Your agents already know the job.

Reporting a number from a live ledger as if the month were closed, so every downstream figure moves after it has been quoted to the board.

Silent division — ad spend over store revenue, presented as ROAS, with two different attribution windows underneath it.

A variance report listing every line that moved, which buries the two that matter and gets skimmed.

A discount given because it was asked for, which reprices the account permanently and teaches every future negotiation that the list price is a starting bid.

Use it when the preview is blank, or verify_build fails unexplained.

Producing a scored ranking whose numbers were invented. A prioritisation framework turns judgements into arithmetic, and arithmetic reads as evidence — so an unsourced score is more persuasive and less honest than the opinion it replaced.

A keyword list ranked by search volume, which reliably produces traffic that never buys — and a programme judged on sessions rather than on the customers it produced.

A positioning statement written from what the company wishes were true, which then contradicts what the product does and what customers say — and every piece of copy downstream inherits the contradiction.

Treating publication as the end of the work. A piece published and posted once reaches a fraction of the audience it was made for, and the team concludes the content was bad rather than that it was never distributed.

Answering an allocation question with "we should do both", which is the answer that leaves the trade-off to whoever is loudest next week.

A generalist agent absorbing specialist work, which makes it invisible on the specialist's board and produces a worse answer than the remit that was written for it.

Calling a winner on a test that never had the volume to detect the effect it was looking for, then rolling out a change that does nothing. The stopping rule decided after seeing the data is the specific failure.

Treating an approved draft as an approved LIST. The recipients need approving too, and a send split into smaller calls to stay under a threshold is the same violation twice.

Reporting a burn number with no breakdown, so a one-off annual renewal reads as a permanent increase in run rate.

Pitching an upgrade to everybody, which converts nobody and costs the trust of the accounts that were doing fine.

Analysing the loss reason recorded in the CRM. That field holds the rep's account of a decision they were not present for, and it disagrees with what the buyer says often enough that a strategy built on it is optimising against a fiction.

A list of best-practice tweaks — bigger button, less copy, social proof — applied to a page whose problem was that the traffic was wrong. Changing the page cannot fix the audience, and the test that follows measures nothing.

A "just checking in" sequence sent to a list nobody has segmented, which trains the recipient to ignore the sender and burns the address for the moment there is something real to say.

A 40-slide deck nobody opens on a call. Enablement fails on retrievability under pressure, not on completeness, and a document written to be comprehensive is written for the wrong moment.

Judging a partnership on the size of the partner rather than on who owns the customer and what the first ninety days actually require from us.

Reporting a single root cause, which is almost always the last human action in the chain. Naming a person as the cause ends the investigation exactly where the systemic contributors start, and it teaches everyone else to report less next time.

Use it when a multi-user business tool, or private per-user data.

Improving a meeting by tightening its agenda. Most bad recurring meetings have no decision to make, and a better-run status meeting is a more efficient way to do something that should not be happening at all.

A single runway number with no assumptions attached, computed from last month's burn rather than a three-month average.

Opening a role described by its title instead of by the work, so the interview loop measures nothing and the debate about the hire restarts at the offer.

A calendar of topics with no owner and no purpose per item, which produces posts nobody can evaluate and a schedule that quietly stops.

Presenting weighted pipeline as a forecast. Stage-probability weighting is arithmetic on optimism and reliably produces a number that has never once been hit.

A cash forecast built on the P&L, which is on an accrual basis — so it shows a profitable month in which the bank account is empty.

A brief that presents analysis and no options, so the reader has to do the structuring work the brief was supposed to do.

Reporting "looks good" from reading the diff. A false pass costs more than an honest "I could not run it".