A CRM shaped around your customers, not the reverse

The cost of a misshapen CRM is invisible. It never shows up as a line in the budget; it shows up as a sales engineer maintaining a private spreadsheet because the official tool does not model the kind of deals they actually close. It shows up as a pipeline review where half the numbers are corrected out loud, and as a forecast the finance team stopped trusting a year ago.

None of this is a failure of the vendor. A generic CRM encodes the average sales process, a lead, a few stages, an amount, a close date, because that is the only process that fits everyone. The problem starts when your process is not the average, and most processes worth running are not.

Listen to one discovery call. Write down what was actually said: the constraint the customer mentioned, the person who has to sign off, the season that matters, the thing they tried before. Then open your CRM and try to record it. Every fact that ends up in a free-text notes field is a field your tool does not have.

When the answer is 'most of it', the gap is your roadmap: literally. Those fields, in that order, are the data model of a CRM shaped around your customers. Nobody has to invent it; the team has been describing it in every call for years.

This does not mean building a CRM from scratch. E-mail sync, calendars, sequences, dialers: rent them; they are commodities and the market does them well. What is worth owning is the model: the objects, the fields and the rules that make a deal in your business what it is. That is usually a small piece of software with a large effect, because every report, every forecast and every handover reads from it.

For a company selling into teams and shift schedules, that model held the sites, the shifts and the constraints per site, nothing a stage-based pipeline could express, and everything the sales team needed to quote correctly the first time. The pipeline stayed in the market tool; the model lived next to it.

Do not migrate. Keep the CRM you have for what it does well and add the model beside it, synchronised on the one identifier both agree on: the account. Start with a single object, the one whose absence costs the most, and give it to the three people who complain most. If they stop keeping their spreadsheet within a month, the model is right. If they do not, you have learned something for less than the price of a licence renewal.

The rest follows the same rule: one object at a time, adopted before the next is built, until the private spreadsheets have nothing left to hold.