Small businesses sit on more information than they realise. The accounting system holds every invoice issued and received. The booking or job system holds every job, who it was for and when. The till or the shop platform holds every sale. What they do not have is answers, because the reports the software offers are a fixed set written by somebody who did not know this particular business - monthly totals, a sales chart, a list of unpaid invoices. Useful, but they answer general questions. The questions an owner actually has are specific: has that customer got slower at paying since spring, do the jobs that came from the website turn out better than the ones from the phone, which month do we always run short.
The obstacle has never been the data. It has been that answering such a question meant either finding it in a report that does not exist, or exporting everything into a spreadsheet and knowing how to pivot it. Both of those are skills, and they are not skills most people running a small business have or should need. So the questions went unasked, and decisions got made on impression instead - which is not irrational, since an owner's impressions are formed from close contact with the business and are frequently right. They are simply untested, and the ones that are wrong are wrong invisibly.
What has changed is that a person can now export the data and ask the question in ordinary words, and get both an answer and the working. The value is not that the arithmetic is difficult; it is that the barrier to asking has collapsed. When asking a question costs five minutes rather than an afternoon, people ask ten questions instead of one, and the useful finding is usually the fourth or fifth - the one nobody would have spent an afternoon on speculatively. The change is in which questions get asked at all, and that is a bigger shift than the change in how they are answered.
Two habits keep this honest, and both are cheap. First, ask for the working, not only the conclusion: which rows were counted, which were excluded, what was assumed about the ones with missing dates. A confident summary with no visible working is the one to distrust, and the assumptions are where the interesting disagreements live. Second, check any surprising answer against something you already know independently - a total you can verify, a month you remember clearly. An answer that contradicts your impression is worth having, since that is the point of measuring. But before acting on it, make sure the disagreement is between you and the business, and not between you and a column that was misread.