Retrics

PLATFORM · THE LEDGER

Numbers you can defend.

Customers who would have returned anyway are excluded — read off the holdout where there is one, off the fixed floor where the audience is too small to carry one.
LIVELIVE IN THE RETRICS APP · HOLDOUT CARVED, SQL-BACKED · FIGURE WITHHELD
GROSS RETURNED

$8,410

ORDERS FROM CUSTOMERS RETRICS FLAGGED AND YOU MESSAGED · 30 DAYS

LESS: THE HOLDOUT'S OWN RETURN

−$1,120

THE CONTROL'S OBSERVED RETURN RATE — WHAT THESE CUSTOMERS DID WITH NO OUTREACH AT ALL

WOULD BE CLAIMED AS RECOVERED

WITHHELD TODAY

THIS SUBTRACTION IS ILLUSTRATIVE — RETRICS PUTS NO RECOVERED DOLLAR IN WRITING WHILE THE COMPARISON IS REBUILT

THE HOLDOUT

10% WITHHELD

DETERMINISTIC AND GLOBAL — A CONTROL CUSTOMER IS NEVER CONTACTED, BY ANY CAMPAIGN

METHOD

CAUSAL BY DESIGN

TREATEDMESSAGED
CONTROLNEVER MESSAGED
LIFTNOT PUBLISHED YET

THE CONTROL IS CARVED ON EVERY ELIGIBLE SEND — NO RATE OR LIFT IS PUBLISHED UNTIL THE COMPARISON COUNTS ONLY WHO A MESSAGE REACHED

ATTRIBUTION

ONE CUSTOMER, ONE CAMPAIGN

THE FIRST SEND THAT REACHED HER — NEVER COUNTED TWICE

TOO SMALL TO SPLIT

UNDER 20 CUSTOMERS

NO CONTROL IS CARVED — THE ROW READS OBSERVATIONAL, AGAINST THE FIXED ~2.7% FLOOR

WHEN IT COUNTS

EXPORTED OR SENT

APPROVING CONTACTS NOBODY, SO APPROVALS ARE INVISIBLE HERE

CONCEPT PREVIEW · NUMBERS ILLUSTRATIVE

01 — THE FLOOR

The number gets smaller. And truer.

~2.7% of lapsed customers come back with no outreach at all. That fixed floor is the bar an uncontrolled result has to clear; a controlled one clears its own holdout's observed return rate instead.

THE RECOVERY LEDGER · 30 DAYS

FIGURE WITHHELD
Returned after outreach$8,410
Less: would have returned anyway−$1,120
Would be claimed — withheld$7,290

CONTROL CARVED

Never messaged

RECIPIENT EVIDENCE

Accumulating

ONLY CUSTOMERS RETRICS FLAGGED AND YOU MESSAGED · READ AGAINST THE HOLDOUT’S OWN RETURN RATE · THE FIXED FLOOR (~2.7% OF LAPSED) APPLIES ONLY WHERE NO CONTROL EXISTS · CONTROL CARVED ON EVERY ELIGIBLE SEND · EVERY METRIC SQL-BACKED

CONCEPT PREVIEW · NUMBERS ILLUSTRATIVE · NO DOLLAR IS PUBLISHED TODAY

02 — THE LIBRARY

Every number, named and defined.

