Why is my subscription margin shrinking over time?

Last updated: June 27, 2026

Subscription margin shrinks over time because three forces stack: the acquisition discount that buys the first order, cost creep in COGS and shipping that the locked subscription price never recovers, and churn that strands fixed acquisition cost across a smaller surviving base. Each one is invisible alone. Together they bend contribution per subscriber down month over month.

Most dashboards show you blended subscription margin as one number, so the decay reads like noise. The number moves a point here, a point there, and nobody can name the cause. The fix is to stop looking at the blend and decompose the decay into its three drivers, then watch contribution per active subscriber by cohort. When you split it, the trend is no longer mysterious. It is arithmetic.

What actually makes subscription margin shrink

Subscription margin shrinks because the price is fixed at signup while almost everything underneath it moves. Three drivers do the damage, and they hit at different points in the subscriber's life.

The acquisition discount hits month one. You sell the first box at 30 to 50 percent off to win the subscriber, so that order carries thin or negative contribution before you even count acquisition spend.

Cost creep hits every month after. Your COGS, your pick-and-pack, and your carrier rates all drift up over a year, but the subscription price you quoted at signup stays frozen. The gap between rising cost and frozen price is pure margin loss, and it compounds the longer a subscriber stays.

Churn-stranded acquisition cost hits the cohort. You paid a fixed amount to acquire each subscriber. When subscribers cancel, that fixed cost does not disappear. It now has to be earned back by fewer survivors, so the effective acquisition cost per active subscriber rises every month the cohort shrinks.

Driver	When it hits	What it does to margin
Acquisition discount	First order only	Thin or negative contribution on month one
Cost creep (COGS + shipping)	Every month after	Frozen price stops recovering rising cost
Churn-stranded acquisition cost	Across the cohort	Fixed acquisition cost spread over fewer survivors

Why does my subscription margin look fine at signup and bad six months later?

Subscription margin looks fine at signup and bad six months later because the two healthy-looking moments are measuring different things. At signup you are looking at a single recurring order against today's cost. Six months later you are looking at a cohort that absorbed a discounted first order, paid down acquisition cost through churn, and now ships at a higher unit cost than it did on day one. The headline price never changed, so the decay hides in the cost line and the survivor count, not the revenue line.

This is why a price increase alone rarely fixes it. If you raise the price but leave the discount, the cost creep, and the churn untouched, you have patched one of three leaks. The other two keep draining.

How to decompose your shrinking subscription margin, step by step

Work through these five steps with one cohort, one product, and real numbers. The point is to attribute the decay, not to admire the total.

Step 1: Pull one acquisition cohort, not the blend

Pick all subscribers who started in a single month. Blended numbers mix new discounted subscribers with mature full-price ones and average the decay into invisibility. A cohort isolates it.

Step 2: Compute first-order contribution with the discount in

Take the discounted first-order revenue, subtract COGS, subtract shipping and fulfillment, subtract payment fees. That is your true month-one contribution before any acquisition spend. It will be far lower than your steady-state number.

Step 3: Compute steady-state contribution and watch the cost line move

For months two onward, use full price minus current COGS minus current shipping minus fees. Then recompute it three and six months out with the higher costs you actually pay then. The slope of that line is your cost creep.

Step 4: Spread fixed acquisition cost across survivors each month

Take total cohort acquisition spend and divide by the number of subscribers still active each month. As the cohort shrinks, that figure climbs. That climb is your churn-stranded acquisition cost.

Step 5: Stack the three and read contribution per active subscriber over time

Put month-one discount, the cost-creep slope, and the rising acquisition-cost-per-survivor on one view. Contribution per active subscriber is what they jointly determine. Read it cohort by cohort.

Worked example: a grain-free dog food subscription at 54 dollars a month

Run a real cohort. The product bills 54 dollars a month. The first box ships at 40 percent off to win the subscriber, so month one nets 32.40 dollars. COGS starts at 19.00 dollars and creeps up about 1 percent a month from ingredient and supplier increases. Shipping and fulfillment start at 8.50 dollars and creep about 1.3 percent a month from carrier surcharges. Payment fees are 2.9 percent plus 30 cents. Acquisition cost is 48 dollars per subscriber, and the starting cohort is 1,000 subscribers.

Here is contribution per active subscriber, month by month, with the cost creep and the shrinking survivor base in plain view.

Month	Revenue	COGS	Shipping	Fees	Contribution per active	Active subscribers	Acquisition cost per active
1	32.40	19.00	8.50	1.24	3.66	1,000	48.00
2	54.00	19.20	8.65	1.87	24.28	780	61.54
3	54.00	19.40	8.80	1.87	23.93	660	72.73
4	54.00	19.60	8.95	1.87	23.58	580	82.76
5	54.00	19.80	9.10	1.87	23.23	520	92.31
6	54.00	20.00	9.25	1.87	22.88	470	102.13

Read the three drivers straight off the table. The acquisition discount shows up in month one: 32.40 minus 19.00 minus 8.50 minus 1.24 leaves 3.66 dollars, and after the 48 dollar acquisition cost that subscriber starts 44.34 dollars in the hole. Cost creep shows up across the contribution column: steady-state contribution slides from 24.28 dollars in month two to 22.88 dollars in month six, a 1.40 dollar drop per subscriber driven entirely by COGS and shipping the frozen 54 dollar price never recovered. Churn-stranded acquisition cost shows up in the last column: the 48 dollar fixed cost, spread over 1,000 subscribers in month one, has to be earned back by only 470 survivors by month six, so the effective acquisition cost per active subscriber more than doubles to 102.13 dollars.

