I've read thousands of cancellation responses. Not because I enjoy it — because I built a tool that analyzes them, and I needed to understand what patterns actually matter.

Here's what I've learned: most founders dramatically undercount the number of distinct churn drivers in their data. They see "pricing" and "competitor" as the big buckets and stop there. But the actual patterns are more nuanced, and the ones you miss are usually the most actionable.

1. The "Value Gap" Is Not the Same as "Too Expensive"

When a customer says "too expensive," most founders hear a pricing problem. But in the data, there are actually two distinct patterns:

Absolute pricing: "$X/month is more than my budget" — This is about ability to pay.
Value perception: "Not worth $X for what I get" — This is about what they receive for the price.

The distinction matters enormously. If most of your "pricing" churn is actually value perception, lowering your price won't help. You need to either increase perceived value (better onboarding, more visible features) or restructure pricing so customers only pay for what they use.

When we run these through RetentionCheck, the AI separates these automatically. In a typical B2B SaaS dataset, about 60-70% of "pricing" responses are actually value perception issues.

2. "Competitor Switching" Tells You What Feature to Build Next

"Switched to [competitor]" seems like a dead end. They're gone, and you can't control what competitors do. Right?

Wrong. When you analyze the competitor mentions in aggregate, you find a roadmap:

Feature consolidation: "Notion does docs AND tasks" — Customers want fewer tools, not better tools
Specific capability gaps: "Linear is just faster" — One specific attribute that's table-stakes
Pricing competition: "Their free tier is better" — Your free/starter plan isn't competitive

The competitor mentions tell you which features are now table stakes, which capabilities you need to add, and where your packaging needs work. This is free market research hiding in your churn data.

3. Support Failures Are Always Underrepresented

Customers who churn because of bad support rarely say "bad support." They say:

"Took 5 days to get a response"
"Nobody got back to me"
"Same bugs, same 'it's on the roadmap' replies"

These get miscategorized as "bugs" or "reliability" when the actual issue is responsiveness. In our analysis, support-related churn is typically 2-3x higher than what founders estimate, because the responses don't use the word "support."

AI analysis catches this because it reads the full context of each response, not just keyword-matching. A response about a bug that went unfixed for months is a support failure, not a product quality issue.

4. "Involuntary" Churn Has a Preventable Subset

Responses like "company shut down" or "got acquired" seem unpreventable. And some of them are. But within the involuntary churn bucket, there are usually recoverable customers:

"Budget cuts" → Could be retained with a discounted tier or annual prepay option
"Project ended" → Could be retained with a pause option instead of cancellation
"Got acquired" → Could be retained by selling the parent company on your tool
"Changed roles" → Could be retained by multi-seat or org-level accounts

We typically find that 30-40% of "involuntary" churn has a retention intervention available. Most companies write off the entire category.

5. The "Ghost Pattern" — What Nobody Mentions

The most interesting pattern isn't in what customers say — it's in what they don't say. If you have 100 cancellation responses and nobody mentions onboarding, that doesn't mean onboarding is great. It might mean customers who had bad onboarding experiences never got far enough to churn — they just never activated.

Look at the gap between your churn data and your activation data:

High signup, low activation, no mention of onboarding in churn → Your onboarding is silently killing growth
High activation, high churn, no mention of bugs in churn → Customers aren't churning because the product is broken — they're churning because it's not differentiated

The absence of a pattern is itself a signal. AI analysis helps here because it can flag when expected categories are missing from the data.

How to Find These Patterns

You don't need to read thousands of responses like I did. Paste your cancellation feedback into RetentionCheck and the AI will surface these patterns automatically — including the nuanced distinction between pricing vs. value perception, the specific competitors being cited, and the hidden support failures.

It takes 30 seconds. The insights might change how you prioritize your next quarter.

Check out our example analyses to see these patterns in action across different business types.