Every number we publish is on a leash.
Our whole pitch is that your numbers are lying to you. That claim would be worthless if you had to take our numbers on faith. So here is the standard every published figure follows, written down where you can hold us to it.
What each number is allowed to claim.
Realized means it happened and is on the engagement record. Modeled means it is a forward projection, and it will always say so in the same breath. We never present a projection as a result, and when a case file carries both, the labels sit next to the numbers, not in a footnote.
Agency-measured means the figure comes from systems we instrumented and read ourselves. Client-attested means the client's own records produced it and the client confirmed it. Neither means third-party audited, and our files say that too: the standing line is "agency-measured and not independently audited," and it stays attached to the claim it qualifies.
Booked is not collected. When a figure counts signed work, initial retainers, or scheduled treatment, the file says which, because those are different dollars. A number that would only survive as the bigger, vaguer version of itself does not get published.
Why some figures are ranges and multiples.
Our proof is de-identified. Owner-run businesses show us their real economics, their broken tracking, and their worst quarters, and they can only afford to do that because we do not put their name on it. The protection has a cost, and we spend it on the number, never on the truth: sensitive figures print as ranges ("about," "under," "a strong multiple of the category benchmark") rather than exact values that could identify the client.
The counting unit is stated in public: one case file is one diagnosis, not one client. A long engagement legitimately produces several files. One client is never claimed twice as two clients.
What never happens.
No invented numbers, ever, including for color. No stitching two clients into one story without saying it is a composite. No attaching one client's result to another client's narrative. No quoting a client verbatim unless those were the words; paraphrase is attributed as paraphrase. No health, legal, or financial claim goes out without a compliance review, and claims we cannot substantiate get replaced with the defensible statement underneath them, on every surface we publish, in every channel we use.
Where published research is part of an argument, we name the researcher and the year, and when a finding is contested we say it is contested.
What you can ask us to show you.
You should not take a seller's numbers on trust, including ours. On a call, ask how any specific published claim was measured and we will walk you through the method behind it: what was instrumented, over what period, and what was excluded. Under NDA, we can go further, including the working behind a sample claim. And the fastest check needs no permission at all: every case file keeps its limitations in the same sentence as its result. If you ever find one that does not, tell us, and we will fix the file and thank you for it.