When I talk to business owners about their data, the conversation often starts in the same place: they know it's not great, they've been meaning to sort it out, but it hasn't felt urgent. The business is running. Customers are paying. The admin is manageable.
What usually changes that calculation is when we sit down and work out what poor data quality is actually costing them. Not in vague terms. In money.
Almost every time, the number is bigger than they expected.
The alarm installer example
I worked with a security systems business — alarms, monitoring, recurring subscription model — that had been operating successfully for years. Good reputation, steady client base, solid word-of-mouth growth. The revenue was there. But the data behind that revenue was in a broken spreadsheet.
Multiple people had edited it over time. There was no consistent method for adding new customers. Some customers had duplicate entries. Some entries were missing contact details. Critically, there was no reliable way to track monthly recurring revenue: which contracts were active, what each customer was paying, and when each subscription was due for renewal.
The business had several hundred customers. When we started working through the data, the financial picture that emerged was concerning — not because the business was in trouble, but because the gaps in the data represented real revenue risk.
Missed renewal reminders to customers whose contact details hadn't been updated. Subscriptions that had lapsed quietly because nobody had a systematic way to notice. Customers who should have been receiving service calls who weren't being contacted because their records were incomplete.
Conservatively — and I want to be precise about what I mean by that — we identified thousands of pounds in potentially missed or at-risk revenue. Spread across hundreds of customers, the incremental losses are small individually. Collectively, across a full year, they're significant.
The four categories of data-quality revenue cost
In my experience, the revenue impact of poor data quality almost always falls into one of four categories.
Revenue you're not collecting. Work that's been completed but not invoiced because the job record is incomplete. Customers whose price hasn't been updated after an agreed increase. Subscriptions where the direct debit was cancelled months ago and nobody caught it. This is where automating invoice chasing pays back fastest.
Revenue you're about to lose. Customers approaching a renewal date who won't receive a reminder because their contact details are wrong. Contracts that lapse by default because no renewal process exists. Relationships that go cold because outreach is being sent to an outdated email.
Time cost that converts to money. Hours spent reconciling records, hunting for information that should be retrievable in seconds, manually cross-checking two systems that should agree. A business owner spending five hours a month on data reconciliation that a clean system would eliminate is spending real money on a problem that shouldn't exist.
Decision cost. Management reports built on inaccurate data leading to wrong decisions: overestimating revenue, missing a margin problem because the cost base isn't captured properly, making staffing decisions on numbers that don't reflect reality. The CA in me pays close attention to this one — reconciled data feeds reliable reporting, raw exports don't.
The fix, and what it enables
For the alarm installer, the approach was methodical rather than complex. We established a single authoritative record for every customer — working through the existing data to fill gaps, resolve duplicates, and flag records that couldn't be confirmed. We defined the minimum fields that had to be complete for every entry: name, contact details, contract value, renewal date. We built a simple process so that new customers were always added consistently.
Once that foundation existed, the follow-on work was straightforward. Automated renewal reminders. A monthly MRR summary that could be trusted. A process to catch lapsed direct debits before they became write-offs. None of that automation was possible while the underlying data was broken. All of it was straightforward once the data was clean — the same principle as building a single source of truth.
The hours saved on manual reconciliation were immediate. The revenue secured through systematic renewal management compounded over time.
What this means before you touch AI
If you're thinking about AI or automation for any part of your business, the revenue calculation above is worth doing before you start. Not as a reason to delay — but as a way to understand where the real value sits.
The highest-return AI implementations I've seen are the ones where the data was good before the automation started. Not perfect — that's an unrealistic standard. But good enough that the AI is acting on information it can trust.
The lowest-return implementations are the ones where automation was layered onto bad data. The process runs. The outputs are wrong or incomplete. The business gets less value than it expected, and the owner concludes — incorrectly — that AI wasn't right for their business.
Do the data audit first. Work out what it would take to clean the fields that matter. Then build the automation. In that order, the payback is much faster.
The same revenue calculation matters if you're preparing to sell, not just to automate. Every pound of "hiding" revenue above is also a pound a buyer's due diligence team will either find and use to argue the price down, or miss because you found it first and can show your workings. Clean, defensible numbers in a data room are worth more than the underlying pounds themselves.
If you'd like help with that sequencing — working out where your data stands and what an implementation would actually look like — book a free conversation.