"The timer assumed every dog eats at the same speed."
A DTC pet-supplies brand ran its replenishment emails on a fixed 30-day timer. The flow converted at 1.1%. Ninety days after Sendnexa switched the trigger to each customer's expected date of next order, conversion sat at 2.4% and revenue per recipient had risen from $1.60 to $2.75. Dana Okafor built the cohorts; Priya Raghavan ran the engagement.
| Metric | Before | After (90 days) |
|---|---|---|
| Replenishment placed-order rate | 1.1% | 2.4% |
| Flow revenue per recipient | $1.60 | $2.75 |
| Flows' share of email revenue | 23% | 37% (platform avg ~41%) |
| Sends to 180-day-inactive segments | 71% of volume | suppressed |
What did the baseline audit find?
Dana: The single worst number was that 71% of sends were going to segments with no purchase in 180 or more days. The replenishment flow was almost a footnote next to that, but it was the clearest example of the same disease: everything ran on assumptions instead of data. Thirty days is not a consumption cycle; it is a guess with a cron job.
How does the expected-date trigger actually work?
Priya: Klaviyo predicts an expected date of next order per profile once an account clears its thresholds: 500+ customers with orders, 180 days of history, three or more orders per qualifying customer. This brand qualified comfortably. We trigger the flow a category-sized window ahead of each customer's predicted date, seven days for consumables, and the models retrain weekly so the windows drift with actual behaviour.
Why did conversion more than double?
Dana: Timing is most of it. The old timer emailed a 60-day-cycle customer at day 30, twice too early, and a 21-day-cycle customer four days late. Both look like "unengaged subscribers" in a report when they are really mistimed sends. Put the same email inside the week the customer is actually running low and 1.1% becomes 2.4% without touching the creative.
Was there a rollout risk with swapping triggers?
Priya: We ran the new flow in parallel for 12 days against a held-back cohort before cutover, which is standard in ourflow-build work. The old timer kept sending the whole time. Cutover happened only after the parallel run cleared the baseline on placed-order rate, so the client never had a week of dead replenishment mail.
What is the honest caveat?
Dana: This works where purchases repeat on a cycle. Pet food, coffee, supplements: strong fit. A furniture brand has no meaningful expected next order date, and pretending otherwise produces noise. Check whether your ESP's predictive thresholds are met and whether your category has a real cycle before copying this; that check is the first hour of aworking session.
Client identity anonymized. Figures are representative engagement data; Klaviyo's published benchmarks (flows at ~41% of email revenue from 5.3% of sends) are the reference frame.
Does your category have a cycle?
If customers reorder on any rhythm at all, timer-based flows are leaving orders on the table. We can show you how many.
Book a working session