the races.

Once eBay payments were scaling, they sought the next “s-curve” of growth and resumed the search for use cases.

They sent salespeople out to find more verticals desperate for better payment options, and eventually discovered gaming, adult, jewellery and watches – risky industries under-served by traditional processors.

Even though payments was a “horizontal” product, they started by finding desperate niches and expanded from there.

The PayPal Mafia understood something crucial: finding desperate users beats building perfect products.

And that's so much faster with off the shelf AI tools! Here's how you can apply their approach.

Signal Mining

Signal mining means systematically scanning qualitative data for patterns of customer desperation.

The good news – desperate customers don't hide. They complain loudly. They leave breadcrumbs everywhere – in reviews, forums, support tickets, and search queries. And now, LLMs can help you find these breadcrumbs at scale – in your own data, and out in the world. Here's how...

Mining your own data:

Upload your customer communications (support tickets, reviews, user interviews) to Claude or ChatGPT and use these prompts:

• Which customer complaints mention trying multiple solutions before finding us? What specific problems were they trying to solve?

• Find all mentions of emotional language like 'desperate,' 'frustrated,' 'finally,' or 'thank god.' What patterns emerge?

• Which use cases have the highest ratio of action words (bought, switched, implemented) vs consideration words (thinking, comparing, researching)?

Mining the market:

Before you send salespeople everywhere, find places where your prospects discuss the outcomes you enable, e.g. Reddit threads, industry forums, competitor reviews, even group chats.

• Search those conversations for frustrations with competitors or tricky tasks. Which specific use cases generate the most emotional complaints or DIY workarounds?

• What are the top 3 reasons people rage-quit [current solution]? Which user segments are most affected?

• Find niche communities building workarounds for [problem]. What have they cobbled together?

Simple next step

If you’re considering a pivot or looking for a new growth lever:

1. Pick your richest data source – support tickets, user interviews, or public discussions. Feed it into an LLM with one of these prompts.

2. Look for desperation: people who've tried multiple solutions and workarounds, emotional language, or situations where current solutions fail.

3. Then ask: Could we help this person?

Remember: Desperate customers don't need convincing – they need solutions.

Happy hunting. 👍

Helpful? Read all my posts and subscribe here.