Home · Services · Data & insight

Service · Data & insight

The numbers already exist. The answers don't.

Sales records, spreadsheets and systems that never speak to each other — your business is already producing the evidence. We gather it into one set of numbers everyone trusts, build the view your team will actually open, and put a signal on the readings that matter.

Software we buildA live dashboard and running signals — they watch and they tell a person; they never act

No obligation, thirty minutes, nothing pitched. Bring the decision you wish were easier. You'll leave knowing whether your figures can answer it today — and what it would take if they can't.

What it actually is

One clean model underneath, one view on top, and a signal watching.

Three pieces, in that order, and a question in front of all of them. Skip the question and the dashboard argues with the spreadsheet; skip the signal and you meet the problem in next month's report.

The questions
We start from the decisions you wish were easier, not from whatever data happens to be lying around. A workshop decides what actually deserves measuring, and every number gets defined once, in writing, so two reports can never disagree about what "active customer" means.
The model
Your scattered sources connected and reconciled into one reliable foundation — with a map of where every figure comes from and how it flows, so a number you doubt can always be traced back to the system it came from.
The view
A live dashboard built around your questions rather than a vendor's defaults, and made for a Tuesday morning rather than for a screenshot. If nobody opens it, it has failed however good it looks.
The signal
AI watching the metrics that matter for readings that break their own history — and raising a hand early, to a named person, with the reading and its history attached.
What it never does
No signal changes a price, cancels an order, reorders stock or sends anything to a customer. It tells someone. The deciding stays exactly where it was, with the people who are accountable for it.
What you get

Numbers you can start a meeting with.

Four outcomes this service is built to produce, and then the artefacts they leave behind — the model, the definitions and the signal rules, sitting in your own accounts on standard tools.

Truth
One set of numbers

Your scattered figures cleaned and gathered, each defined once — so a meeting starts from the same facts instead of three spreadsheets and a memory.

Use
A view people open

Built around the questions you actually ask, in the order you ask them. A dashboard nobody opens is a report with better lighting.

Early
Trouble while it's small

Signals watch for readings that break their own pattern and flag them to a person — before the damage has finished landing.

Speed
Answers without the wait

Ask something about the business and get the answer while it can still change what you'd do — not in next month's pack.

The artefacts, and where each one lives

Yours to keep

What you keep

  • A KPI workshopdeciding together what actually deserves measuring, before a line of anything is built
  • A metric dictionaryevery number defined once, in writing — no more duelling spreadsheets
  • A data-source mapwhere every figure comes from, how it flows, and which system is the one that decides it
  • The live dashboard itselfbuilt for daily use, not for show
  • AI early-warning signalswatching the metrics that matter and raising a hand early, to a named person
  • An interpretation guide and trainingso your team reads the numbers without needing us in the room
  • Handover documentation in plain languagewhich source each figure is read from, where the answer lands, and how to change a definition without calling anyone
Order of work

How it is built

  • We start from your real questionsnot from whatever data happens to be lying around
  • The first working dashboard appears earlyearly enough to be argued with, then it grows on your feedback
  • What may be read, and what may be stored, settled in writingbefore the first source is connected
  • Everything documented and handed overthe model, the rules and the accounts — your team owns the numbers
Where a human stays in charge

A signal raises a hand. A person decides what it means.

An alert that acts on its own is how a bad reading becomes a bad decision at machine speed. Everything we build here watches and tells — and the line it may not cross is written down, in your account, in these words.

The signal rule, as it is actually written plain language · yours to edit
  1. IFa metric crosses a line you set yourself
  2. ORa reading breaks the pattern its own history has kept
  3. ORthe source behind it is late, incomplete, or newly disagreeing with another source
  4. THENraise it to a named person, with the reading, its history and its source attached — and change nothing

The lines live in your account, in these words, from the day the dashboard goes live. Move one and the signal moves with it — no ticket, no approval, no waiting for us. And a signal that keeps crying wolf is one you switch off yourself: we will not argue for keeping a rule your team has stopped believing.

How it runs

Five phases, and no dashboard before phase two.

The order never changes. Here it matters most at the seam between two and three: nothing gets built on a number that has not been defined yet — and every phase ends at a door: whatever is finished when you stop is already sitting in your accounts.

