Understand and decide
Who buys, how often, and who is going quiet
Everything here already happened. There is no predicted lifetime value and no churn probability, because your own data cannot support a model of either — and a confident prediction is worse than an honest count.
14-day evaluation · no card · start before you connect anything
What it counts, and how it counts it
Where this sits in the loop
Understand. It learns the business. It follows Improve and hands on to Decide.
- 1
One person, assembled from your orders
Buyers are identified by address and brought together across orders, so a repeat customer reads as one person rather than as several purchases.
- 2
Quiet is defined by your business, not by a default
What counts as "gone quiet" is derived from your own repeat-purchase rhythm. A bakery and a furniture shop have completely different answers, and a fixed ninety-day rule would be wrong for both.
- 3
Below a real sample, it declines to say
Until there are enough repeat purchases to establish that rhythm, nobody is marked lapsed and the page says so with a dash rather than guessing.
What you can rely on
Counts you can stand behind
How many bought once, how many came back, how many have gone quiet, and how many of those can actually be emailed — which is usually the number that matters and the one nobody reports.
The Retention agent works from this
Where regulars have gone quiet and can be written to, an agent drafts what to say, for you to approve. It never adds anybody to a list.
Consent is part of the picture
A buyer who never subscribed has not agreed to marketing. They are counted as a customer and excluded from anything that would email them, and the gap between the two numbers is shown rather than hidden.
What it will not tell you
Figures this deliberately withholds, and what it would take to earn them.
No predicted lifetime value or churn score
Both are offered routinely elsewhere and neither is supportable from a single business’s order history. What you get instead is what happened, counted correctly.
Identity is by email address
Somebody who orders under two addresses is two people here. There is no probabilistic identity matching, because a wrong merge is worse than a missed one.
See what it already knows about your business
The evaluation starts before you connect anything: Xyranik reads your own site into facts you can check and correct. What it cannot yet measure, it says so rather than filling in.