We track 42 active ghostwriting clients at EcomGhosts. Over 24 months we've watched 11 cancel. The notice arrives in month 7 or month 13 — but the behavior that predicts it shows up 90 days earlier, every single time.
This is the opposite of the "client expansion signals" post we wrote earlier this month. Same data set, opposite tail. What clients stop doing in the quarter before they leave.
We treat these four signals as a rolling watchlist now. When two of them light up on the same account, we pull the founder onto a 30-minute call inside two weeks. Of the last 9 accounts where we did this, 7 stayed past 18 months.
Signal 1: DM forward velocity drops below one per 14 days
Healthy ghostwriting clients forward inbound DMs to us within minutes of receiving them. The DM hits their phone, they screenshot or paste, we get it before the lead cools.
For our top quartile of clients, median forward time is under 90 minutes and volume runs 3–8 forwards per week.
The pre-churn pattern is specific. Forward volume doesn't go to zero — it drops to one every 10 to 14 days. Founders who are about to cancel don't say "the leads stopped" — they quietly stop sending them to us. Sometimes because they're handling them in-house. Sometimes because the founder has stopped reading their own inbox. Sometimes because they've decided the leads aren't valuable and they haven't told us yet.
The number to watch is forwards per week per active client. When it falls below 1.0 and stays there for three weeks, the next 90 days are at risk.
Signal 2: Edit feedback becomes generic
Specific edit feedback is the strongest retention signal we measure. "Change 'we' to 'I' in the second paragraph." "This number is wrong — it was 14.2%, not 14%." "Don't open with 'I've been thinking.'"
Founders who are paying attention give us edits that name a clause, a number, or a structural choice. Across our retained book, the median edit message contains 2.4 specific corrections.
Pre-churn edits flatten. They become "looks good," then "approved," then "yeah send it." Or they switch from inline corrections to a single line: "make it punchier" or "tighten this up."
The mechanic is simple. The founder has mentally checked out of the content. They're no longer running each post through their own filter — they're rubber-stamping or rejecting on vibes. Specific edits require thought; generic edits require a thumb-tap.
When median specific-corrections-per-edit drops below 0.6 for three consecutive weeks, the account is at risk.
Signal 3: Voice-sync attendance becomes rescheduled, then "let's do async"
We run a 30-minute voice sync every other week with each active client. It's how we capture new framings, fresh client stories, and the founder's current obsessions. It's the input that keeps the output sounding like them.
Healthy clients miss maybe one sync per quarter and reschedule the same week. They show up with 2–3 things they specifically want to talk about.
Pre-churn pattern: the sync gets pushed once, then pushed again, then "let's just do this async — send me 5 questions and I'll record voice memos." The voice memos arrive late, get shorter, and start sounding generic. Within 8 weeks of that switch, 64% of accounts in our churn cohort had cancelled.
The async-only request looks like a productivity move. It's almost always a disengagement signal. Founders who are leaning in want the live conversation because the conversation surfaces things they didn't know they were going to say.
Signal 4: Zero sales-asset pull-through for 30+ days
Our highest-LTV clients pull the LinkedIn content into their sales process. They paste posts into proposal decks. They send specific articles to inbound leads as a credibility prop. They reference the writing in discovery calls.
We track this through one question on our monthly client check-in: "Where did this month's content show up outside LinkedIn?" Healthy accounts list 3–5 places. The post got pulled into a deck, forwarded to a prospect, quoted in a podcast pitch, used as a sales artifact in a renewal.
When the answer becomes "nowhere" or "I'll have to think about it" for two months in a row, the ROI story has gone dim in the founder's head. They're no longer using the content as a business asset — they're treating it as a marketing line-item. Marketing line-items get cut. Business assets don't.
This is the most predictive single signal in our data. 78% of accounts that hit 60 days without sales-asset pull-through cancelled within the next 90 days.
What we do when two signals light up at once
We've built a weekly review where we score each account on these four signals. Any account scoring 2 of 4 in the red gets a founder call inside 14 days, and the call has one agenda: what would have to be true for this content to be worth 10x what it costs.
That call is not a save attempt. It's a re-scoping. Sometimes the answer is to cut cadence, change format, or refocus on a different audience. Sometimes the answer is that the founder genuinely doesn't need ghostwriting anymore and the right move is to part well. Either outcome beats the surprise cancel.
The agencies that lose clients in this space are not losing them on quality. They're losing them because the customer disengaged in month 6 and no one noticed until month 12.
FAQ
How early can these signals show up? We've seen Signal 4 (sales-asset pull-through dropping to zero) appear as far out as 5 months before cancellation. Signal 1 (DM forwards) typically appears 8–12 weeks before. Signal 2 (edit specificity) and Signal 3 (voice sync drift) tend to cluster in the 60-day window.
What's the false positive rate? About 22% of accounts that light up 2 of 4 signals do not cancel within 90 days — but in nearly all those cases, the founder call surfaced a real issue we then fixed. We treat the false positives as wins.
Does this apply to non-ghostwriting client relationships? The structure does. The specific signals are LinkedIn-content-shaped. The underlying pattern — disengagement shows up as quieter behavior before it shows up as cancellation — applies to any retainer business.
If you're running a content service for ecom founders and want to compare notes on the leading indicators we track, the team at EcomGhosts is happy to swap data.