Most brands treat their subscription programme as a feature. Something the platform switched on, sitting in a widget on the product page, owned by whoever set up Recharge or Skio. Conversion work happens somewhere else entirely, on the one-time purchase path, run by a different person against a different number. That split is why so many subscription programmes stall at a low single-digit share of orders and never move.
Subscription design and conversion design are the same job. Every decision that shapes whether someone subscribes is a conversion decision: how the buy box is built, what the default frequency is, what the discount says about the product, what happens in the ninety seconds after checkout, and how easy it is to pause instead of cancel.
This is what we mean by subscription conversion rate optimisation. Not testing button colours on a subscribe widget, but treating the whole programme as one funnel and optimising it end to end.
Each section below includes the specific numbers we pull from Recharge or Skio to check whether the problem is actually there. Opinions about subscription UX are cheap. The data tells you which of the twenty possible fixes is the one worth building.
Subscription CRO audit versus standard CRO audit
The two look similar on a scope document and share almost no methodology. A standard CRO audit follows a session. A subscription CRO audit follows a person across months.
| Standard CRO audit | Subscription CRO audit | |
|---|---|---|
| Primary metric | Conversion rate, AOV, revenue per session | Subscription share of orders, billing cycles completed, subscriber LTV |
| Unit of analysis | The session | The subscriber lifecycle |
| Measurement window | Two to four weeks | Three billing cycles minimum |
| Where the funnel ends | Order confirmation | Somewhere around cycle three |
| Surfaces reviewed | PDP, cart, checkout | Buy box, checkout, post-purchase, subscriber portal, cancel flow, dunning emails |
| Core data sources | GA4, Shopify, heatmaps, session recordings | All of those, plus Recharge or Skio subscriber, cohort and cancel-reason data |
| Biggest failure mode | Testing a change too small to matter | Winning the sign-up and losing the subscriber |
| Who usually owns it | Ecommerce or CRO lead | Nobody. It sits between dev, retention and customer support |
That last row is the real finding on most audits. Subscription conversion fails in the gaps between teams, which is why it survives years of CRO work that never touches it.
What to pull before you start
From Shopify: subscription versus one-time order split by month, by product and by new versus returning customer. Repeat purchase rate for one-time buyers. Median gap in days between repeat orders of the same product.
From Recharge or Skio: active subscriber count over time, new subscriptions per week, churned subscriptions per week, cohort retention by sign-up month, cancellation reasons with volumes, and cancellations grouped by which billing cycle they happened in.
From GA4: PDP sessions, add-to-cart rate and checkout completion, segmented by whether the subscription option was selected.
From your helpdesk: ticket volume tagged as subscription admin. Every one of those is something the portal should have handled on its own.
Run these before forming a view. The cancellation-by-cycle number in particular will tell you which section of this post to work on first.
Why subscription CRO is a different discipline
Standard ecommerce CRO optimises for a single decision. Someone lands, considers, buys or does not. The test is clean because the outcome is immediate.
Subscription asks for something harder. You are asking a shopper to commit to a purchase they have not made yet, at a cadence they have to guess at, for a product they may not have tried. The friction is not in the interface. It is in the commitment.
That changes what you optimise for. A test that lifts subscription sign-ups by a fifth but pushes churn up in month two has made the business worse. The number that matters is not the sign-up rate. It is the sign-up rate multiplied by how many billing cycles those subscribers survive.
The metric you are actually optimising
Before running a single test, agree what a win looks like. On most of the programmes we work on, that is subscription share of orders and average billing cycles completed, tracked together. Either one on its own is easy to game.
Set the measurement window at three billing cycles minimum. A weekly coffee subscription will show you the answer in a month. A quarterly beauty refill will not tell you anything useful until the following quarter, and running a two-week test on it is theatre.
What to pull from Recharge or Skio: cohort retention by sign-up month, so you can see whether a change made in March is still holding in June. Pair it with subscription AOV against one-time AOV. If sign-ups climb while cohort retention drops, the test lost even though the dashboard says it won.
The subscription buy box
The buy box does more work than any other element on a subscription site. It has to present two purchase models side by side, explain the difference, justify the discount, and set a delivery frequency, without turning into a form.
Most default app widgets fail at this. They stack radio buttons, a dropdown and a small-print discount line into a block that reads as admin. Shoppers skim it and pick the option that requires no thinking, which is one-time.
Should subscription be the default option?
Usually, yes. Pre-selecting subscribe moves the effort onto the shopper who wants a one-off, and most product categories that suit subscription in the first place are ones people buy repeatedly anyway.
The exception is a considered first purchase at a high price point, or a flavour-led product where trying before committing is genuinely the sensible route. Defaulting to subscribe on a forty pound serum bought by a first-time visitor tends to shift the drop-off point rather than remove it.