PHASE ONE

Listen

A 30-minute call about the decisions you wish were easier, then time with the people who assemble today's reports by hand.

PHASE TWO

Map

The KPI workshop: what deserves measuring, what each number means, and where every figure comes from. One truth per number, written down.

PHASE THREE

Build

Your sources connected and reconciled into one model, and the first working dashboard put in front of you early enough for you to disagree with it.

PHASE FOUR

Prove

It runs beside the spreadsheet it replaces, on real weeks, until the two agree — and your team trusts the one that took nobody's morning.

PHASE FIVE

Hand over

Model, dashboard, signal rules, documentation and training. Your numbers, in your accounts, in your hands.

No lock-in by design — the model, the dashboard and the rules are yours, on standard tools any developer could maintain.

What we won't claim

What a dashboard cannot do for you.

A screen can look finished and still be useless, and data work is where that is easiest to hide. Both lists below stay on the page as policy: a dashboard is an easy thing to pitch and a hard thing to undo, so the reasons against are printed beside the reasons for.

The honest shape

What a dashboard is not

  • Not a fortune teller. A signal compares today against the history it has been given. It does not know your market, your competitor's plans, or what the season will do.
  • Not a replacement for your accounting or point-of-sale software. We read from what you already run, and those systems stay the record.
  • Not a data warehouse programme. We build the smallest clean model that answers your questions, and grow it only when a question needs it to grow.
  • Not a screen for the boardroom wall. If nobody opens it on an ordinary Tuesday, it has failed — and we would rather find that out in the pilot than after handover.
Reasons to hold off

When we'd tell you not to build it

  • The data isn't there yet. If the figures you need have never been recorded, no dashboard can invent them. The first job is recording them — a smaller piece of work than the dashboard would have been.
  • One person reads one spreadsheet and it works. If the current way is not losing you decisions, keep it and put the effort somewhere it hurts.
  • The question is "what should I do", not "what is happening". That is a judgement call. A dashboard can inform it; nothing here will make it for you.
  • Nobody can be spared to define the numbers. Without the workshop you get a fast, confident, disputed dashboard — worse than the spreadsheets it replaced.
Questions

Four answers, before you give up an hour.

The questions this particular service raises, written out here rather than saved for the meeting — because an answer you can re-read is worth more than one you heard once.

Q01 Our data is a mess. Do we have to clean it up first?

No — sorting it out is the work, and telling you to tidy up before we arrive would just delay the project forever. What happens first is that we find out how bad it is and say so plainly: which of your questions the data can answer today, which need a fix first, and which cannot be answered at all yet.

You get that answer before you commit to a dashboard, not after it has been built on top of the mess.

Q02 Who fixes the dashboard when a source system changes?

For an ordinary change — a renamed column, a new field, a definition that has moved on — your own team can, and the handover documentation is written for exactly that: what the dashboard reads, where each figure comes from, and how to point a definition at something new. The data-source map is the page you open first.

When a source system changes shape entirely — a new accounting package, a replaced point-of-sale — that is a piece of work rather than an edit. You can hand it to us, to your own developer, or to whoever installed the new system: it runs on standard tools, in your accounts, so nobody has to wait for us to be available.

Q03 Can the AI tell us what's going to happen?

It can tell you when something has stopped behaving the way it always has, which is the useful half of that question. Signals compare a reading against its own history and flag the break while it is still small.

What it will not do is dress a guess about your market up as a forecast. Confident numbers about a future nobody has seen are exactly the kind of number this firm does not publish — not on a dashboard, and not on this website.

Q04 Who can see the dashboard, and where do our numbers sit?

You decide, and it is written down before anything is connected: which sources may be read, what may be stored, and who may open which view. Access is yours to grant and yours to revoke — the accounts are created in your systems, in your name, with the narrowest permission the job allows.

Two commitments cover that on every service, and both are contract terms rather than page copy: an NDA comes back signed before you send us a document or a login, and nothing of yours is ever used to train AI models. They are set out in full among the four standing orders on the services page.

Thirty minutes · zero pitch

Bring the decision you keep making blind.

Thirty minutes, with us, on the numbers you already collect. You get a plain answer on whether they can settle your question yet — and on what it would take if they can't.

Ask us to call you back

with FuseLoom · [email protected] · and if the figures you already keep answer the question, we will say so