I ask every new client the same early question: "If I needed a list of every active customer, with their current contact details and what they're paying you each month — how long would it take to produce that list, and how confident would you be in it?"
The answer to that question tells me almost everything I need to know about the state of their data — and, by extension, how ready they are to start automating anything.
The answer is usually some version of: "It would take a while, and I'd want to double-check it."
That's the data problem in plain language. Not a technical failure. Not a system failure. A process failure: the business has been operating without a single, reliable place where customer information lives.
What a broken data source looks like in practice
I worked with a security systems business that had been operating for a number of years with a good client base — recurring subscriptions, steady monthly revenue, solid word-of-mouth. When we started working together, it became clear that their customer data was living in a spreadsheet that had been edited by multiple people over time, with no consistent process for how new customers were added or how changes were recorded.
Some customers had two entries — a legacy record and a current one, neither clearly marked as authoritative. Some entries had contact details that hadn't been updated in years. Some entries were missing key information entirely. And crucially, there was no systematic way to track monthly recurring revenue: which customers were on active contracts, what each was paying, and when each subscription was up for renewal.
The business had several hundred customers. It couldn't say with confidence which of them were active, what the total contracted monthly value was, or which renewals were coming up in the next 90 days. That's the kind of gap I explore further in the revenue hiding in broken data.
Why "good enough to operate" isn't good enough for AI
There's an important distinction between data that's good enough to run the business day to day, and data that's good enough to automate processes on top of.
When a human is in the loop, imperfect data gets corrected in real time. You notice the wrong name. You know that the email for this customer is out of date because you remember they moved. You hold the context in your head that the system doesn't hold.
When you automate, that human correction disappears. The workflow reads the data as it is and acts accordingly. If the email is out of date, the automated renewal reminder goes to an inbox nobody checks. If the customer has two records, they might get two reminders — or none, depending on which record the workflow reads. If the contract value is wrong, the revenue report you're trying to automate is wrong at scale.
The AI doesn't know. It acts with confidence on whatever it reads. That's why most AI implementations fail for data reasons, not technology reasons.
What an authoritative source actually is
An authoritative source isn't a perfect database. It's simpler than that: it's the place you've agreed is the truth, maintained by a defined process.
For a small business, this usually means:
- One place for customer data. Not a spreadsheet and a CRM and a folder of emails. One system, defined as the master, with everything else either feeding into it or being secondary.
- A consistent intake process. Every new customer gets added the same way — same fields, same format — by whoever is responsible for that step.
- Someone responsible for accuracy. Not a committee. One person who can be asked "is this record current?" and is expected to know.
- A process for changes. When a customer moves, changes their email, or upgrades their contract — there's a defined step for updating the master record.
That's it. That's master data management at SMB scale. The wealth management firms and financial institutions have entire programmes built around these principles, because the consequences of getting client data wrong in a regulated environment are severe. But the underlying logic is the same whether you have ten clients or ten thousand.
The security business outcome
For the alarm installer, the fix wasn't complex — but it was methodical. We established a single authoritative spreadsheet (with a roadmap to move to a proper CRM), defined the minimum fields that had to be complete for every customer, and built a simple intake process so that every new customer was added consistently. We also worked through the existing records systematically, filling in gaps, deduplicating entries, and flagging the ones that genuinely couldn't be reconciled.
The business now has a clear picture of its monthly recurring revenue. It can identify renewals coming up in the next 90 days in minutes. It knows which customers don't have a valid email address on file, and there's a process to update those records before they matter.
That foundation — clean, current, single-source customer data — is what makes everything else possible. The automated renewal reminders we built later work because the data they're reading can be trusted.
How to start
If you don't have a single authoritative source for your customer data, the right first step isn't buying a new CRM. It's answering three questions:
- Where does your customer data currently live? List every place — CRM, spreadsheets, email threads, accounting software.
- Which of those is the most complete and most current? That's your candidate for the master.
- What's missing from it that would make it genuinely authoritative?
The gap between where you are and where you need to be is almost always smaller than it looks. But you do have to close it before AI can help you. Use the data readiness checklist to work out what's usable now.
If you're heading toward a sale rather than day-to-day operations, the same exercise matters for a different reason. A buyer's due diligence team will ask exactly the question I opened with — can you produce a reliable customer list, MRR figure, and renewal schedule, and how quickly. "It would take a while, and I'd want to double-check it" is not an answer that survives a data room. Fixing this before you're mid-transaction is materially cheaper than fixing it under deal pressure.
If you'd like help mapping that gap, book a free call. We'll work out what your data actually looks like and what it would take to build something you can trust.