What a Number Needs Before Two People Compare It

Take the word lead. It is used in a business to mean at least four distinct things, and each of them is a different count over the same population of enquiries. Each is defensible. They are not interchangeable, and mixing them in one meeting produces a conversation about performance that is really a conversation about vocabulary.

The spread is 40 against 14, and both figures are true, and the difference is a definition rather than a mistake. Now add the detail that makes it worse in practice: seven of the forty were booked as site visits, and only five of those seven are linked to the enquiry that produced them. The other two were logged as new records by whoever accompanied the visit. So the count of site visits is seven and is unarguable, while the count of visits traceable back to a source is five — and a conversion rate built on the first and a report of where leads come from built on the second will not reconcile, and both are being quoted in the same meeting.

The arithmetic, so it can be checked: 14 ÷ 40 = 35% on the narrowest definition, and 7 ÷ 40 = 17.5% on the site-visit definition. Same month, same forty records, a factor of two. A business arguing about whether lead quality improved is very often arguing about which of these two it meant.

Everything below is invented so the arithmetic is checkable by hand. The quotes, the customers and the outcomes are ours; none of it is a customer or a measured result. The 18% rate appears only so the totals can be verified, and the rate and treatment that actually apply are for your own chartered accountant to confirm.

Now the arithmetic that matters, because it is the whole argument: 11,16,280 − 7,19,800 = 3,96,480, and 7,19,800 − 4,27,600 = 2,92,200, and 3,96,480 + 2,92,200 = 6,88,680. Three people walk out of one meeting holding three figures that differ by 6,88,680, and every one of the three is a correct count of real records produced by the same system on the same day.

Notice also what the outstanding row proves. It reconciles two ways: 2,12,400 − 50,000 = 1,62,400 for the part-paid invoice, plus 1,29,800 for the invoice nothing has been received against, and 7,19,800 − 4,27,600 = 2,92,200 from the invoiced and collected figures. The same set of numbers also closes from the other end, because 8,85,000 of converted value less 7,19,800 invoiced is 1,65,200, which is exactly Q7 — the quote that converted but was not delivered and so is correctly absent from the invoiced figure. A set of numbers that reconciles from more than one direction is a set somebody can agree on even when they disagree about the label. A set that reconciles only one way is a set where a definition error and a data error look identical, and there is no way to tell them apart after the fact.

The thing that makes these three figures legitimate rather than sloppy is that each one is a real state of the commercial relationship, not a version of it. Quoted is what was promised. Invoiced is what was documented. Collected is what arrived. They are related by the chain — a quote becomes an invoice becomes allocations — and the differences between them are not errors at all. Q7 is 1,65,200 of promised work that has not been delivered, so its absence from the invoiced figure is a fact about the business and not a gap in the data. Confusing those two things is what turns a healthy pipeline into an argument.

The honest technical position is that the system has no way to know which of the three figures you meant, and this is not a limitation that more work would remove. Every one of the three is a legitimate projection over a stated set of records, computed the same way from the same tables. Choosing between them is a statement about what the business considers revenue, and only the business knows that. A tool that guessed would be worse than useless, because a guess looks like an answer.

What the system can do is make the choice visible rather than embedded, and that is a different and achievable thing. If quoted, billed, collected and outstanding are read and reported as four separate states, then the person quoting a figure has to pick one, and the act of picking one is where the definition gets written. If revenue were a single stored field — a number the system maintained — none of this conversation would happen, and the business would be relying on a value whose construction nobody could inspect, updating silently as records change. We regard the four-state model as the honest option even though it is the more demanding one, precisely because every number it produces can be taken apart.

The same logic applies on the other side of the chain, and it is where a definition can ask for something the platform honestly cannot supply. A metric like margin per customer, or revenue per salesperson, needs cost and effort data. NoxOrigin has no timesheet record and no payroll module, and it does no expense tracking — expense tracking is a Nox-Billings capability, and purchasing sits in Commerce. So a definition built on cost per customer may be entirely legitimate, and the platform may simply not be the place to compute it. Writing that into the definition is better than quietly dropping the metric, because a definition nobody can compute is one that quietly stops being reported.

A definition that only names the noun is not worth having, because the noun is the part that was never in dispute. 'A lead is a qualified lead' teaches nobody anything. What makes a definition usable is that it settles the questions people actually argue about, and there is a short, complete set of those questions. This is the shape we would want one of our own to take, written out once and then left alone.

The pair at the bottom is the part most definitions skip and the part that does the most work. An example that counts and an example that does not, using the same two records a reader already has in front of them, settles almost every practical argument — because the argument stops being about the definition and becomes about whether this specific record satisfies it. That is a conversation two people can finish.

None of this requires a project. It requires a page, a name, and a date, and the three have to be separate things — because the failure mode is not that a business lacks definitions, it is that a business has definitions that live in one person's head and are carried by one person's memory of what they meant last quarter.

There is a fourth thing that is not on the list and is worth more than the first three. Quote the filter and the definition together, every time, the way a careful person quotes a unit. 'Revenue of 7,19,800, invoiced, October, credit notes included' is a figure somebody can repeat next month, challenge, or compare. 'Revenue of 7,19,800' is a number that only means something to the person who produced it, and that is a polite way of saying it does not survive leaving the room.

The test of whether this is working is uncomfortable and simple. Pick a metric two people in the business quote. Read its definition aloud. If at any point somebody says 'but we count that one', you have found a real disagreement worth having — and it is a disagreement about a definition, held in daylight, with a person who can change it. If nobody objects, the definition is doing its job. The failure state is not a bad definition. It is a missing one, discovered when two people who both trust their own arithmetic discover they were talking about different things.

Three boundaries. First, a definition is a business decision and a good one can still be the wrong decision for your business. Calling a quoted invoice revenue is a choice with consequences, and the right choice depends on your market, your cycle and what you are trying to understand. What we can offer is a method for making the choice and a way of living with it; the choice itself is yours. Second, definitions do not repair the data underneath them. A perfectly written definition of a qualified lead, read from records where the requirement field is empty for half the team, will produce a perfectly reproducible number about almost nothing. Third, this is not financial reporting and we are not your accountants. Which state of the chain counts as revenue, how advances and part-payments are treated, and what your statutory position is are questions for your own chartered accountant. NoxOrigin reads the records and reports the four states separately; it does not file GST returns or any other statutory return, and it does not post to your books.

Nothing here is measured. We have no data on how often businesses agree a metric definition before or after a reporting disagreement, no count of metrics in the wild that have a named owner, and no measurement of what a written definition is worth. Every figure in the tables is arithmetic we constructed so that a 6,88,680 spread between three true readings could be traced line by line, and the quotes, customers and outcomes are invented. What we would want to instrument before making a claim is a list rather than a number: which metrics a team actually quotes, how many of those have a written definition with an owner, and how many month-on-month disagreements are resolved by a definition rather than argued past. Until those are counted, this article is a method. You can test it this month by taking the last figure you disagreed about and writing down, in one line, what it counts — and seeing whether the other person recognises their own number in it.