What a data audit really finds in a small business
Six findings that come back almost every time, whatever the sector. None of them is a tooling problem, and that is exactly what makes them expensive.
· 6 minute read
A data audit in a small business takes one to two weeks. You open the systems, you look at what comes in and what goes out, and you talk to the people who enter the data rather than the ones who comment on it. At the end, you hand over a document.
That document rarely surprises anyone technically. It surprises by what it reveals about a company that is otherwise running perfectly well. Here are the six findings that come back most often, and what they actually cost.
1. There are always more sources than management says
Ask how many tools are in use and the answer almost always names three or four: the ERP, the CRM, the online shop, payroll. Look properly and you find two to three times as many.
The gap is not made of forgotten software. It is made of spreadsheets. A tracking file created one day by one person for a one-off need, which became the only source of truth on a subject, without anyone deciding it and without anyone backing it up. Those files often carry the most critical information in the company: negotiated discounts, product compatibilities, commitments made.
The cost is not the mess. It is that part of your company depends on a file you cannot name, on a machine you cannot point to.
2. The same figure has several definitions, and none of them is wrong
In one distribution business we found four definitions of revenue living side by side: with or without year-end rebates, counted at order or at invoice, shipping included or not, returns deducted or not. Each was defensible. Each answered a legitimate question, asked by a different department.
The problem does not show up in the calculation. It shows up when two people walk into a meeting with two figures, and the discussion becomes about the figure instead of the decision. Half a meeting lost every month arbitrating a gap nobody can explain.
It is also the least technical finding on this list, and the most profitable to fix: half a day with management settles it, provided somebody has the authority to settle it.
3. Master data exists twice, and nobody owns it
The same product carries two codes depending on whether you look in the ERP or in the online shop. The same customer exists twice across two branches. The same temp has three separate histories because they worked through three offices.
As long as everyone works inside their own tool, nobody sees it. The day you try to consolidate, everything goes wrong at once - and not obviously wrong: wrong in a way that does not show, because the totals stay plausible.
Reconciling what exists is done once. What matters more is that somebody is named to keep the master data afterwards. Without that, the duplicates are back within six months and you will have paid for a clean-up, not for a solution.
4. Nothing is historised, so nothing is comparable
Your business software is perfectly good at telling you today's state. It almost never tells you the state six weeks ago: it overwrites, because it was built to run operations, not to tell a story.
The direct consequence: you cannot anticipate a stockout, because you do not have the curve leading to it. You cannot measure whether your fill rate is improving, because you only have today's. You cannot tell whether an action worked, because you do not have the before.
It is the finding with the most consequences and the slowest to repair: history starts the day you build it. Six months earlier would have been better, and that stays true every single day.
5. The knowledge that matters is written down nowhere
Compatibility between spare parts, at a manufacturer. Key account preferences, at a distributor. The rules for arbitrating a schedule, at an agency. That knowledge exists and is in fact very reliable - it sits in the heads of two people, sometimes in a workbook only they maintain.
That is the real risk in a small business, and it has nothing to do with IT. One retirement, one long sick leave, and part of the trade becomes unreachable.
No artificial intelligence fixes this. It has to be written down, structured and validated with the people who know. It is work, it is priced separately, and it is often the most profitable investment of the whole engagement.
6. What blocks is almost never technical
On the technical side the answer is usually good: the APIs exist, the databases are reachable, and the volumes of a small business raise no architectural problem at all. When an audit concludes you should not go ahead, it is rarely for infrastructure reasons.
Two blockers recur. The first is contractual: some vendors charge for access to your own data, or refuse it outright. That is the first thing to check, before any talk of architecture, because it changes the budget entirely.
The second is human: you need somebody with the authority to settle the definitions, and somebody willing to keep the master data over time. If those two people do not exist, the best foundation in the world degrades within a year.
What it means if you recognise yourself
None of these six findings is the sign of a badly run company. You find them in profitable, well-managed businesses that grew faster than their tools. That is exactly what makes them hard to see from the inside: nothing breaks, everything costs a little.
It is also why we always start with a short, paid audit rather than with a project proposal. Two weeks to know what your data actually allows, what it would cost, and in what order to do it. If the answer is that nothing should be done for now, the document will say so, and it will be yours regardless.
Want to know what yours would find?
Thirty minutes is enough to say whether an audit makes sense at your company. If it does not, we will tell you.