The cancellation email is not the event. It is the paperwork.
The decision was made weeks earlier, in a conversation you were not in, and by the time the email arrives you are not doing retention any more. You are doing recovery, and recovery works far less often than anybody selling retention software will tell you.
Which means the useful question is not how to save them. It is what you could have seen in the six weeks before, and the answer is that you could have seen quite a lot.
The signals were not in your product analytics. They were in your inbox and your calendar, which is exactly why software misses them and a person does not.
The signals are in your inbox, not your dashboard
Five things, all observable without asking anybody anything, and all free.
Response latency from their side. They used to come back within two hours. Now it is three days, and your last message has not been answered at all. This is the single best signal available and it is sitting in your sent folder, uncounted. Nobody tracks it because it is not a field in any system, which is also why it still works.
Who is on the email. Your champion drops to cc. A name you have never seen appears. The relationship has been delegated downward, and delegation downward is the clearest pre-cancellation signal in a service business, because nobody delegates something they are personally invested in.
The shape of the questions changes. This is the one I would watch most closely. A healthy client asks “could you also look at X”. A client who has started thinking about leaving asks “could you send a summary of everything delivered this year”.
That second question is almost never curiosity. It is being asked because somebody upstream asked for it, and a documentation request is a procurement signal wearing an administrative costume. When you get one, the decision window is already open.
Meeting behavior. The monthly call gets rescheduled, then rescheduled again, then becomes “let us just do this over email”. Each step is reasonable and the sequence is not.
The extras stop. In a working engagement, small additions keep appearing, because the client keeps thinking of things. When the extras stop entirely and the scope goes exactly flat, the relationship has been frozen pending a decision somebody has already started making.
One related thing worth saying plainly, because it inverts most people’s instinct. A complaint is good news. Complaining costs the customer time and social energy, which means they are still investing in the relationship and still expect it to continue. The client who has stopped complaining is not satisfied. They have stopped bothering, and that is much later in the process.
About the published warning windows
You will see numbers for how far ahead this is detectable, so here is the honest state of them.
Published figures cluster around one to three months: thirty to ninety days in some write-ups, four to eight weeks in others. Every one I tried to trace led back to a vendor page selling churn software, with attributions to large research firms and in one case to a journal paper, and none of them to anything I could actually read. So I am not going to quote you a figure.
The direction is all you need. There is a window, it is measured in weeks rather than days, and the entire question is whether anybody at your company is looking during it.
A five-minute review, done by a person
Here is the whole process, and it is deliberately not software.
Once a month, for each account that matters, one person spends five minutes answering three questions. How long is their current reply time compared with three months ago. Who is on the last three emails. When did they last ask for something new.
Write the answers in a sheet. Three columns, one row per client, updated monthly.
That is it. The reason it works is not sophistication. It is that these three facts are never written down anywhere, so no trend in them is ever visible, so the drift happens in full view of people who are each looking at a single email.
A client whose reply time has tripled, whose champion has gone to cc, and who has not asked for anything new in two months is not a maybe. Go and have the conversation this week, before the summary request arrives.
Ask the five who already left, and ask it differently
Worth doing once, properly, and it is a different exercise from asking about lost deals.
I went through why the person with access to the buyer is usually the one whose judgment is on trial in the context of lost deals, and the who-should-ask problem is the same here, so read that for the staffing of it.
What is different is the trap. A customer who has already left is motivated to be kind. They may want the door open, they may know you socially, and they have nothing to gain from the truth. So the default answer is “it was a budget thing”, and the default answer is usually a courtesy.
Two questions get around it.
“When did you first start thinking about stopping?” This dates the decision, and the date is almost always well before anything you noticed. It also moves them off justification and into recall, which is a different and more honest mode.
“What were you asked internally that you could not answer?” This is the one that produces the real reason, and in my experience the real reason is rarely price. It is that at some point somebody senior asked what we are getting from these people, and your contact could not answer in a sentence.
That is a positioning and reporting failure rather than a service failure, and it is fixable in a way that “they went cheaper” is not.
Which saves are worth attempting
Not all of them, and chasing the wrong ones is how retention budgets get wasted.
Worth a real effort: the client whose champion changed, where the work is genuinely good and the new person simply inherited a vendor with no story attached. Go and give them the story, framed as their win rather than yours.
Also worth it: the one who left over a specific unresolved failure. They will tell you exactly what it was, and if you have fixed it, that is a reason to talk rather than a plea.
Let go: anybody who left on price after you had the pricing conversation twice. Nobody switches back for a discount, and offering one teaches your remaining clients what the real price is.
Also let go, and this is the harder one: the client who was never a fit and whose departure quietly improved your week. Count them separately in your churn number rather than mourning them, because mixing them in makes the number say nothing.
Before lunch
Thirty minutes, and you can do all of it from your own email.
Open your last three email threads with your five biggest clients and write down their reply times. Compare with the same threads from spring.
Check who is on the most recent message in each. Note anybody who has moved to cc.
Then pick the one account where the pattern looks worst, and book a call this week with no agenda other than asking how it is going and what they are being asked internally about the work.
If you’d like help working out whether a client has gone quiet or just gone busy, drop me a line on email, WhatsApp, phone or LinkedIn and we can have a quick chat, and my notes on how I work cover the rest. I’m taking on a small number of contracts at the moment, so I’ll tell you straight whether it’s something I could help with.