Service businesses lose customers quietly.
Not every dissatisfied customer complains. Some simply change their behaviour or stop returning — and every system you own records that as nothing at all. UMI exists to make that visible while it can still be acted on.
About UMIis broken
Customer-experience software got very good at collecting and stayed poor at concluding. Dashboards multiplied, sentiment became a score, and the number of things a manager is shown went up while the number of decisions those things support stayed roughly where it was.
Meanwhile the most useful information kept arriving in pieces. A review lands in one place. A booking history sits in another. A complaint is resolved in a third and closed the same evening. Each record is correct, and no single one of them tells you that a four-year regular has stopped coming.
The customers you lose do not always tell you why. Almost none of them complain — they simply stop.
building instead
Software that starts from the decision a manager has to make on a normal Tuesday morning, and keeps only what helps them make it. In practice that means three things.
Join the records up first. Before anything can be understood, the review, the booking, the complaint and the visit have to be recognised as one person. That is most of the work, and it is where UMI starts.
Say what it means in a sentence. Not a score, not a confidence percentage — a plain conclusion a manager can act on, or argue with, because the records it came from stay attached to it.
Keep the list short. A morning with three things on it is useful. A morning with forty-one is a report. If a quiet week produces a short list, that is the system working, not failing.
market
A great deal of what matters here is spoken rather than written, Arabic as often as English, or simply behavioural — the regular who stops coming, the guest who mentions something once at checkout, the patient who reschedules twice and then does not.
So Arabic is not a translation layer. A bilingual customer base has to read as one voice rather than two datasets, and the interface is built right-to-left as a first-class layout. And the emphasis sits on what people did rather than how they scored something, because behaviour exists in every business, in every language, and is harder to argue with.