Retention marketing is the line in your budget nobody argues about. Then the win-back codes go out, the saves get counted, and nobody divides the discount by the margin. What is that motion costing you per customer?
The order came in on a Tuesday and the win-back campaign reported a recovery. A customer who had not bought in eleven months used the code, spent $120, and your dashboard turned green. What it did not show was your gross profit.
At a 60% margin, a 30% discount does not take 30% off your profit. It takes half: $84 in revenue against $48 in cost, so $36 of gross profit instead of $72. The recovery was real. So was the $36. Every retention marketing report I have seen counts the save. Almost none count the $36.
One order is survivable. The trouble starts when a whole segment learns the pattern. That campaign bought back a customer and taught a group of them the exact conditions under which your price drops.
The Retention Math Nobody Shows
Retention gets sold as the cheap channel. Sometimes it is. What decides anything is retained gross profit against the full cost of the motion, and the full cost almost never includes the discount.
Three numbers decide it: what the customer would have spent if you had done nothing, what the motion costs per customer including the discount, and what gross profit is left after both. The first number is never zero, which is why the figure you can actually claim is smaller than the one in the report.
Plenty of spreadsheets stop at the second number and call the difference retained revenue. Retain a customer who was coming back anyway and you have bought nothing.
The discount has the worst arithmetic. Divide it by your margin to see the share of gross profit it removes. A 30% discount against a 60% margin removes half. A 15% discount against a 30% margin removes the same half. That is why the same promotion feels trivial to a software business and catastrophic to a grocer.
Now the payback. The $36 you gave up comes back after one full-price order at $72 of gross profit. One order. That is why the tactic spreads: it pays back fast enough to survive a quarterly review.
The problem is what happens next. If the customer now only returns when there is a code, every future order earns $36 instead of $72. Six orders later you have given up $216 in margin on a customer whose behavior has not changed, only their price has. That is discount drag: the discount stops being a one-time cost of recovery and becomes a permanent line in your margin.
The three signals that expose it are the ones any honest set of marketing KPIs already carries: gross margin per order before and after the promotion, the share of orders that used a code, and the repeat rate of discounted customers against full-price ones. If the third is lower, you are not creating attachment. You are training a price.
Depth is usually copied from someone else's offer. Model 10%, 15% and 30% against your margin, then pick the shallowest one that still moves behavior. The $120 order above is an illustration.
The Four Motions, in the Order They Pay
Four retention motions are worth funding, and they are not interchangeable. Run them cheapest first, because each one changes the baseline for the next. Skipping ahead is how teams pay a discount to fix a problem an onboarding email would have prevented.
Motion | What it costs | What it buys | Where it stops paying |
|---|---|---|---|
Onboarding completion | Attention and product work, no margin | The first session reaching the moment of value | Never. It is the floor under the other three |
Usage and activation | Instrumentation plus triggered messages | A signal that arrives weeks before revenue does | When you cannot act on the signal you already collect |
Win-back | A discount against gross profit, plus the message | A customer whose lapse had a nameable cause | The moment the code is the only reason they return |
Loyalty and advocacy | A platform fee plus the reward; advocacy costs less | Repeat behavior and referrals | When it funds behavior the first three should have created |
Onboarding Completion: The Cheapest Retained Dollar
Onboarding is the one motion whose cost is mostly attention: you remove friction from the first session instead of buying media or giving away margin.
Amplitude's 2025 product benchmark, drawn from more than 2,600 companies and 10,600 products, finds the median product activates 5% of new users on day one, while the top decile reaches 21%. That gap is the size of the prize.
Fix the first session before you fund anything with a discount attached. A user who never reaches the moment the product is supposed to deliver cannot be retained by a reward, a reminder or a code. They were never in the market you think you are pricing.
Usage and Activation: The Signal That Arrives First
Purchases are a lagging indicator. They tell you what happened last month. Usage tells you what is about to happen.
The same benchmark puts a day-7 return rate of 7% in the top quartile, and finds that at the median product 98% of new users are inactive two weeks after their first action. By the time a customer looks lost in your revenue data, they left weeks earlier.
Triggered behavior is where this motion earns its money. Omnisend's 2025 ecommerce benchmark, built on more than 20 billion campaign emails and 470 million automated sends across 27,000 brands, puts automated email at 3.41 USD per email sent against 0.155 USD for scheduled campaigns, about 22 times better. That ratio separates marketing automation tools from a send button. The gap is timing: a message fired by behavior lands on a decision in progress, a scheduled campaign lands on a calendar.
One caution. ChartMogul and Kyle Poyar's December 2025 retention report, covering 3,500 software companies, shows median gross revenue retention for AI-native businesses moving sharply upward through the year, with the best carrying roughly twice the gross retention and two and a half times the net retention of their early peers. Read that as direction, not a target. Last year's good number is this year's median.
Win-Back: When It Works, and What the Discount Costs
Win-back has the clearest story and the worst economics, which is why it gets funded first and judged last.
It works when the lapse had a cause you can name and remove: a failed payment, an expired card, a plan that no longer fit, a seasonal need that comes back. Those customers left for a reason that has since changed. A reminder plus a fix is enough, and no discount is needed.
It fails when the customer was never really yours. Baymard's cart abandonment research, updated in September 2025 and averaged across 50 studies, puts average abandonment at 70.22%, and 42% of US online shoppers say they abandoned a cart in the past three months because they were just browsing or not ready to buy. That is the largest single block of abandonment, and no retention motion recovers a decision that was never made. Aim the budget at lapsed buyers, not at browsers.
