Welcome Email Sequence Template

Start with two exports: subscribers, with the date each joined and how (a discount popup, a free guide, checkout), and orders, with a customer ID and date. Follow everyone who joined in one month for 90 days, or longer if customers take months to decide.

A running total of orders only rises, so quiet stretches are easy to miss. Ask instead, for each stretch of days: of the people who haven't bought yet, what share buy now? Statisticians call this a hazard rate. NIST's handbook of statistical methods, which applies it to product failures, notes that failed units play no further role, since only the survivors count¹. Here, a buyer leaves the count.

1. 1

   Group by signup date

   Take everyone who joined in the month. Picking by order date pulls in people who joined years ago.

2. 2

   Keep first orders only

   A customer with four orders counts once, on the day of the first.

3. 3

   Use bands that widen

   Days 0-1, 2-3, 4-6, 7-10, 11-14, 15-21, 22-34, 35-59 and 60-89, because orders thin out. To compare bands of different lengths, divide by their days.

4. 4

   Divide by who is left

   For each band, divide its first orders by the people with no order at the band's start.

In this template's worked example, 288 of 12,000 new subscribers ordered on day 0 or 1, a rate of 2.4%. The rate fell to 0.6% on days 11 to 14, then rose to 2.2% on days 15 to 21. Even per day, the rate doubled. No welcome email caused it: the last one went out on day 6.

One curve averages people who joined for different reasons. A 10% popup catches people already looking at a product, and a care guide catches people still learning. Sort sources into three to five groups by what people were doing when they joined, and build a table for each. Merge or set aside any group under about 500 signups a month: its table will be mostly noise.

Blended, the dip and climb look like one curve's late bump. Split by source, the bump is another group of people. Popup signups buy at once and fade, while guide signups barely buy in week one, then jump to 5.6% on days 15 to 21 and convert better overall.

Read each group's schedule off its own curve, not a stock one. Klaviyo, which sells email marketing software, recommends sending 3 emails over the course of a week² for a standard welcome series. That might suit the popup group, but ends more than a week before the guide group's peak.

Put emails in and just before each group's high bands, not in its quietest stretch. Where under 1% buy, an email won't sell much, so make it useful: answer the question that stalls a first order, such as which size to buy. Save the selling for the bands where a group buys.

Don't discount people who would pay full price. In the example, guide signups were 29% of the list but placed only 12.5% of orders using the early discount, then paid full price at their peak. On a product with a 40% margin, 10% off gives away a quarter of the sale's profit. The product pricing template works out each product's margin.

Then write each email for its group: pick up where popup signups left off on the product page, and answer what guide signups came to learn. The welcome email sequence writer drafts seven emails you can fit to these days.

Plan for people who stop opening. In the example, subscribers who opened nothing in their first 14 days later bought at 0.6%, against 7.7% for those who opened something. That argues for dropping non-openers at day 14. Split by source, it holds for only one group: popup non-openers bought at 0.3%, guide non-openers at 2.3%.

Set the cutoff per group, weighing the orders you would lose against complaints. Google's guidelines for Gmail senders ask you to keep spam rates reported in Postmaster Tools below 0.10%³ and never reach 0.30%. Check clicks as well as opens, since Apple's Mail Privacy Protection helps users prevent senders from knowing when they open an email⁴.

In a longer sequence, people may unsubscribe from an email sent weeks before. In the US, the FTC's CAN-SPAM guide says an opt-out mechanism must be able to process opt-out requests for at least 30 days after you send your message⁵. Each request must be honored within 10 business days.

A click before an order doesn't prove the email caused it. In field experiments on eBay's search ads, economists found that because clicks and purchases go together, returns from paid search are a fraction of conventional non-experimental estimates⁶. Revenue your email platform credits to a message has the same blind spot.

Make the first test a holdout. Randomly keep about 10% of one group out of the new emails, and compare first orders per subscriber, not opens or credited revenue. In the example, the held-back group places about 28 orders a month, so the test needs about three months.

Evan Miller, who writes about A/B testing statistics, advises you to decide on a sample size in advance and wait until the experiment is over⁷. Stopping at the first good result inflates false positives.

Small lists can only detect big effects. Split the example's day-17 email and each side gets about 1,700 people and 94 orders a month. At the usual 95% confidence and 80% power, that can only reliably detect a lift of about 40%. So test big changes first: timing, price and whether to send at all.

To skip the arithmetic, open this template in River with your two exports: it works out when each group buys before suggesting a schedule. When the sequence ends, the newsletter writer drafts the regular emails subscribers move on to.