Field note
The Category Nobody Named: Retention Execution
Every company measures churn. Almost none execute the save. The gap between knowing and doing is a category that has been empty.
Every company has a churn number. Almost none have a churn team.
They have dashboards. They have scores. They have Slack alerts that tell someone, somewhere, that a customer looks at risk. What happens next is improvised. A CSM fires off an email. A discount goes out because it always goes out. A renewal comes up and no one is prepared. The churn happens anyway. The post-mortem says "we need better tooling."
The tooling was never the problem.
What prediction promised and didn't deliver
The last decade gave us remarkably good churn prediction. Models got sharper. Data got cheaper to move. Every billing platform added a risk score. Retention software learned to surface the accounts most likely to cancel.
None of it moved the churn number.
The reason is obvious in retrospect: prediction is not execution. A model that tells you someone is 73% likely to churn has done nothing for you until a human picks it up, writes a message, finds an offer, sends it at the right moment, and follows up when it lands. That sequence takes time your team doesn't have and judgment your tooling can't supply.
The insight was there. The action was missing.
The gap has a name
What sits between knowing and doing, between the risk score and the saved account, is a function that most companies have never staffed. It is not customer success. CS protects your highest-value accounts through relationship management. It does not have bandwidth for the mid-tier, the self-serve, the long tail. It is not a loyalty program. Loyalty rewards behavior that is already happening; it does not intervene when behavior is about to stop.
This function is Retention Execution: the systematic, repeatable, measurable work of taking every at-risk signal and turning it into an account that stays.
Not a dashboard. Not a score. The work.
Why nobody built it
The market got split. One side built insight tools for enterprise buyers who could afford analysts and CS teams to act on the output. The other side built loyalty and messaging platforms for SMBs who could not afford the overhead. Neither side built the execution layer: the thing that actually runs the save at scale, on every account, and proves whether it worked.
The split made each tool defensible in its lane. It also left a gap large enough to build a company in.
What Retention Execution actually looks like
It starts with agent teams that read your data: billing events, usage signals, support tickets, behavioral patterns. They build a complete picture of every account. Those agents detect the risk before the cancel request, craft the save play specific to that account's situation, and run the intervention with every customer-facing action human-approved before it ships. Outcomes feed back into the model so it gets sharper over time.
Concretely: three weeks before a renewal, you know a mid-market account has dropped usage and the last CSM note was eight months old. You do not file a ticket. You do not set a reminder. You execute the save, human-approved, and the account shows up in your retained book.
This is a function measured on retained accounts you can name, not on activity.
The credibility problem
There is one question every retention buyer eventually asks: did you actually save those customers, or would they have stayed anyway?
It is a good question. For most tools, the honest answer is: we do not know. A headline save rate conflates re-engagement with natural retention. A vendor that reports a single percentage may have done nothing; those customers may have renewed regardless.
The answer to this is not a better dashboard. It is the work on your own book, measured on your real accounts. You see every save the agents worked, the signal that flagged each account, and the play that brought it back. Every customer-facing action is human-approved before it ships, so nothing is claimed that you did not approve.
A three-week proof process settles the question directly. The team works your real at-risk accounts alongside yours, and you watch the saves land on names you recognize.
Retention Execution, done correctly, runs the save and shows you the result.
The category is open
Prediction tools won the last decade by naming their category first. The companies that built "customer health scoring" owned that market before anyone else described the problem.
Retention Execution is the next category. The problem description is already in the buyer's language. They know their churn score does not pick up the phone. Their CS team cannot cover every account. They cannot prove their saves are real. The solution is a function most of them have never had.
The company that best describes this problem will be assumed to own the solution.
That description is Retention Execution. The floor is open.
For a detailed look at how the execution layer works in practice, including Sign-off and how saves are measured on your real accounts, see the Platform page.