Failed CRM rollouts are rarely dramatic. Almost nobody cancels the subscription in a fury. What happens instead is quieter and much more expensive: the system stays, the invoices keep clearing, and the actual business slowly relocates back into group texts and one person's memory. The CRM becomes a place where the office types things in afterwards, which is the most costly possible outcome — you are now paying for software and doing double entry.
The good news is that this drift is visible early and it is caused by a short list of things. Six signs, in roughly the order they appear.
Sign 1: everything was migrated
The most reliable predictor of a failed rollout is an ambitious migration. Six years of contacts, every closed job, all the old notes — imported faithfully, and now the system is full of dead records that make every search worse and every list untrustworthy.
Move active customers and open quotes. Archive the rest to a CSV you keep, and accept that you will almost certainly never open it. A system that is 90% dead weight teaches everyone in week one that it cannot be trusted, and that lesson does not get un-taught.
Sign 2: it went live everywhere at once
Big-bang rollouts fail because they ask a business to change six habits simultaneously, during a normal working week, with no one having built confidence in any single part. When something goes wrong — and something always does — nobody can tell which change caused it, so the whole thing gets blamed.
The version that works turns on exactly one automation, the one addressing your largest leak, and lets it run for two weeks before anything else is switched on. It feels slow and it is dramatically faster in practice, because the first win buys you the credibility to ask for the second change.
Sign 3: the field is not in it
If technicians are still texting photos to the office, the rollout has already failed and everybody is being polite about it. The record is being assembled second-hand, hours late, by someone who was not there.
The cause is almost always friction rather than attitude. Watch one technician do one job and count the times they reach for their personal phone instead. Every instance is a place where your system asked for more effort than a text message did — and no amount of training beats a workflow that is genuinely slower.
Sign 4: the old system is still editable
Running the spreadsheet alongside the CRM for two weeks as a read-only backup is prudent. Leaving it editable is fatal, and the reason is not discipline — it is that under pressure, people use whatever is fastest, and the old system is always fastest because they already know it.
Set a date. On that date the sheet becomes read-only, and a week later it is archived. If that feels frightening, the fear is data telling you the migration was not finished, which is a specific and fixable problem rather than a reason to delay indefinitely.
Sign 5: nobody owns it
Software with no owner degrades. Not through malice — through a hundred small unresolved questions: where do we put commercial jobs, what do we do with the recurring customer who books by email, who fixes it when a sequence sends twice.
The owner does not need to be technical and it should not be the busiest person in the company. They need thirty minutes a week and the authority to decide how things are done. Without that role, every ambiguity resolves itself into "I will just do it the old way", one person at a time.
| Question that needs an owner | What happens without one | Cost by month three |
|---|---|---|
| Where does an odd job type go? | Three people invent three conventions | Reports nobody believes |
| Who fixes a misfiring automation? | It gets switched off | The leak reopens silently |
| What do we do about the old email list? | It stays in Outlook | Two customer databases |
| Who trains the new hire? | They shadow whoever is free | Workarounds get taught deliberately |
Sign 6: no one can name what improved
Ask anyone in the business what is better since the CRM went in. If the honest answer is "it is more organised", the rollout has no anchor and will lose the first argument it has with a busy week. If the answer is "we have not missed a quote follow-up since June" or "our median response time went from four hours to six minutes", it is safe.
This is why naming the leak before you buy matters so much. The named leak becomes the measurement, and the measurement is what carries the change through the season when everybody is tired.
TARGETS
The recovery plan
If you recognise three or more of the six, do not restart. Restarting is what businesses do instead of narrowing, and it fails the same way twice. Narrow instead:
- 01Pick the single biggest leak again
Not the whole system — one leak, named in a sentence with a dollar figure attached. Everything else is out of scope until this is working.
- 02Archive aggressively
Get every record out of the active system that nobody will act on this quarter. Search quality is the thing that most quietly determines whether people trust the tool.
- 03Go and watch the field for one day
Not a meeting, not a survey. Watch, count the workarounds, and fix the three biggest sources of friction in the order they cost you the most taps.
- 04Name an owner and give them thirty minutes a week
Publicly. The role is to answer questions and decide conventions, not to do data entry.
- 05Put one number on the wall
Whatever the leak was, measure it weekly and show everyone. A rollout with a visible number recovers; a rollout justified by "being more organised" does not.
None of these are technology problems, which is the whole point. The software was probably fine. What failed was scope, sequence, and ownership — all three of which are recoverable in a fortnight if you are willing to do less rather than more.
Common questions
Why do CRM implementations fail?
Rarely because of the software. The usual causes are migrating too much data, switching everything on at once, leaving the old system editable, the field not actually using it, nobody owning day-to-day decisions, and no single named number that proves it worked.
How do you know a CRM rollout is failing?
The clearest early sign is technicians still texting photos and updates to the office in week three, which means the record is being assembled second-hand. The second is that nobody can name a specific thing that improved — "more organised" is not an anchor and will not survive a busy week.
How much data should you migrate into a new CRM?
Only what someone will act on within ninety days: active customers and open quotes. Everything else belongs in an archived CSV. A system full of dead records teaches the team in week one that its lists cannot be trusted.
Should you restart a failing CRM rollout?
No — narrow it instead. Pick one leak, archive everything inactive, spend a day watching the field and removing friction, name an owner with thirty minutes a week, and publish one weekly number. Restarting usually reproduces the same failure at the same scale.