Ignore the reactivation rates quoted around the web. They trace back to aggregator posts with no published sample, and none says what a recovered customer costs you. Take last quarter's win-backs, subtract the discount from their gross profit, and see what is left.
Loyalty and Advocacy: The Last Motion, Not the First
Loyalty comes last because it is the most expensive way to buy behavior the first three motions should have created: a monthly platform fee plus the rewards themselves, for a benefit onboarding should already have delivered. Whether a rewards scheme earns that fee is a different argument, with a different set of numbers behind it.
Advocacy is the cheaper half of the motion, and the half that gets skipped. A customer who refers does the acquisition work for you at a cost you can count, which is why customer advocacy belongs next to retention in the budget rather than under brand.
What Changes at 500, 5,000 and 50,000 Customers
The thresholds below are our arithmetic, not an industry benchmark. Searches for a published study that ties retention motions to customer count turned up recruiting guides and vendor pages with no sample and no method behind them. What follows is a model with the assumptions on the table so you can substitute your own.
The model assumes a 60% gross margin, an average order of $120, and a hire priced from eCommerce Placement's 2026 salary guide: 110,000 to 165,000 USD base plus a 10 to 20% bonus for a Head or Director of Retention and Lifecycle Marketing, 75,000 to 105,000 USD for an Email, SMS or CRM Marketing Manager, 70,000 to 95,000 USD for a Retention Data Analyst. Those are a recruiting guide range, not a survey.
Under 500 customers. A hire at the low end of the manager range costs about $150 per customer per year. At $120 an order and two orders a year, a customer produces $144 of gross profit. The hire costs more than the customer is worth. At this size there is no retention department: whoever already owns marketing runs onboarding and activation, because those cost attention rather than cash. A discount-led win-back is the most expensive thing you can do per customer, and it is what usually gets done first.
Around 5,000 customers. The same hire costs about $15 per customer per year, a tenth of what a customer returns. Tooling becomes affordable, segmentation becomes possible, and the constraint stops being money and starts being measurement. A segment is finally big enough for a code to change behavior across it, which means a badly designed win-back here is the first one that can reset a price for good.
Put a payback figure on that. A 30% code on a $120 order at the same margin drops gross profit from $72 to $36, so each recovered customer costs $36. Recover 100 a quarter and it costs $14,400 a year, roughly a fifth of the manager hire. Whether that is cheap depends on a number you probably do not have: how many of the 100 would have returned anyway. If the answer is half, your real cost per saved customer is $72.
Tooling scales with contacts rather than headcount, which is why teams undercount it. Assume $5,000 a year: $10 per customer at 500, $1 at 5,000, ten cents at 50,000. The salary falls the same way. The platform often does not, and the two costs cross at different sizes.
Around 50,000 customers. The hire costs about $1.50 per customer per year, which is noise. The discount does not. Promotion intensity at this size is already high: McKinsey's work on precision revenue growth management finds that 28% to 50% of retail sales volume in European CPG categories is promotion-based, and some companies invest up to 20% of gross revenues in promotions. McKinsey's own framing is that visibility into net promotion impact is poor relative to that spend. If your retention budget at this size is mostly discount, you are running a pricing strategy through a marketing team.
Run the division with your own numbers. If your average order is $60 and customers buy once a year, that customer returns $36 of gross profit, so $150 per customer per year is four times what they produce. The 500 threshold moves up.
When Your Churn Report Has No Cause Column
Churn arrives in the report as a number with no cause attached. The cause field is empty because nothing in your stack ever joined what a customer did to whether they stayed.
The consequence is predictable. If the only per-customer record you keep is order history, the only explanation available for a stop is that the customer did not want the product any more. Product gets blamed by default, because it is the only story the data can support.
Two questions tell you whether you are in that position. Can you name, for each customer who lapsed last quarter, the last meaningful action they took before they stopped? Can you split churn by acquisition cohort and by what happened in their first thirty days? If both answers are no, you are counting departures.
This is the problem revenue attribution solves on the acquisition side, applied to the customers you already have. Until behavior and outcome sit in one record, every budget decision is a guess with a chart attached.
Retention Marketing: Frequently Asked Questions
What is retention marketing?
Retention marketing is the work of getting a customer who has already bought to buy again, stay longer and spend more, instead of replacing them. It covers onboarding, activation, win-back and advocacy. The useful question is not whether to do it, it is which of those four to fund first.
Is retention marketing cheaper than acquisition?
The claim that keeping a customer costs a fraction of winning one is repeated in almost every article on this subject, usually without a source attached. Treat it as a premise, not a finding. Retention is cheaper when the motion is onboarding or activation, where the cost is mostly attention. It stops being cheaper the moment the motion is a discount, because a discount is a price cut with a marketing label.
What is a good retention rate?
There is no universal good number. Published tables vary by category, and Omnisend's ecommerce benchmark report is one of the few that publishes one at all. Its own rows sit far apart, which is the useful part: a rate only means something next to your own trend and cohort split. A rate that stays flat while your acquisition mix shifts is an average hiding two different businesses.
Conclusion
Retention marketing is a sequence before it is a channel: onboarding before activation, activation before win-back, win-back only where the lapse had a cause, and advocacy once the first three work. The motions that cost attention go first because they are cheap and they compound. The motion that costs margin goes last.
So before the next win-back campaign goes out, answer one question. What will that code cost you in gross profit, and how many orders will it take before the answer stops being a rounding error?
Cost the motion before you fund it
Four motions, one order, and a discount that does not hide in the margin. Put the numbers in one place before the next win-back campaign goes out.