Test it properly rather than assuming. And when you do default to subscribe, make the one-time option visible and easy to reach. Hiding it reads as a trap, and it is the fastest way to earn a cancellation in cycle one.
What to pull to validate this:
- New subscriptions created per week by product, from Recharge or Skio. Compare that against Shopify's one-time order volume for the same product to get a true subscription share per SKU rather than a site-wide average.
- Cancellations occurring in cycle one, from the cancel-reason report. A spike here, especially with reasons like "did not mean to subscribe", means the default is trapping people rather than persuading them.
- Refund and dispute rate on first subscription orders in Shopify. Same signal, harder to argue with.
- First-order subscribers against subscribers who converted later. If almost nobody subscribes on first purchase in a high-consideration category, stop fighting the PDP and move the effort post-purchase.
Frequency choice and cadence
Offering eight delivery frequencies feels generous. It is not. Every extra option is another decision, and the shopper has no idea how fast they get through the product because they have never bought it.
Three options, with the middle one preselected and labelled with the reason, converts better and churns less. "Every 4 weeks - most popular for two coffee drinkers" does more work than a bare dropdown. You are not offering choice, you are offering a recommendation.
Get the default cadence from actual reorder data rather than instinct. If the average gap between repeat orders is thirty-eight days, a four-week default guarantees a pile-up and a pause request by cycle three. See subscription ecommerce design examples for how this plays out visually across different categories.
What to pull to validate this:
- Distribution of chosen frequencies across active subscriptions in Recharge or Skio. If ninety percent of subscribers sit on the default, your other options are decoration. If they are scattered evenly, your default is not doing its job.
- Frequency changes made in the first ninety days, and the direction of travel. Subscribers lengthening their interval means the default is too fast. Almost nobody shortens it, which is why erring slower is the safer bet.
- Skip and pause volume per frequency. A high skip rate on one cadence is the same signal as a frequency change, from people who could not be bothered to edit their settings.
- Median gap between repeat one-time orders of the same product in Shopify. That is your real consumption rate, and it should set the default.
Converting one-time buyers into subscribers
The largest untapped pool of subscribers on most stores is the people who have already bought once and liked it. They have solved the trust problem. They know the product works. They just have not been asked at the right moment.
Product page conversion gets almost all the attention here, and it is the hardest place to win. The second and third purchase are easier ground.
What to pull to size the opportunity: the count of Shopify customers with two or more one-time orders and no active subscription. That is your addressable pool, and on most stores it is larger than the current subscriber base. Cross-reference it against Recharge or Skio to see how many existing subscribers started as one-time buyers rather than subscribing on first order. That ratio tells you where your programme actually recruits from.
What discount actually shifts behaviour
Under 10% and the discount reads as a rounding error, not a reason to commit. Above about 20% and you have taught the customer that the product is overpriced at full rate, which makes every one-time purchase feel like a mistake and every price rise a fight.
The useful range sits between those two. What matters more than the exact number is what sits alongside it. Free delivery on subscription orders, early access to new releases, or a member-only size often move more people than another five percent off, and they cost less margin.
Discount is the weakest lever available and most brands reach for it first. It is worth exhausting the others before discounting deeper.
What to pull to validate this:
- Cohort retention split by discount level, if you have ever run more than one. Deeper-discount cohorts that churn faster are a clear signal you bought sign-ups rather than subscribers.
- Subscription AOV against one-time AOV in Recharge or Skio. If the subscription figure sits well below, the discount is doing the work a bundle or a larger default quantity should be doing.
- Contribution margin per subscription order after discount and shipping, from Shopify. Some programmes are running at a loss until cycle three, which changes what an acceptable churn rate looks like.
- Cancellation reasons tagged as price. If price barely registers as a reason, discounting deeper solves a problem you do not have.
The post-purchase window
The best moment to convert a one-time buyer is immediately after they have bought. They have committed, the card details are already there, and the decision is a small extension of one they have just made rather than a fresh one.
Both Recharge and Skio support post-purchase conversion, and the mechanics differ between them. Which one suits a brand depends on the wider programme design rather than this single feature, something we cover in our comparison of Recharge and Skio.
The second window opens when the first order is nearly used up. That is an email job rather than an interface one, and it depends entirely on knowing the product's real consumption rate. Our guide to Klaviyo flows for subscription brands covers how to build that trigger.
What to pull to validate this:
- Impression and acceptance rate on any post-purchase subscription offer. A low impression count usually means the offer is misconfigured rather than unpopular, and it is worth checking before you rewrite the copy.
- Retention of post-purchase converts against PDP converts. In our experience they behave differently, and if post-purchase subscribers churn faster the offer is being accepted without being understood.
- Klaviyo performance on any replenishment or subscribe-prompt flow: conversion rate and revenue per recipient, not open rate.
- Days between first order and subscription start, for subscribers who converted later. That distribution tells you when to send the prompt.
