The vocabulary, defined.

Forward CLVForward-looking customer lifetime value

What a customer is predicted to contribute over the next 12 months: predicted purchases times expected order value, at your product margin, less the returns the model expects, with money arriving later counted for less. It is expected contribution, not revenue, which is why a customer can top the revenue leaderboard and still score low once returns land. A discounted order is simply a smaller one, so promotions already sit inside it. Payment fees, fulfillment and ad spend are not deducted, so it is a product-margin figure rather than a fully costed one. Scored per customer, never averaged across a cohort, and distinct from historical LTV, which measures what a customer has already spent.Read more: How Forward CLV is built

Customer lifetime value (CLV)The field-standard term

The value a customer generates over the whole relationship with a brand. The term is used loosely: some tools mean revenue to date, others a revenue forecast, and not all of them take out product costs or returns. Tacet’s measure is Forward CLV, which fixes the horizon at 12 months and counts contribution rather than revenue.

Still buyingAlso written P(Active)

The probability that a customer is still a customer at all, rather than gone without telling you, often written P(Active). It weighs their own purchase rhythm against how long they have been silent: a frequent buyer gone quiet scores low, and an infrequent buyer with the same gap scores high, a distinction recency thresholds and RFM buckets cannot make. It is not a forecast that they will buy. Low single-digit percentages are normal on a base with many one-time buyers. Paired with Forward CLV, it sorts a lapsed list into who to revive, retarget, or sunset.Read more: Nobody knows when a customer churns

Return propensityPredicted, not counted

The share of a customer’s future revenue the model expects to come back, learned from their own purchase and return behavior. A prediction, not a count of past returns: the observed refund share is the measured counterpart, and the same customer can rank differently on each. It is how returns enter Forward CLV. Blank for one-time buyers, who have no repeat history to learn from. An exchange that leaves no refund isn’t visible to it today, so a customer who only ever exchanges reads lower than they should.Read more: Returns are a forward-value problem

Value concentrationWhere value sits in the base

How Forward CLV is distributed across a customer base, read as the share held by the top 1%, 10% and 20% of customers ranked by their own score. Value is rarely spread evenly: a thin top tier tends to hold much of the forward value. Concentration tells an operator where retention and acquisition dollars should go, and it replaces flat demographic segmentation. Earlier Tacet material calls it CLV concentration.

One-time buyerExactly one order, ever

A customer with exactly one order, ever; the portal calls them one-timers. In the DTC books Tacet has scored, most customers buy exactly once, so how they are read shapes the whole base. With no repeat history to learn from, parts of the model treat them differently: their return propensity is left blank, for one. Whether a one-time buyer comes back for a second order is where much of a brand’s forward value is won or lost.Read more: Most of your customers will never buy again

PlayA proposed action you approve

A specific action Tacet proposes on a pool of customers the scores identify: the audience, the message, and the send time. An admin in your organization approves each one; Tacet staff can’t, and a proposal nobody approves expires after 7 days. Today a play sends an email campaign from your own Klaviyo account, inside Tacet’s standard sending rules, which apply by default and which your admin can adjust. Every play is read against a held-back share of the same customers, and the result is reported as incremental profit.Read more: When Tacet acts for you

Held backThe comparison that makes a result real

The share of a play’s audience kept out of the send on purpose, so the rest can be read against it. A held-back customer is a customer not emailed: that is what makes the read possible, not a gap in the count. Tacet also keeps a persistent holdout group that none of its sends reach. Measured this way, a result is the difference between two groups of the same customers, not credit an attribution model hands out.Read more: An email that worked, according to every dashboard

Incremental profitWhat a play actually changed

Profit measured against customers Tacet deliberately did not contact: what the group that was reached did, compared with the held-back group that was not. It is shown as a range, a confidence interval, because it compares two groups rather than counting one. It is not the revenue Klaviyo credits to a send, and not a forecast the way Forward CLV is: it is what was measured, and a result of zero is reported as zero. Older Tacet material calls it DeltaCLV.Read more: The metric you optimize decides which strategies survive

Confidence intervalHow sure a measured result is

The range a measured result plausibly sits in, given how many customers were measured and how much they vary. Tacet puts one on measured outcomes, such as a play’s incremental profit against its held-back group, and not on per-customer scores, which are predictions rather than measurements. A narrow interval is a result you can act on. One that spans zero means the play has not shown an effect either way.Read more: What a confidence interval is doing on a marketing report

The models underneathProbabilistic, not heuristic

Forward CLV is produced by probabilistic models: one estimating how often a customer buys and when they lapse, another estimating what they spend when they do. Both are fitted per customer, not per cohort, which is what makes the output a score rather than a bucket. What separates Tacet is the layer on top: product costs and expected returns taken into the score itself, so the number is contribution rather than revenue.

Holdout validationTesting a model on data it hasn’t seen

The discipline of testing a forecast on data it has not seen: fit the model on an earlier window of order history, predict the most recent months held out from it, then compare the prediction with what actually happened. It checks the model rather than a play, which is what a held-back group is for.

Predictive vs. descriptive CLVForward-looking vs. backward-looking

Descriptive CLV, often called historical LTV, reports what customers have already done. Predictive CLV, or Forward CLV, estimates what they will do next. Most dashboards are descriptive: they chart the past. Forward CLV is predictive, and only a predictive number can be acted on before the money is spent. The two won’t tie: a cohort LTV is measured backwards over customers who have already completed a year, while Forward CLV is averaged across everyone scored, dormant customers included.

After product costs and returnsWhy revenue overstates value

Revenue is booked at checkout. Returns land weeks later, and product costs never appear in revenue at all. Forward CLV takes out product costs and the returns each customer is expected to make, so the number reflects contribution rather than gross sales. In categories with high return rates, the returns are most of the story.

Contribution marginWhat a sale earns after its own costs

What a sale earns after the costs that come with it: product cost, returns and discounts, and in fuller versions fulfillment, payment processing, and advertising. Forward CLV works at product margin: it takes out product costs and expected returns, but not payment fees, fulfillment, or ad spend, so it sits above a fully costed contribution margin. Read the two together.Read more: Revenue is not a proxy for the health of your business

The DiagnosticTacet’s free first step

Tacet’s free first step: every customer in your store scored by Forward CLV, from one read-only Shopify connection, with where your value concentrates and where contribution leaks. You keep it whether or not you subscribe. Where a store has too little repeat history for the scores to be reliable, the Diagnostic says so.Read more: A sample Diagnostic

TrackerThe portal’s daily sheet

The page the Tacet portal opens on: your store’s own figures, one column per day, a month at a time, with each row carrying its definition and its source. It downloads as a CSV. Where Forward CLV looks ahead, the Tracker is the day-by-day record of what the store did.Read more: Who can see the portal

How the numbers are built is documented on the methodology page, and plain answers about the product are on the FAQ.