The Definitions Appendix, and Why the Pack Is Worthless Without It

Here is the test, and it is worth running before reading the rest of this. Pick one measure off your KPI page. Pick last month. Hand the definition and the source to two people who both use the number, send them to separate rooms, and ask each to compute it. No conversation.

Most companies at this size have never done this and would not enjoy the result. The two figures come back different, or one person says the definition does not contain enough to compute anything, or both come back with the number that is already printed on the pack because they found the spreadsheet instead of following the definition, which is the same as failing.

An undefined measure is not a measure. It is a number with a name on it, and a page of them is decoration that people argue in front of. The appendix is the thing that converts the KPI page from decoration into a management tool, and it is a record with fields, not a glossary. A glossary is a paragraph per term written once by whoever built the page. A record has an owner, a version and a date, and it can be tested.

None of the definitions in this piece are yours. That is deliberate, and it is the point rather than a limitation. Publishing a set of standard formulas would recreate the exact failure the appendix exists to prevent, which is a company measuring itself with words it did not choose and cannot defend. What follows is the shape of the record and the work to build it.

The ten fields

One row per measure, and every field filled. A field left blank is the field that will be argued about.

Field	What goes in it	Why it is separate from the others
Name	Exactly the words people say out loud, including the abbreviation they actually use	If the appendix calls it one thing and the meeting calls it another, the appendix is not the definition of anything
Purpose	One sentence on what decision this measure informs	A measure that cannot name a decision is a candidate for deletion, and this field is where that becomes visible
Formula, in words	The numerator and the denominator, each described as a thing rather than as a cell reference	A formula written as arithmetic can be copied. A formula written in words can be checked
Source	The system and the specific report, named	Somebody has to open the same report next year and get the same figure, which the phrase “the accounting system” gives them no way to do
Source accounts or fields	Account codes, or the field in the operational system, listed	This is what lets somebody reproduce the number without the person who built it
Exclusions	What is deliberately left out, and why for each one	See below. This field carries most of the value in the table
Period convention	Which month a transaction lands in, and whether the measure is monthly, rolling or year to date	Two people can agree on everything else and still differ because one used invoice date and the other used delivery date
Owner	One person, by name	Not a department. A measure owned by finance in general is owned by nobody in particular
Version and effective date	The version number, the date it took effect, and the first month it appeared on the pack in this form	Without this the trend line is a lie the moment anything changes
Superseded definition	The previous version, kept, with the dates it applied	Deleting the old one destroys the ability to read your own history

Building it, in order

This is a day of work for a company that has never done it, and about two hours a year afterwards. Do the steps in this order, because two of them exist to stop you doing more work than necessary.

1. List every measure currently on the pack, then delete the ones nobody has ever asked a question about. Do this before defining anything. Defining a measure that no one uses is the most common way a one day job becomes a three day one, and I have rarely opened a KPI page at this size without finding a measure that is on it because it was easy to produce.

2. Ask the person who quotes each surviving measure most often what they think it means, and write down their answer verbatim. Do not correct them while they are talking. Where two people give you materially different answers about the same measure, you have just found the rows that are actually costing the company something, and those get built first.

3. Write the formula in words before you write it in a spreadsheet. The words are the definition. The spreadsheet is an implementation of it, and when the two disagree later, you need to know which one was supposed to be authoritative.

4. Name the source down to the report. Then open that report and confirm it still exists and still contains the field you are relying on.

5. Reproduce one real month by hand, from the definition alone. This is the step that finds the problems. If you cannot get to the number printed on last month’s pack, either the definition is incomplete or the pack has been computing something else, and you now have to find out which. Both answers are worth the afternoon.

6. Write the exclusions, with a reason attached to each.

7. Assign one named owner per measure, and tell them. Ownership that somebody learns about while being queried has arrived too late to help.

8. Run the independence test described at the top, on the two or three measures that carry the most weight in decisions. Fix what it finds.

9. Version, date and publish. The appendix goes in the pack, at the back, every month. Not in a shared drive that people are told about once.

