Glossary

The words, defined plainly.

Direct mail carries a lot of vocabulary, some of it postal, some of it statistical, and some of it invented by vendors to make attribution sound like proof. These are the terms that come up when mail is run as a measured channel, defined the way we use them.

Measurement

HoldoutA group of accounts deliberately excluded from a campaign so their outcomes can be compared against the accounts you did mail. The only design that answers what the channel caused rather than what merely followed it.
IncrementalityThe share of an outcome that would not have happened without the channel. It is what a holdout measures, and it is almost always smaller than what attribution reports.
Lift, absolute and relativeAbsolute lift is the difference in percentage points: 8.9% against 7.0% is 1.9 points. Relative lift expresses the same gap as a proportion: 27%. Absolute is what converts to pipeline; relative is what appears in marketing.
MatchbackJoining closed opportunities back to the accounts that were mailed. Useful for sizing a program, but silent on causation, because it has no comparison group.
Attribution windowThe period after delivery during which an outcome is counted as related to the piece. Longer windows credit the channel with more, which is why the window has to be fixed before the campaign runs.
Intention to treatAnalyzing every account as randomized, including those whose mail never arrived. Dropping undelivered pieces looks like data hygiene and reliably inflates the result, because bad addresses cluster in dead accounts.
Minimum detectable effect (MDE)The smallest lift a given design could reliably find. Computed from base rate, sample size and power. A program whose MDE is larger than any plausible effect cannot learn anything, however it is reported.
Statistical powerThe probability a test detects a real effect of a given size. Conventionally 80%. Low power produces null results that get misread as evidence the channel failed.
Confidence intervalThe plausible range around an estimate. A 27% lift spanning +9% to +48% is a finding; the same 27% spanning −5% to +61% is noise. Reporting the midpoint alone hides which one you have.
Direct responseA response traceable to the piece itself — a QR scan or a landing page unique to that mailing. Evidence rather than inference, and it needs far less volume to measure, because the baseline is zero by construction.

Data quality

Address resolutionFinding a deliverable postal address for a contact you may know only as a work email. Distinct from verification, which checks an address you already have.
CASSCoding Accuracy Support System. A USPS certification for software that standardizes and validates US addresses. Canada Post's equivalent is SERP.
NCOANational Change of Address. A USPS dataset of forwarding orders, used to update addresses for people and businesses that have moved.
MailabilityWhether a given address is fit to receive a piece before you buy postage: format valid, deliverable, not suppressed, and confident enough to be worth the spend.
Suppression listAccounts and individuals who must never be mailed — opt-outs, competitors, active support escalations, anyone a rep has flagged. Applied globally rather than per campaign.
Return to senderA piece the postal service could not deliver and sent back. Costly, but the most useful feedback a mail program gets: every return is a labeled example that improves the next address decision.

Production

IMbIntelligent Mail Barcode. The USPS barcode that makes a piece trackable through the mail stream, which is what lets follow-up be timed to delivery rather than to the send button.
First class vs marketing mailFirst class is faster and more expensive, with predictable transit. Marketing mail (formerly standard class) is cheaper and slower, with a delivery window measured in weeks. Cheaper postage costs you the ability to time a follow-up call.
Self-mailerA piece that travels without an envelope, folded and sealed. Cheaper than an enveloped letter and more visible than a postcard.
Dimensional mailAnything with depth — a box, a kit, an object. Sometimes called lumpy mail. Higher open rates and an entirely different supply chain from flat print.
Robotic handwritingA machine holding a real ballpoint pen, writing in a handwriting style. Not a printed font that imitates handwriting, and not a person. A separate industry from print, which is why it is always a separate vendor.

Programs

ABMAccount-based marketing. Treating a named list of accounts as the unit of targeting and measurement rather than individual leads. Mail fits it well, and it is also why holdouts have to be assigned by account.
Buying committeeThe group of people at one account who influence a purchase, typically six to ten in B2B software. The reason contact-level holdouts leak: mail one member and you have partly treated the whole account.
SignalA public or CRM event that makes an account worth mailing now — a funding round, a job posting, a renewal window, a stalled deal. What separates a campaign from a batch send.
Cost per opportunityProgram cost divided by incremental opportunities created. The unit that decides whether a channel gets funded again, and the reason cost per piece is the wrong comparison.
ProvenanceThe record of where a claim in a piece came from and when. On paper it matters more than in email, because a letter with a wrong fact about a company's funding round cannot be recalled.

The holdout testing guide covers the measurement terms in depth, and the measurement page shows how they appear in a real report.

Write to us

Say it on paper. Prove it in pipeline.

Launch a pilot to your top accounts: research-backed pieces your reps approve, follow-up timed to delivery, and an account-level holdout that reports what the channel actually created.

Start freeSchedule a call

Setup takes an afternoon. Delivery takes days, because paper travels. The report takes a sales cycle, because pipeline does too.

Write to us: adam@trysincerely.com

Sincerely,