Buying a CRM takes an afternoon. Getting people to use it is a different job entirely, and it is the one nobody plans for.
Most failed rollouts are described afterwards as a people problem. The team would not adopt it. They went back to their own notes. They never filled the fields in.
That is almost never what happened. What happened is the system asked people to trust it before it had earned any trust, and they were right not to.
Eight steps. In order, because each one makes the next possible.
01
Fix the data before you ask anyone to trust it
People stop using a CRM the first time it lies to them. One wrong phone number, one duplicate, one deal in the wrong place, and the team quietly goes back to their own notes. Clean first, launch second. Never the other way round.
02
Find the duplicates before your team does
On one job, a business ran two separate boards. Of the 1,568 people on the smaller one, 712 were already on the big one. They believed they had two sources of leads. They had one source and a copy of it.
03
Give every stage an entry rule and an exit rule
Write both down in plain words. A deal enters this stage when X has happened. It leaves when Y has happened. If you cannot write those two sentences, the stage is not a stage, and your pipeline report is guesswork.
04
Decide who owns each field, by name
Not the sales team. A person. Who is allowed to change a deal value, who fixes a wrong number, who closes something as dead. Shared ownership means nobody fixes anything, because everybody assumes somebody else will.
05
Roll it out to one team, not the whole company
Pick the group with the most to gain and the fewest habits to unlearn. Get it working there for a month. A small group that loves it will sell it to the rest of the business far better than an announcement ever will.
06
Make the right thing the easy thing
If logging a call takes four clicks, it will not happen on a busy day. Calls should log themselves. Notes should be one box, not seven fields. Every extra step is a place where the record and the reality come apart.
07
Never wire anything to a stage's ID number
Use the stage name instead. Wire your automations to an ID and the day someone adds a new stage, part of your system stops working silently. Nobody gets an error. Things just stop arriving.
08
Test every change on a copy first
Every script we run does a practice pass and shows exactly what it would change before it touches anything. That is how 601 records got corrected in one go with nothing to undo afterwards. A live first run is how you lose a week.
Moving data is not the risky part
People worry about losing records in a migration. In practice that is the most controllable part of the whole job.
We have moved 14,611 records into a single CRM, everyone with a phone number and a real call history, in one place instead of spread across lists. We have joined two separate boards into one, 856 deals moved, none lost, with a written list of every change. We have pulled 8,682 deals onto one board from three different lists and counted them all.
None of that is the hard bit. The hard bit is the week after, when somebody opens the new system, sees something they know is wrong, and decides the whole thing cannot be trusted.
How to tell it has actually stuck
Not by logins. People log in because they are told to.
Watch for these three instead. Somebody corrects a record without being asked. Somebody quotes a CRM number in a meeting and nobody reaches for a spreadsheet to check it. And the private side-notes stop appearing, because the system finally holds what people need.
When those three are true, the change has landed. Until then you have an expensive address book.