Keep the customers you already paid to win.
Acquisition gets the budget while customers you already earned quietly lapse. Decifer watches every customer’s behaviour over time and flags the segments slipping away, while there is still time to act, and with the value at stake measured from your own data.
The expensive customer is the one you already had.
Winning a new customer costs several times more than keeping one you have. Yet a buyer who ordered three times and then went silent for two months is usually more recoverable, and far cheaper to recover, than a stranger you pay to reach.
The catch is timing. Spotting that moment means watching every customer’s cadence and noticing the break the week it happens, which a dashboard of totals will never do for you. Decifer does exactly that part, continuously.
What we won't do
- Promise a fixed win-back rate, we measure yours against a holdout.
- Treat one-time buyers and loyal repeat buyers as the same list.
- Invent a reason a segment lapsed when the data doesn't show one.
What the module tracks.
Three views of your customer base, each computed from real orders and kept current as new ones land.
Segmentation
Recency, frequency, value
The standard RFM model: we group your customers by how recently they bought, how often, and how much they have spent, computed continuously from your real order history, not a survey or a guess.
What we measure
Which segments hold your revenue, and which are quietly shrinking.
Lifecycle state
Defined by each customer's own cadence
New, active, at-risk, dormant, lapsed, set against how that customer actually buys. A monthly buyer silent for six weeks is at-risk; a twice-a-year buyer at six weeks is not. A fixed 90-day rule misreads both.
What we measure
Who is moving the wrong way, and when the window to act is still open.
Reactivation value
What a segment is worth winning back
For a dormant segment we read its past spend and order frequency to estimate what recovering it is realistically worth, so effort goes where the return is, not where the list is largest.
What we measure
The realistic value at stake, grounded in that segment's own history.
How it works, step by step.
The work an analyst would do by hand across a spreadsheet of orders, run continuously instead of quarterly.
Build the customer timeline
We read every customer's purchase history from your store and CRM, read-only. Nothing is written back and nothing leaves your isolated workspace.
Score recency, frequency and value
Standard RFM, recalculated as orders arrive. Each customer carries a live position, not a number from a quarterly export.
Define lifecycle by cadence
At-risk and dormant thresholds are set per buying pattern, so the same 60 days of silence means different things for a weekly buyer and an annual one.
Find the movements that matter
We surface segments crossing from active to at-risk and cohorts decaying faster than their own past, weighted by the revenue attached, so the biggest leaks rank first.
Issue the decision
One ranked move: which segment, why now, what it is worth, and the customers in it, with the figures attached. Where the cause of a drop is unknown, we say so rather than invent one.
One decision, reasoned end to end.
A measured signal, the value at stake, what changed, and the move, with the figures attached. The numbers below are illustrative.
A high-value segment is slipping from active to at-risk.
Signal
320 customers who ordered three or more times in the last year have now gone 70 days without an order, past their own 45-day median gap. Their combined spend over the last twelve months was $148,000.
What changed
These are proven repeat buyers, not one-time purchasers, so the value at risk is real. The same cohort’s gap a year ago was 40 days, so this is not their normal seasonal lull. Why they slowed is not in the order data, that part needs your read.
The move
- Target this segment with a reactivation offer sized to their past spend, not a blanket discount.
- Exclude one-time buyers, whose behaviour and economics are different.
- Hold out a random slice and measure win-back against it, so the result is the program's, not chance.
If win-back matches your own historical reactivation rate, recovering even a fifth of this segment returns a meaningful share of that $148k over the next quarter, measured against the holdout, not assumed.
What the module reads.
Connects to Shopify, Klaviyo, Braze and HubSpot, or enter your order history by hand. Every figure is computed, never invented.
See which customers are slipping.
Connect your sources, and the at-risk segments, with the revenue attached, appear the same day.