Case study · Food delivery · August 2026

The pace of sampling

This client's sequence gave anyone who had not yet tried a product group a sample of exactly that. The idea converted 17.3% of contacts, but at its own pace: almost half of the attributed purchases took more than five days to happen.

In August 2026, a food delivery chain ran three targeted sampling actions on WhatsApp at the same time. The month closed with 962 recipients delivered, 166 unique buyers, 182 orders and R$ 15,269.48 in attributed revenue, within a window of 15 calendar days after each message was delivered. Client names and any store identifiers were removed from this material. The operating numbers were not.

A sample does not convert the same day

Unlike an offer with an immediate discount, giving people a sample of something they have never tried takes time: they receive it, taste it, decide. Only 6.6% of attributed buyers bought on the day of the contact. Almost half, 47.0%, took more than five days.

Same day 1 day 2 days 3–5 days 5+ days 6.6% 12.0% 13.3% 21.1% 47.0%
Distribution of the time between message delivery and the first attributed purchase, across the month's 166 unique buyers.

Handing out a taste is not selling on the spot. It is planting the next purchase.

Recency beat the product

What most separated buyers from non-buyers was not which sample they received. It was how long it had been since they last bought anything. People who had bought in the 30 days before the contact converted 42.9% of the time. After 91 days of silence, the rate falls to less than a fifth of that.

0% 10% 20% 30% 40% 50% 0–30 days 31–60 days 61–90 days 91–120 days Over 120 days 42.9% 24.0% 13.8% 7.6% 7.7%
Sales conversion by days since the last purchase before the contact. Reliable history from July 2025 onward.

Three samples, three audiences

The sequence ran three actions in the month, each one sampling a different product group to people who had not yet ordered from that group. The action with the smallest reach, aimed at an audience already classified as frequent customers, brought the lowest volume but the highest return per person contacted.

acao_borda
15.8% conversion
Recipients delivered
158
Unique buyers
25
Attributed revenue
R$ 2,372.96
Average order value
R$ 91.27
Revenue per recipient delivered
R$ 15.02
Average time to purchase
6.4 days
acao_sobremesa · priority of the month
17.6% conversion
Recipients delivered
692
Unique buyers
122
Attributed revenue
R$ 10,720.74
Average order value
R$ 80.01
Revenue per recipient delivered
R$ 15.49
Average time to purchase
5.4 days
acao_vip · restricted audience
17.0% conversion
Recipients delivered
112
Unique buyers
19
Attributed revenue
R$ 2,175.78
Average order value
R$ 98.90
Revenue per recipient delivered
R$ 19.43
Average time to purchase
5.4 days

acao_vip, limited to customers with 8 or more purchases in their history, reached 6 times fewer people than acao_sobremesa, yet earned R$ 19.43 per person contacted, against R$ 15.49. A smaller, better-qualified audience beat volume on revenue per contact.

The loyalty ladder shows up here too

Even within a strategy built around gaps in what customers consume, rather than how often they buy, purchase history still mattered. Customers classified as VIP converted at more than twice the rate of customers with few purchases in their history.

0% 10% 20% 30% 40% 50% Three purchases Recurring (4–7) VIP (8+) 10.3% 17.2% 24.3%
Sales conversion by purchase-frequency profile before the contact, 13-month rolling window. The "two purchases" category had only 3 recipients delivered in this slice and was left out of the chart because the sample was too small.

How we read these numbers

Only confirmed purchases count toward conversion, attributed to the most recent eligible delivery within a window of 15 calendar days. Being in the scheduled queue, being accepted by the API or having the message read does not count as a conversion. None of the 962 recipients delivered in the month had received another contact in the 30 days before their first delivery, which rules out fatigue from over-sending as an explanation for the differences observed.

What is not here. The figures come from a single monthly data slice and were not calibrated for statistical comparison between actions. The readings in this case study are qualitative, not causal. With no cost-per-send data in this slice, there is no ROI or profit to calculate, and none was estimated.
What was removed. Client names and any store identifiers were taken out of this material before publication. The internal action identifiers (acao_borda, acao_sobremesa, acao_vip) were kept because they do not reveal the client. The operating numbers are real.
Want a result like this

Find out where your customer base holds recoverable revenue.

The Recoverable Revenue Diagnosis helps you map the starting point of your post-sale operation.

I want a diagnosis to recover revenue