How to Choose a CRM and Migrate to It
I ran a company that had outgrown its CRM. The evaluation took a scoring sheet. The migration took six weeks of data cleanup and a parallel run, and that is where the value was.
· 4 min read
Patrick Laughlin, Founder, Chief AI Officer
Naval Academy, class of 2011. Decorated naval aviator, 10+ years. President, 2023 Inc. 5000 company.

Before Elite Workflow I ran a services company. Revenue was growing, and every dollar of it took heroics. The sales team kept their real pipeline in their heads and their notes in a spreadsheet, because the CRM was slower than remembering.
We decided to move. The stakes felt high, because a bad choice meant six months and a budget gone.
So we built a scoring sheet with 110 criteria and ran 74 platforms through it.
What the scoring sheet was actually for
The categories were the obvious ones. Migration path and API depth. Whether it would talk to the marketing stack we already ran. Whether we could configure it without hiring a developer. Reporting depth. What it was like on a phone for a salesperson sitting in a client's parking lot. Total cost including implementation, not the license line. Vendor stability. Whether the documentation was any good, which turns out to predict a lot.
Here is the part I did not expect. After all that scoring, the top four finished within a few points of each other. Any of them would have worked.
The sheet did not pick a winner. What it did was force us to write down what we actually needed, in one document, before anyone sat in a demo being sold to. That document became the implementation plan. If I ran this again I would still do the evaluation, and I would do it for that reason rather than for the ranking.
The migration is 80 percent of the work
Choosing was maybe a fifth of the job. We ran the rest as 30 numbered steps, published where everyone could see them.
Six weeks of data audit and cleanup came first, before a single record moved. Duplicates, dead accounts, contacts with no owner, three spellings of the same client. Nobody enjoys this and everybody wants to skip it, and skipping it is how you pay a vendor to move your mess into a nicer building.
Then field mapping and the custom objects, which is where you find out that two departments have been using the same field to mean different things for four years. Then the integrations into marketing, service and accounting. Then training that was a real program rather than a one-hour demo. Then a parallel run, both systems live, until the new one won on its own. Then go live, with a rollback plan we did not need and would have been reckless not to write.
What changed, and what I will not claim
The company grew fast the following year. I am not going to tell you the CRM did that. Two dozen things were happening at once and I cannot separate them, and any consultant who hands you a straight line from a software migration to a growth number is selling you a story.
The number I will defend is smaller and it came afterwards.
Once the data was in one place and clean, I automated invoicing and collections. Our clients had been taking about 74 days on average to pay a 30-day invoice. After the automation that came down to 36. The last time I checked, it was near 25. We had budgeted an administrator to absorb the growing invoicing load and we never made the hire.
From 74 days to about 25 is a working capital change of about seven weeks on every invoice we sent, and it came from a scheduled job and a set of templates. No model. No AI. It was possible only because the underlying data had been cleaned first, which is the actual lesson.
Why I keep telling this story to AI buyers
An AI rollout has the same shape as a CRM migration, and it fails for the same three reasons.
Nobody owns the data, so the system inherits four versions of the truth and confidently picks one.
Training is a demo, so people go back to the spreadsheet in week three and nobody finds out for a quarter.
And there is no parallel run, so the new way has to be perfect on day one instead of merely better by day 30.
Three things I would do again on any system that touches how work moves. Write down what you need before you shop, and let that document run the project afterwards. Clean the data before you move it, on a schedule, as its own line item with its own owner. And run the old way beside the new one until the new one wins without anyone being told to prefer it.
Most companies think they are buying software. What they are really doing is redesigning how work moves through the building. Price it that way and you will be roughly right.


