Learning Log
What a welcome flow should actually say
A log, not a template. We inherited a welcome flow that opened with money off and a timer, rebuilt it to open with the brand, and moved the reward to the end. The first email is not the one people think it is.
Almost every welcome flow I am handed opens the same way. Discount, expiry, shop now. It is the default because it works well enough to never get questioned - the first email always outperforms the rest, so it looks like the sequence is doing its job.
What that shape actually produces is a coupon delivery system with four emails attached that nobody reads.
This is the log of rebuilding one, for a brand where that mattered more than usual.
What was there
Tahan Outdoors is eight years old and close to a household name in Malaysia. It started as a marketplace for white-labelled kit and rebranded a couple of years ago after a hard look at what it wanted to be - a brand serving the outdoor community with genuinely good gear rather than one that moves units. The position it landed on is specific and true: gear built for the tropics, for the heat and rain its customers actually walk into.
New subscribers met none of that.
The welcome flow opened with RM15 off and a countdown. Which is exactly what the old marketplace would have said. Nothing in the sequence explained why this gear exists, who makes it, or what "built for the tropics" means when you are standing in a shop comparing two stoves.
What we saw
Three things, in the order we noticed them.
The first email did almost all the work. Opens, clicks, orders - concentrated in email one, then falling away through the rest of the sequence.
Opens fell through the sequence. Not a gentle taper. A drop.
The share of new subscribers who went on to buy was falling month after month. That was the number that made this urgent rather than interesting.
For a while I read the first two as normal. Welcome flows do taper; everyone's does. It was the third that changed the reading, because a sequence that is merely front-loaded should hold its conversion rate steady over time. One that is losing conversion month after month is not front-loaded. It is being ignored.
What I think was happening
The countdown taught people what the relationship was.
If the first thing a brand says is here is money off, and it expires, the subscriber learns the sequence is a transaction with a deadline. Emails two through six have to earn attention from a reader who has already been told what this is about, and already has the only thing they were offered.
Worse, the discount was doing the qualifying. People who wanted a deal took it. People who wanted to know what Tahan was got nothing, and left. A rebranded company was introducing itself in the voice of the company it stopped being.
I want to flag that this is a reading, not a measurement. We did not run an exit survey. It fitted the pattern and it fitted the brand history, and that was enough to act on - but it is interpretation, and I would rather label it than let it pass as finding.
What we built
Six emails. Story first, reward last.
The sequence opens with why Tahan exists and what gear built for the tropics actually means in practice - the specific problems of heat, humidity and sudden rain, and what a product has to do to survive them. Then the range, the people, the use cases. The offer arrives at the end.
We also simplified the offer itself: 10% off with no minimum spend, matched exactly in the sign-up pop-up so the promise made at capture is the promise kept in the inbox. The previous version had a value and a threshold, and any gap between what a pop-up promises and what the first email delivers is a small betrayal at the worst possible moment.
And we staged the testing rather than changing everything and hoping: subject lines and timing first, then clicks and orders. If you move all the variables at once you get a result you cannot attribute, which is the same as no result.
What happened
Open rate on the first email went from 58.2% to 68.1%, a 9.9 point lift. The share of recipients who ordered from the sequence rose by 225%.
Both figures are published on the case study, with the client's approval.
The second number is the one that matters. A better first email is easy to produce and easy to fool yourself with. A rising share of subscribers who actually buy means the sequence is doing the job the sequence exists for, rather than just presenting a coupon more attractively.
The discount, incidentally, converted better from the back of the sequence than it had from the front. Offered to someone who now knows what the brand is, the same percentage reads differently. That is the discount argument from Lab Note 04 showing up in a flow instead of a campaign.
What I got wrong
I expected to lose the quick wins. My assumption was that moving the offer to the end would cost us the impulse buyers who convert on email one, and that we would make it back on engagement further down. That is not what the numbers say. Email one got better without the offer in it, which I did not predict and still find slightly surprising. The best explanation I have is that a specific, concrete opening about tropical conditions is simply more interesting than a generic discount, and interest is what earns an open.
I underestimated the pop-up mismatch. I treated matching the pop-up offer to the email as hygiene, a tidy-up rather than a lever. Looking back at where the drop-off sat, I think it did more than that. Untested, so I am not claiming it.
The limits
- One brand, one category, one market. Malaysian outdoor gear, a rebranded company with a genuinely interesting story to tell. A brand without that story cannot copy this and expect the same result - the move is not "delay the discount", it is "have something worth saying first".
- Not a clean A/B. We rebuilt the sequence rather than running old against new in parallel, because the old one was actively losing conversion and holding it live to be fair to the test was not defensible.
- Open rates are softer than they were. Mail Privacy Protection inflates them. The conversion figure is the one to weigh.
What we are testing next
- Whether six is the right length, or whether the same content lands harder in four.
- Whether the reward needs to be at the very end or works as well in the middle, once the brand has been established.
- Whether the effect holds for subscribers who arrive from paid traffic rather than organic. Different intent at capture, possibly a different tolerance for a story before an offer.
Open your own welcome flow and read only the first email. If a stranger read nothing else, what would they now know about you that they could repeat?
Frequently asked questions
What should a welcome email series include?
Enough for a new subscriber to describe the brand to someone else. In practice that means why the brand exists, what makes the product what it is, and the use cases that show it working, with the offer placed where it converts rather than where it qualifies. A sequence that contains only a discount has told the subscriber nothing to remember.
How many emails should be in a welcome flow?
Enough to carry the content and no more. The sequence described here is six, because that is what the story needed. Picking a number first and finding content to fill it is how welcome flows end up with three emails nobody opens.
Should a welcome email include a discount?
A discount converts well, but leading with it teaches the subscriber that the relationship is transactional and gives away the only thing you were offering before they know what they are buying into. Placing the reward later in the sequence, after the brand has been explained, converted better in the rebuild described here.
Why do open rates drop through a welcome sequence?
Some taper is normal. A sharp drop usually means the first email set an expectation the rest cannot meet - most often because it led with an offer, so every later email is competing against a reader who already got what they came for. Watch the share of subscribers who eventually order rather than opens alone; if that is falling month over month, the sequence is being ignored rather than merely front-loaded.
