Your company has already recorded everything it's done. Every proposal, every order, every complaint, every hour logged. Years of it, stored somewhere.
The question isn't whether the data exists — it almost always does. It's whether it's stored so it can be read, or only so it can be filed. The difference shows itself in ten minutes, with five questions.
Five questions to ask today
Try answering each one without asking anyone to export anything. If you can, the system is working for you.
- On average, how long passes between a request coming in and a proposal going out?
- Of the proposals you sent this year, what percentage were accepted? And the ones above ten thousand euros?
- Which clients bought less this year than last year?
- How many times a month does that exception — the one everyone complains about — actually happen?
- Which day of the week brings in the most work?
These are mundane questions. None of them needs artificial intelligence or expensive tools. And in most companies, none of them has a quick answer.
Why you can't answer
It's rarely down to a lack of information. It's because the information is stored around whoever wrote it, not whoever is going to read it.
The date the request came in is in the email. The date of the proposal is in the invoicing system. Whether it was accepted or not is in the salesperson's head, or in a column that sometimes gets filled in. To answer the first question you have to cross-reference three places by hand — which is why nobody does it more than once a year, when the accountant insists.
The same happens with the exception everyone complains about. It isn't recorded anywhere as an exception: it's scattered across free-text notes, each written in someone's own words. Everyone knows it happens "a lot". Nobody knows if it's four or forty times a month, and that difference completely changes what's worth doing about it.
What those answers are worth
A company that can answer the five questions makes different decisions.
If it knows proposals above ten thousand euros have half the acceptance rate of the others, it changes how it presents them. If it knows work comes in mostly on Mondays and Tuesdays, it rosters the team differently. If it knows that exception is forty a month, not four, it stops treating it as an exception and starts treating it as a process.
None of this is advanced analytics. It's arithmetic on data that's already there.
Start with the question, never with the data
The temptation is to tidy up all the information first and only then see what can be asked. That's a project with no end, because there's no way to tell when it's done.
The other way round: pick one of the five questions — the one you'd find most useful — and tidy up only what's needed to answer it. Usually that's two or three fields and a link between systems. A week of work, not a quarter.
Once that one's answered, the next one costs less, because half the tidying is already done. That's how you build the records that make artificial intelligence useful later on: without ever running a project called "organise the data".
Which question do you need answered?
Tell us what you'd like to know about your operation and where the data lives today. We'll tell you what's missing to get there.