Now the cohort-over-cohort signal, which is the part a blended margin number will never show you. As your book matures and you acquire later cohorts against accumulated cost creep, steady-state contribution per active subscriber declines from one cohort to the next. A first cohort at 24.28 dollars, a second at 23.23 dollars, and a third at 22.18 dollars is a decay of 2.10 dollars, or 8.6 percent lower contribution per active subscriber across three cohorts, before you account for any difference in churn or discount depth. That cohort-over-cohort slide, not the within-cohort churn everyone already tracks, is the quiet reason a subscription book that looks healthy on revenue keeps losing margin.

Across the full six months, each original subscriber returns about 74.90 dollars in cumulative contribution. Subtract the 48 dollar acquisition cost and the cohort nets 26.90 dollars per original subscriber over the window. That is the real lifetime value picture, and it is far below what month-two contribution alone would suggest.

What it costs to skip this decomposition

Skipping the decomposition costs you the ability to act on any single driver. If all you have is a blended subscription margin trending down, your only levers are blunt: raise price across the board, cut the discount and lose conversion, or chase retention without knowing whether retention is even the binding constraint. In the dog food cohort, retention was not the main margin leak between months two and six. Cost creep and stranded acquisition cost were. A founder who reads the blend and reflexively pours budget into a win-back campaign is treating the wrong driver and watching margin keep sliding.

The deeper cost is timing. Cost creep and cohort decay are slow. They do not trip an alert, they do not show up in a single bad month, and by the time the blended number is obviously bad you have already shipped thousands of orders at the lower contribution. The expensive part of margin decay is not the rate. It is the months you ship before you notice. For a related view of how slow erosion hides until it is large, see how to detect margin erosion early, and for the per-order math that feeds these cohort numbers, see how to calculate true net margin.

How to find your own subscription margin decay figure

Use the five steps above on your single highest-volume subscription SKU. Pull one cohort, compute discounted first-order contribution, plot steady-state contribution at month two, four, and six, and divide acquisition spend by active survivors each month. The number you want at the end is contribution per active subscriber, tracked by cohort. When a newer cohort's steady-state contribution sits below an older one's, you have measured your cohort margin decay directly, and you can name which of the three drivers moved.

If low order value is also part of your subscription math, the per-order analysis in what low AOV per order really costs pairs directly with this cohort view.

Where Agentis fits

Decomposing one cohort by hand is doable once. Doing it continuously, across every SKU and every cohort, while costs creep in real time, is where a spreadsheet stops keeping up. Agentis is a real-time profit governance platform for high-volume Shopify Plus and ShopLine merchants. It monitors margin at the order and SKU level and flags or blocks unprofitable activity before it reaches the P&L. Profit governance is the practice of monitoring and enforcing margin rules in real time across every order, SKU, and channel, so unprofitable activity gets caught and corrected as it happens instead of discovered in a month-end report. For a subscription book, that means a creeping COGS line or a cohort sliding below its contribution floor surfaces the week it happens, not in the quarter-end review.

Frequently asked questions

Why is my subscription margin shrinking even though revenue is flat?

Subscription margin shrinks while revenue stays flat because the decay lives in the cost line and the survivor count, not the price. Your billed price is frozen at signup, but COGS and shipping creep up every month, and churn forces fixed acquisition cost onto fewer survivors. Revenue can hold steady while contribution per subscriber falls underneath it.

Is churn or cost creep the bigger driver of subscription margin decay?

It depends on the cohort, which is exactly why you decompose rather than guess. In the worked dog food example, cost creep and churn-stranded acquisition cost together outweighed within-cohort churn between months two and six. Run the five steps on your own cohort to see which driver dominates yours before you pick a fix.

How does the acquisition discount keep hurting margin after month one?

The acquisition discount itself only hits the first order, but it interacts with churn to do lasting damage. A subscriber acquired at a deep discount who cancels after two or three boxes never repays the discounted first order plus the acquisition cost. The deeper the discount and the faster the early churn, the longer the cohort stays underwater.

What is cohort margin decay and how is it different from churn?

Cohort margin decay is the decline in contribution per active subscriber from one acquisition cohort to the next, driven by accumulated cost creep and discount depth. Churn measures how many subscribers you lose. Cohort margin decay measures how much each remaining subscriber is worth, and a book can hold churn steady while cohort margin still slides.

Can I fix subscription margin decay with a price increase alone?

A price increase alone fixes one of three drivers. Raising the price recovers cost creep on existing subscribers, but it does nothing about the acquisition discount or the stranded acquisition cost from churn, and a clumsy increase can raise churn and make the stranded-cost problem worse. Decompose first, then target the driver that is actually binding.

Concrete next step for today: pull your single highest-volume subscription cohort, compute discounted first-order contribution and steady-state contribution at months two and six, and write down contribution per active subscriber for each. If the later number is lower, you have found your decay and which driver moved it.