Most broken CRMs did not start out broken. They started clean, then got worse gradually as fields were added without being removed, imports created duplicates, deals were left open past their close date, and small workarounds compounded into a system nobody trusts.
By the time someone decides it needs to be fixed, the natural instinct is to rip it out and start fresh. In almost every case that is the wrong move. A CRM cleanup, done in the right order, gets you a trustworthy system in weeks rather than months and without losing your existing history. What follows is the sequence I use.
1. Start with an audit, not a cleanup
Do not touch any data until you understand what you have. The temptation to start merging duplicates or deleting records on day one is strong, and it always creates more problems than it solves.
Instead, spend the first few days running reports. How many total contacts, companies, and deals do you have? How many deals are open, and how old are they? What percentage have missing required fields? How many duplicates exist, and by what criteria? What automations are currently running, and are any of them failing silently?
The audit gives you a baseline. Without it, you cannot tell whether your cleanup is working, and you cannot explain to leadership why the pipeline number is about to drop.
2. Fix the pipeline stages before anything else
Pipeline stages are the frame everything else hangs on. If the stages are wrong, every downstream fix compounds the problem.
Good stages describe what the buyer has done, not what the seller has done. "Sent proposal" is a seller action. "Proposal reviewed by economic buyer" is a buyer state. The difference matters because buyer states are objectively verifiable and predict conversion, while seller actions do not.
Most SMB CRMs I audit have too many stages, and the definitions are vague enough that reps disagree about which stage a given deal belongs in. Cut to five to seven stages, write a one-sentence definition of each, and specify the exit criteria that must be true before a deal advances.
3. Merge duplicates
Duplicates are the single largest source of unreliable reporting. When one company exists as three records, activity is split across all three, and no one report shows the complete picture. Same with contacts.
Every CRM has native tools for finding and merging duplicates. Use them. Start with companies, because contact and deal duplicates often resolve automatically once companies are consolidated. Set a rule for which record survives (usually the oldest, with the most associated activity) and stick to it.
This step is unglamorous and takes real hours. It is also the step that moves the needle most on report accuracy.
4. Trim the fields
Most CRMs I audit have somewhere between 40 and 100 custom fields, of which fewer than a third are actively used. The rest sit empty on every record or contain values so inconsistent they cannot be reported on.
List every custom field. For each one, ask two questions: is anyone using this to make decisions, and is the data in it reliable? If the answer to either is no, hide or delete the field. Rebuilding it later is far easier than fighting through a hundred fields to find the ten that matter.
This step is also where reps notice something has changed for the better. Fewer fields means faster data entry and less cognitive load.
5. Close out the stale pipeline
Every broken CRM has a large chunk of pipeline that is not really pipeline. Deals sit at "sent proposal" from four months ago with no activity. The rep has mentally moved on but has not closed them out because closing a deal as lost feels like failure.
Set a rule: any open deal with a close date more than 30 days in the past and no logged activity in the last 30 days gets closed as lost, unless a rep can defend it in one sentence. Batch this in a review session with the sales team rather than doing it unilaterally. Reps need to see the logic and buy in.
Your pipeline number will drop, sometimes dramatically. That is the point. What is left is closer to what is actually in play.
6. Set new data entry requirements
Cleanup only lasts if you fix the input. Otherwise the CRM degrades right back to where it started within six months.
Pick the small number of fields that genuinely matter, and enforce them at stage transitions rather than at deal creation. A rep should be able to create a new deal in seconds. To move it to a later stage, they must fill in the fields that make that stage report correctly.
Do not overdo this. Every required field is friction. Ask yourself whether the data would ever be used, by whom, and for what decision. If you cannot answer, do not require it.
7. Give the team a running start
The final step is not technical, it is human. A cleaned-up CRM will only stay clean if the team understands what changed, why, and what is expected of them going forward.
Run a 30-minute session covering the new stage definitions, the fields that are now required, the ones that were removed, and where reports now live. Follow up with a one-page reference document. Then check in weekly for the first month to catch any drift before it becomes habit.
Momentum matters here. If you fix the system but never adjust behaviour, the cleanup was theatre.
Getting it done
Most CRM rescues can be done in three to six weeks with focused attention. The blockers are almost never technical. They are the audit that gets skipped, the difficult decisions about what to delete, and the team conversation about how things will work going forward.
This is a core part of what Fulcrum does. We audit the current state, run the cleanup in the right order, and hand over a CRM your team actually uses. If you are looking at a CRM that has stopped serving the business, our Health Check will tell you where to start. Or if you'd rather talk it through, book a free discovery call.