Step 5 is the one that gets skipped and it is the only step that proves anything. Everything before it is describing what you believe the pack does. Step 5 checks.

Exclusions are where the disputes live

Almost every reporting argument I have sat in turned out to be an argument about an exclusion, and almost none of them arrived looking like one. They start as a disagreement about the business, and half an hour in it emerges that one person has been counting something the other has been leaving out for a year.

So the exclusions field has a rule: state what is excluded and why, in the same sentence, every time.

“Excludes intercompany” is a fact and it will be reinterpreted within a year. “Excludes intercompany billings, because those are eliminated on consolidation and counting them here would show group revenue twice” is a definition that survives the person who wrote it, because the next reader can tell whether a new situation falls inside the reason or outside it.

Two exclusions worth being explicit about in almost any company, because they are the ones that get handled inconsistently and silently: whatever the company treats as a one-off, and whatever is billed to a customer as a pass-through rather than earned. Both change the measure materially, both are judgement calls, and neither is wrong. What is wrong is having the answer live in somebody’s habit.

The word that means two different things at once

If the company has a facility, the credit agreement defines terms for itself, in its own definitions section, and those definitions are contractual rather than accounting. Your internal margin measure and the agreement’s version of what looks like the same word are different quantities, and both are correct inside their own document.

The appendix has to say so. Each measure that shares a name with a defined term in an agreement carries a line stating that it is the management definition and is not the covenant definition, with a pointer to where the covenant build actually lives. The covenant work is its own exercise and it is driven by the agreement’s wording rather than by anything you choose.

Companies that leave this unstated fail in one of two directions. Either the management measure quietly drifts towards the lender’s definition, and the business ends up managed on a set of adjustments a bank negotiated, or somebody computes a covenant from the management definition because it was the one on the page, and reports a compliance figure that was never the tested figure. The second one is the expensive direction.

Changing a definition mid year

Definitions do need to change. A business changes what it sells and a measure built for the old shape stops informing anything. The rule is not that definitions are frozen. It is that changing one has a fixed price and the price is paid every time.

1. A written request, naming who wants the change and what decision the current definition is getting wrong. Not what it fails to capture in the abstract. Which decision.
2. Restated comparatives across the whole trend shown on the page, on the new basis, or the change waits until the year boundary. A trend line where the first nine points mean one thing and the last three mean another is worse than no trend line, because it looks continuous.
3. A note on the page in the month the change takes effect and in the two months after, saying what changed and pointing at the appendix entry.
4. The superseded definition kept with its date range, not overwritten.

There is one request to refuse, and refusing it is the whole reason the change control exists. A definition change proposed in the month the number looked bad, by the person the number reflects on, does not get made that month. It waits for the year boundary.

The honest complication is that person is sometimes right. The measure may genuinely have been wrong all along, and it took a bad month for anyone to look at it properly. The way to tell is to ask them to write the request as though last month had been strong. If the argument still holds in that version, it is a real argument and it should be made at the year boundary anyway, where it costs nothing to accept.

What does not belong in the appendix

Targets do not. A definition carrying a target inside it gets edited when the target is missed, and that is exactly how a measurement system loses its independence. Targets belong on the page that reports performance against them, with their own owner and their own date.

Comparisons to other companies do not, in any form. Whatever the source, the definitions behind an external figure are not yours and cannot be reconciled to yours, so a number sitting next to your measure implying a comparison is doing something the appendix exists to prevent.

Commentary does not. The appendix says what the number means, permanently. What happened this month goes in the commentary, which is a different document with a different life expectancy.

What we do not do

We do not put a measure on a pack before its definition exists, however obvious the measure seems. Obvious is the category that produces the worst disagreements, because nobody thinks to check.

We do not maintain the appendix as a separate document that is referred to. It ships in the pack, every month, at the back, alongside the rest of the standing set, because a definition nobody can see while reading the number is a definition nobody reads.

And we do not change a definition and restate a trend in the same month without saying so on the page. A silently restated trend is the single fastest way to lose a reader who was actually paying attention, which is the only reader worth having.