AI audience analysis
Most lists are mailed like everyone on them is the same person. The data says otherwise: in one baseline audit we found 71% of sends going to segments with no purchase in 180+ days. This service turns your ESP's predictive scores, churn risk, predicted CLV, and expected date of next order, into segments and suppressions that decide who gets mailed at all.
What do the models actually predict?
Klaviyo computes three scores per profile and retrains the models weekly: churn risk (the probability a customer will not buy again), predicted customer lifetime value, and expected date of next order. Omnisend and Customer.io expose comparable signals. The scores exist in most accounts already. What is usually missing is anyone acting on them.
We translate the scores into three working assets. A suppression list holds high-churn-risk profiles out of routine campaigns. CLV tiers split the list so your best customers stop receiving discount-led mail they do not need. And next-order-date cohorts feed the replenishment and win-back flows that convert timing into orders.
What does an engagement produce?
| Deliverable | What it is | Where it lives |
|---|---|---|
| Baseline audit | Send-vs-engagement map, complaint rate vs the 0.1% Gmail guidance, consent coverage | Report + working session |
| Suppression lists | Churn-risk and long-inactive exclusions applied to every campaign | Your ESP |
| CLV tiers | 3-tier segmentation with per-tier cadence and offer rules | Your ESP |
| Timing cohorts | Expected-date-of-next-order windows sized to your category (7 days for consumables, 14 to 30 for considered purchases) | Your ESP + flow triggers |
| Runbook | One page per segment: contents, trigger, review cadence | Shared doc, yours to keep |
How is this different from "AI email tools"?
We do not sell a tool, and we do not promise that a model writes your strategy. The predictive scores come from your platform; the judgment about what to do with them comes from people who have run this playbook across DTC and SaaS lists up to 350k subscribers. The coffee-subscription case study shows the full sequence, including the two weeks where cutting sends made the dashboard look worse before placed-order rate moved.
What data do you need access to?
Read and build access to your ESP, nothing more. The models we act on are computed by the platform itself: Klaviyo, for example, produces churn risk, predicted CLV, and expected date of next order once your account has 500+ customers with orders and 180 days of history. We never export subscriber data off your systems.
What if our account is below the predictive thresholds?
Then we build the proxy version first. Recency, frequency, and monetary segments approximate the same decisions while your order history accumulates: a 90-day non-engager suppression behaves much like a churn-risk suppression in practice. One pet-supplies client started exactly this way, and the model-based triggers replaced the proxy segments at the 6-month mark.
Does suppressing subscribers actually help revenue?
Suppression trades reach for reputation, and the trade is usually favourable. Our coffee-subscription client cut send volume 22% with revenue flat, while spam complaints fell from 0.19% to 0.06%, safely under Google's 0.1% guidance. Deliverability gains then compound: mail that lands in the inbox earns more than mail sent to more people.
Who maintains the segments after the engagement?
You do, and that is deliberate. Everything is built inside your ESP with naming conventions and a one-page runbook per segment: what it contains, what triggers it, when to review it. Most teams need about an hour a month of maintenance. We stay available for model reviews quarterly if you want them.
Find out what your list is hiding
The baseline audit runs in the first two weeks of any engagement. Bring your ESP login to a working session and we will show you the first three findings live.
Book a working session