CONCEPT PREVIEW · NUMBERS ILLUSTRATIVE
METRICS CATALOG

130+ DEFINED

OPEN ANY FIGURE AND READ HOW IT IS COMPUTED

REPORT PAGES

16

ORDERS · CUSTOMERS · REVENUE · BEHAVIOR · GEOGRAPHY & MORE

RECOVERED REVENUE

PER CAMPAIGN, PER PERIOD

CAUSAL OR OBSERVATIONAL — THE ROW SAYS WHICH

COHORTS

MONTHLY TRIANGLE

SURVIVAL BY ACQUISITION MONTH

RFM GRID

5×5, REVENUE PER CELL

RECENCY BY FREQUENCY, ACROSS THE WHOLE BASE

RETENTION

REPEAT RATE, BY PERIOD

THE CURVE EVERY OTHER NUMBER ANSWERS TO

FORECAST

MODELLED, AND LABELLED

NEXT 30 AND 90 DAYS, MARKED AS PROJECTION

DISCOUNTS

WHAT MARGIN COSTS

THE QUALITY BEHIND THE REVENUE LINE

PROMISE KEPT

5 OPERATIONAL PROMISES

SHIPPING, CANCELS, STOCK, BASKET, VALUE — VS YOUR OWN HISTORY

OWNER'S BRIEF

ONE PRINT-CLEAN PAGE

THE NUMBERS A FOUNDER FORWARDS BEFORE STANDUP

CSV EXPORT

EVERY LIST, EVERY REPORT

NO SEAT FEE FOR YOUR OWN DATA

PUBLIC API

REST + webhooks

PULL THE SAME FIGURES AND CHECK OUR MATH

130+SQL-backed retention metricslive in the product today — browse the catalog in-app. Verify →
16analytics report pagesorders, customers, cohorts, RFM, behavior, geography, forecast & more.

WATCH IT PROVE

Claimed only
above the baseline.

The recovered figure is what survives subtraction: a withheld control never hears from you and sets the bar with its own return rate, the fixed floor stands in where an audience is too small to carry a control, and only customers Retrics flagged and you messaged are counted. The control is carved on every eligible send today; the figure stays withheld until the comparison counts only who a message reached.

READ AGAINST THE HOLDOUT’S OWN RETURN · CONTROL CARVED, NEVER MESSAGED · SQL-BACKED

Drift radarSCANNING 14,208 CUSTOMERS
C-4821

8 orders · due in 6 days

HELD
C-1177

3 orders · 71 days quiet

WIN-BACK RANKED
C-9204

5 orders · 44 days quiet

WIN-BACK RANKED
C-3358

12 orders · steady

HELD
WOULD BE CLAIMED · WITHHELD TODAY$0

CONCEPT PREVIEW · NUMBERS ILLUSTRATIVE · YOU WRITE THE MESSAGE

03 — EXPORT & API

Your numbers leave whenever you like.

Every report exports to CSV, and the public API pulls the same figures into your own warehouse — where you can check our math.

BEFORE YOU ASK

Straight answers.

Q01What counts as recovered revenue?

Only orders from customers Retrics flagged and you then messaged, minus what they would have done anyway — for a controlled send that is the holdout's own observed return rate, and for an audience too small to carry a control it is the fixed ~2.7% natural-return floor. If she was coming back anyway, we don't take the credit. The amount itself is withheld today: the comparison is being rebuilt to count only the customers a message actually reached, so no recovered dollar is published yet.

Q02Why are your numbers smaller than my email platform's?

Because 'platform-attributed revenue' typically credits every order a message touched, including customers who were returning regardless. Retrics nets out what would have returned anyway — the holdout's own rate where there is a control, the fixed floor where there is none — and only counts customers it flagged and you actually messaged. Smaller numbers by design — and until that count comes from recipient-level send evidence rather than audience membership, the figure stays withheld rather than estimated, because a number you cannot defend in a board meeting is worse than no number.

Q03What makes a figure causal rather than observational?

A withheld control. When an audience is large enough, a tenth of it is held back at send time and never contacted, so the return rate of the treated group can be compared against a group that got nothing; everything else is labelled observational and measured against the floor instead. That difference is the only thing that would ever make a figure causal here. The recovered-revenue figure is withheld today. The holdout is still carved and never messaged on every eligible send, so the evidence accumulates — but the comparison is being rebuilt to count only the customers a message actually reached, and Retrics won’t publish a dollar measured over anything looser.

Q04Can I audit the math?

Yes. Every metric in the library is SQL-backed and defined in the in-app catalog, every list exports to CSV, and the measurement method — holdout design, the ~2.7% floor, backtest windows — is published in full on our results page.

Run the honest math on your store.

Connect with read-only analytics scopes and see your own cohorts, your own floor, your own windows.