Lifecycle and CRM
Lay out lifecycle mail as a stage-by-segment matrix, price one send in caused upgrades against the paid cost of a customer, and see why the list is the limit.
On this page
Lifecycle marketing writes to people who have already signed up, so it skips the most expensive link in every other channel: paying to be seen. One send to 20,000 users costs about $120 and can move dozens of upgrades. Its limit is not cost but the list, which is finite, already paid for and worn down a little by every send, and its honest count is the upgrades a send caused, not the ones that happened to follow it.
The stage-by-segment matrix
A lifecycle program is a grid. The columns are the stages a user passes through: new, active on the free plan, paying, lapsed. The rows are segments that want different things at the same stage. For the invoicing app from the funnel and AARRR, the useful split is freelancers against small agencies of two to ten people, because an agency's account is worth more and succeeds by a different action.
Each column is a program with its own goal and its own trigger:
- Onboarding, days 0 to 14. The goal is the activation event: a first invoice sent within seven days for a freelancer, a colleague invited for an agency. It is triggered by what the user has not yet done, not by the calendar.
- Upgrade, active free users. The goal is the paid plan, triggered by a sign of need, such as a third invoice in a month or a client who pays late.
- Expansion, paying customers. The goal is an annual plan for a freelancer, which brings cash forward and lowers churn, or another seat for an agency.
- Win-back, users quiet for 30 days. The goal is one return visit that does the activation event again.
The naive program is one newsletter to everyone, which serves no cell: the new freelancer needs to send an invoice, the agency needs a second user, and the paying customer needs nothing from a message about the free plan. The grid also says where to start. Onboarding comes first because every later column inherits its rate: an activation lift raises paying customers from every channel at once, which is why the funnel lesson picks activation as the stage where a 10% lift is usually cheapest and fastest to read.
One send, worked
The upgrade send goes to the 20,000 active free users in the two shaded cells. At the course's email rates:
sent 20,000
opened x 40% -> 8,000
clicked x 5% -> 400
upgraded x 10% -> 40
cost 20,000 sends at $1 per 1,000 = $20, plus an hour of writing at $100 = $120
That is $3 an upgrade. Compare it with the right paid figure. An upgrade is a payment, so the comparison is the $333 a paying customer costs on the course's paid chain, not the $67 sign-up that the arithmetic a marketer carries quotes for the platform's event. Forty paying customers from paid would take about $13,300 of media; at the $521 the app's own funnel pays in module 2, more.
Count from clicks, not opens. Mail clients that load images on delivery record opens that nobody made, so the 40% is the softest number in the chain, and the 400 clicks are the first figure a reader can trust.
The mistake: counting every upgrade that followed the send
Some of the 40 would have upgraded that week anyway. They hit the free plan's limit, or a client paid late and they wanted reminders. The report that credits the send with 40 credits it with those too.
The fix is a holdout. Keep a random 2,000 active free users, a tenth as many as are mailed, out of every upgrade send, and compare the two groups' upgrade rates. Assume the holdout's rate says 15 of the 40 would have paid without the send. The send caused 25.
Re-priced on what it caused, the send costs $120 ÷ 25 = $4.80 an upgrade and brings 25 × $368 = $9,200 of 12-month margin, the value paid media basics uses so that every channel is judged inside the same 12-month payback window; from paid, the same 25 customers would cost 25 × $333 ≈ $8,300. The conclusion survives. What changes is the program that gets built on it, because a team reading 40 plans for 40.
One send's holdout is too small to read. At the holdout's rate of 15 in 20,000, its 2,000 people produce about 1.5 upgrades in a send's window, which noise swamps. Keep the same 2,000 out of every upgrade send for a quarter: after six sends they should show about 9 upgrades, give or take 3, where the mailed rate would predict 24, and that gap is readable. A slow, repeated effect needs a standing holdout rather than a one-off test, which is the argument in when an A/B test is the wrong instrument.
The list is the limit
Paid media's limit is a price that rises as spend grows, as paid media basics prices it. Lifecycle's limit is the list. There is no second set of 20,000 people who already use the product, and none can be bought at $120. Twenty-five caused upgrades a month is the program's size until acquisition grows the list, which is why lifecycle multiplies the other channels and cannot replace them.
Every send also spends some of the list. Assume 0.3% of recipients unsubscribe per send, 60 people who cannot be mailed again. A list member is worth the upgrades future sends would have caused: at 25 a month from 20,000, about 1.5% of members a year, so roughly 1.5% × $368 ≈ $5.50 of margin a year each. The 60 cost about $330 a year, more than the send's $120 and still far below its $9,200. The cost that can be large is the one the arithmetic hides: when complaints rise, mailbox providers start routing the sender's mail to spam, and the onboarding mail, the most valuable column, stops arriving with it. That cost behaves like a cliff rather than a slope, so set a complaint-rate limit before raising the frequency.
In practice
To the team: one goal per cell, one standing holdout per program, and a monthly report of caused upgrades rather than attributed ones. Ask of every proposed send which cell it serves and which number it moves. To whoever sells the email platform: ask whether it can hold out a random share of every program and report the difference, because a dashboard without one reports the 40.
The mistake that costs money is crediting a send with every upgrade that followed it, then sending more because the ratio looks so good. Forty counted, 25 caused; and the extra sends spend the list, and the sender reputation, that onboarding depends on.
Check yourself
1. Marketing proposes four upgrade sends a month instead of one. Assume each extra send causes a third as many upgrades as the first, about 8, and each send unsubscribes 0.3% of the list. Estimate the month and decide.
Caused upgrades rise from 25 to 25 + 3 × 8 = 49. The extra 24 bring 24 × $368 ≈ $8,800 of 12-month margin. The extra cost is 3 × $120 = $360 of sending and writing plus 180 more unsubscribes at about $5.50 a year each, roughly $1,000, so about $1,400 in all. Go weekly, on two conditions: the standing holdout confirms the extra sends cause what is assumed, and a complaint-rate limit stops the program before deliverability suffers.
2. After a quarter, the holdout upgrades at 0.19% per send window against 0.20% for the mailed group. What do you do with the upgrade program?
The difference is 0.01% of 20,000, about 2 upgrades a send, and over six sends the holdout's count of about 23 sits within its own noise of roughly ±5. The send causes almost nothing; the 40 were people upgrading anyway. Stop counting it as growth. Change the offer, for example an annual plan at two months free, test it against the same holdout, and move the writing hour to onboarding meanwhile.