What the owner could not see before
The pain that produces this request is not usually reporting. It is that the business runs across several systems that do not talk to each other, so the owner cannot answer a simple question without asking three people and waiting.
How many jobs are actually live. What is sitting unquoted. Which client has gone quiet. Whether this month is better or worse than the last one, on the measure that matters rather than the one that is easy to pull.
Every one of those answers already exists somewhere in the software you run. The problem is that it exists in four places, in four shapes, and reconciling them is somebody's Friday.
What is on the screen
Less than people expect, deliberately. A dashboard that shows everything gets read once. The ones that stay open show a small number of things that change and that somebody can act on.
- What is in flight, and which items have stopped moving. Not a count, but the specific ones, so it is a to-do rather than a statistic.
- What is waiting on us and what is waiting on someone else, kept apart, because they are different problems and mixing them is why the number never goes down.
- The one or two measures the business is actually run on, against the same period last year rather than against a target somebody set once.
- Anything that looks wrong: an invoice that does not match its order, a job past a date, a quote that never got sent. This is the part that earns its keep, because it is attention nobody has to spend.
Image not yet supplied
The dashboard itself, at the size a person actually uses it, showing the four things listed above and not much else. The restraint is the point, so do not choose the busiest screen available. Must be anonymised: no client or customer name, no dollar figure, no email address.
What stays manual, on purpose
A screen that shows you something is a different thing from a screen that acts on it, and the second is where these builds get into trouble. So some steps stay with a person, and the rule is that anything with a consequence somebody else experiences needs a human between the decision and the doing.
- A quantity a customer typed does not become an order. A member of staff signs off in between, every time.
- Generated material is staged, never published. One run put the wrong client’s artwork in the wrong folder, so candidates now wait for a person, with no exception for a confident-looking result.
- Outbound goes one at a time, capped per day, with somebody approving each one before it leaves.
None of that is a limit of the software. Each one is a place where the machine was perfectly capable and was not allowed, and the reason is the same in all three: the cost of the rare wrong one is much higher than the saving on the ordinary right ones.
Where the numbers come from
Getting a trustworthy number is the part never shown in a demo, and it is where essentially all of the cost sits.
The screen is a few days' work. Getting a reliable number onto it means reading from systems that were not built to be read from, deciding which one is authoritative when two disagree, and handling the cases where the same customer exists twice with slightly different names. That is not AI work. It is the unglamorous reconciliation that decides whether anybody trusts the screen.
Image not yet supplied
The part of the tooling where two records from different systems are matched to each other and a person confirms or rejects the match: the reconciliation screen, not the dashboard. This is the image nobody else publishes, which is exactly why it belongs here. Must be anonymised: substitute the company names shown, and remove any figure.
The question to ask anyone showing you a dashboard
Not can it show me this. Ask where does that number come from, and what happens to it when the two systems disagree. The answer tells you whether you are looking at a product or a picture of one.
What it costs to keep true
A dashboard is not a delivery, it is a dependency. Every system it reads from will change. A field gets renamed, an export changes shape, somebody upgrades something. When that happens the number either breaks visibly or, much worse, silently goes wrong while still looking plausible.
So the real cost includes keeping it true: checks that notice when a source stops matching, and somebody who owns it. A dashboard nobody maintains becomes a dashboard nobody trusts, usually within a quarter, and a screen nobody trusts is worse than no screen because the business has stopped looking anywhere else.
The way in is one process, not the whole picture
Everything above is what people describe when they tell us what is wrong, because fragmentation is what the problem feels like from the inside. It is also the most expensive of the three ways we work. Those two facts pull in opposite directions, and which one we act on is the thing to watch.
We do not think you should start here. Everything above is level three, the largest of the three levels we work at. Not because the connected version is wrong, but because it is the most expensive way to find out whether you want to work with anybody.
The entry is level one: one process where the front of your business meets a customer, quoting or intake or the follow-up that keeps getting missed, connected properly. It has the same joins in it as the whole picture, at a size you can check, and it either runs or it does not.
Two things follow from doing it that way round. You will have watched us work on your systems before committing to anything larger, which is worth more than any proposal. And the decisions made along the way get written down, which is most of what makes the connected version cheaper and easier to trust when it comes.
One more thing, and it is about us, not about you: sales and marketing is where our own tooling is deepest and where a first result usually arrives fastest. Back-end systems work is possible and it is a larger thing. We would rather say that here than shape an engagement to fit it later.