How LTV.ai shapes exactly who receives a campaign
Blasting the list and crude engagement suppression are both wrong. LTV.ai shapes audiences over calibrated segments, with value floors and built-in holdouts.
Nairit BanerjeeHead of Machine Learning
There are two common ways brands decide who gets a campaign, and both are bad. The first is to blast the whole list, on the theory that more sends means more revenue. The second, in reaction to that, is aggressive engagement-based suppression that trims the list down hard.
The first quietly damages deliverability and margin; the second routinely drops valuable customers who simply have not opened lately. And almost no one runs a control group, so no one can actually prove any of it worked.
How it is done today, and why it is weak
Blasting treats reputation as free, which it is not; mailing people who will never engage teaches inbox providers to distrust the sender, which hurts the whole program. Crude suppression over-corrects, using recency as a proxy for value and silently removing high-value customers who buy in bursts.
And without a holdout, "this campaign drove revenue" is unfalsifiable, because you cannot separate what the campaign caused from what would have happened regardless. The industry measures attribution and calls it proof.
Why we think this is worth getting right
Two values meet here. Economic alignment: precise audiences protect the brand's most important asset, its sender reputation, and stop margin leaking to people who were never going to convert. And accountability: to prove the system adds value, you have to be willing to withhold it from someone.
Both push toward the same capability, deliberate control over the audience with a holdout built in. Getting this right is what lets every other claim in the platform be stated as incremental lift rather than hopeful attribution.
How LTV.ai approaches it
Everything resolves to a single audience object the send pipeline consumes, so shaping stays consistent whether it comes from a smart list, a follow-up trim, a suppression rule, or an upload. Inclusion and exclusion lists are built as boolean formulas over the calibrated segments from the prediction suite, for example include high purchase-intent and discount-insensitive customers, or exclude high-churn and low-click ones.
When a campaign is a follow-up, the system suggests excluding the unlikely-to-engage, but bounded: a cap on how much can be trimmed, and value floors so high-value customers are never silently removed. A preview shows the impact before anything commits, and the resulting audience is materialized with a distinct name, an auditable record of exactly who was sent to.
How it stays honest and compounds
The audience system carves holdouts, a control group that receives nothing, so lift is measured against people who did not get the campaign. This is what turns a revenue number into an incrementality number, and it is the mechanism behind the up to 22% lift figure.
Withholding is not lost revenue, it is how the claim becomes provable. As the prediction models sharpen, the segments these audiences are built from get more accurate, so the same logic targets better over time.
Frequently asked questions
Could a follow-up trim drop my best customers? No. Value floors protect high-value customers from exclusion.
What is the holdout for? Measuring true incremental lift against a control that received nothing.
Part of the machine learning behind LTV.ai.
See it on your store: book a demo.