Why your systems disagree on how many customers you have

David Helon, Co-founder and COO·3 min read

Here's a test you can run this afternoon. Ask three people how many customers the company has: one who works in the finance system, one in the CRM, one in the customer-success tool.

You'll get three different numbers.

I can guarantee you that in any company you go to, the number of customers in the finance system is different from the CRM, which is different from the customer-success tool. Most owners believe their data is in better shape than it is, until they ask.

Why the numbers drift

Mostly it happens because each system was set up to do its own job, and nobody owns keeping them in step.

  • Names don't match. One system has "Acme", another "ACME Inc.", a third an account number.
  • Parent and child accounts are recorded differently in each platform, or not at all.
  • Duplicates pile up. Old records that should have been purged never are.
  • "Customer" means different things. Finance counts who paid. The CRM counts who signed. Customer success counts who's active. Customers, users, active users and subscriptions can be counted dozens of ways.

Financials will always be different across platforms unless someone is actively keeping them in sync. In most companies of 20 to 200 people, it's nobody's actual job.

The number can be wrong in two places at once

I've seen this more than once. The CEO tracks MRR in a spreadsheet, and the spreadsheet is wrong.

Then you check the billing platform and find it wasn't set up correctly when it was implemented, so its MRR is wrong too. Fixing it means going back into the implementation, correcting it, and proving the number down to the penny. Then you can retire the spreadsheet.

Most companies never find out, because nobody checks the spreadsheet against the source.

Why it stays broken

Smaller companies, say 20 to 50 people, don't have anyone whose job is the data. They react to whatever the CEO or the board asks for.

Bigger ones, 51 to 200, usually have a couple of people in operations handling it, but not strategically. They have a grand idea of what they want, and when they cost it out it's a big number, something like $500,000, and there's no way they can go ahead with that plan.

So the three numbers stay three numbers. And now every AI tool the team uses inherits the disagreement, because each one is connected to a different system.

One count, kept current

What it takes is unglamorous:

  1. Bring the systems together. Connect finance, the CRM and the customer-success tool to one place.
  2. Recognize the same customer across all three, and check the links before trusting them. altspark finds related records and shows you inferred links to confirm or correct.
  3. Keep it current on a schedule you set, so Friday's count reflects the latest sync, not Monday's export.
  4. Ask once. Everyone asks from any AI tool that supports MCP, through the project's own server, so every answer comes from the same data.

Your systems stay the system of record. altspark keeps a synced copy of what you choose, so the answer can be checked against the source.

What this looks like in your seat

If you run the company, you stop getting three customer counts in the same board meeting.

If you run operations, every region or team asks the same server, so they work from the same data.

If you run finance, total spend with one vendor across three systems that spell its name differently becomes one question.

If you hold the credentials, there's the project's own server for everyone to ask, with access set per connection and roles deciding who sees which data, instead of a connector per person.

Count your customers

Run the test this week. Ask the three systems. If the numbers match, I'd like to hear how you did it.

If they don't, you already have the whole answer. It's just in three places.