Build-a-box as a conversion mechanic
Build-a-box gets treated as a merchandising feature. It is really a conversion tool, and it works because it removes the two objections that kill subscription sign-up: boredom and the fear of being stuck with the wrong thing.
When someone picks their own six items, the commitment stops feeling like a commitment. They are choosing, not signing up. The subscription becomes the delivery method rather than the product.
It does not suit every catalogue. Build-a-box needs enough range to make the choice feel real, and a price architecture where a mixed box makes sense. Below roughly eight or ten SKUs the builder feels thin and a curated bundle converts better.
Where it does fit, the conversion detail matters more than the feature itself. Prefill the box with a recommended selection so the shopper edits rather than starts from nothing. Show the running total and the saving live. Let people swap items between deliveries without contacting support, because the ability to change is most of the reason they subscribed.
What to pull to validate this:
- Box completion rate: builders started against subscriptions created. A low completion rate is a UX problem, and it is invisible in every standard conversion report because the session never reaches checkout.
- Average items per box against the minimum. If most boxes sit exactly on the minimum, the incentive to add more is not landing.
- Swap volume per subscriber per cycle, from Recharge or Skio. Regular swapping is the healthiest engagement signal a subscription programme produces. Zero swaps across a cohort usually means people have forgotten the subscription exists, and they will notice it next time the card is charged.
- Cohort retention of build-a-box subscribers against fixed-product subscribers. This is the number that justifies the build cost.
The subscriber portal is a conversion surface
Most brands think of the portal as account admin. It is the highest-intent page on the site. Everyone who opens it is an existing paying customer taking a deliberate action, and what happens next decides retained revenue.
Login friction alone loses subscribers. A customer who wants to delay one delivery, cannot get into their account, and emails support instead will often just cancel while they wait. Passwordless access removes a whole class of avoidable churn, and it is one of the first things we look at on any subscription build.
The portal is also where upsell actually works. Adding a product to an existing subscription is a two-tap decision for someone who already trusts the brand. Most portals never ask.
What to pull to validate this:
- Monthly portal login rate as a share of active subscribers. Low engagement is not a good sign, whatever the churn number currently says.
- Password reset requests and failed login volume. On password-based portals this is often the single largest source of avoidable support contact.
- Portal actions broken down by type: skip, pause, date change, swap, add product, cancel. The ratio of pauses to cancellations tells you whether the portal is presenting pause as a real option.
- Helpdesk tickets tagged subscription admin, matched against the actions the portal already supports. Anything a customer emailed about that they could have done themselves is a design failure with a measurable cost.
- Acceptance rate on any add-on or upsell shown in the portal.
Cancel flows that save without discounting
A cancel flow that opens with a discount offer teaches every subscriber that clicking cancel is how you get money off. It saves the current subscription and damages the next twelve.
Ask the reason first, then answer that specific reason. Too much product gets a delay or a longer interval. Wrong item gets a swap. Going on holiday gets a pause with a return date. Only price gets a price response, and even then a smaller size often works better than a discount.
Pause needs to be as prominent as cancel, with a specific return date rather than an open-ended skip. Under the DMCC Act, cancellation also has to be straightforward, so the flow needs to be a genuine offer of alternatives rather than an obstacle course. We covered what that means in practice in our piece on the DMCC Act and subscription rules.
What to pull to validate this:
- Cancellation reason distribution with volumes, from the Recharge or Skio cancel flow. If the top reason is "other", the reason list is wrong and every save offer downstream is guesswork.
- Save rate by reason and by offer type. Most flows have one offer doing all the saving and three that never work, and you cannot see that from an overall save rate.
- Cancellations grouped by billing cycle number. Cycle one cancellations are an acquisition problem. Cycle four onwards is usually product fatigue, and it needs a different fix.
- Reactivation rate of paused subscribers against churned ones. If paused subscribers come back at a decent rate, pushing pause harder in the flow pays for itself.
The first billing cycle decides everything
More subscribers are lost between the first and second delivery than at any other point. The sign-up worked, the product arrived, and then nothing happened until a card was charged again and the customer felt ambushed.
Three things fix most of it. Tell people clearly when the next charge lands and how to change it. Use the gap to teach them something about the product, so the second box arrives with more context than the first. Warn before renewal rather than after.
Failed payments belong in this bracket too. A meaningful share of what looks like churn is a card that expired, and a properly built dunning sequence recovers a good proportion of it. That is conversion work, even though it never touches a product page. Our approach to Klaviyo email marketing for subscription brands goes into the sequence in detail.
What to pull to validate this:
- Cohort retention from cycle one to two, and two to three, from Recharge or Skio. The drop between one and two is the number to fix first, and it is usually the largest single leak in the programme.
- Failed payment rate and dunning recovery rate. Involuntary churn hides inside the headline churn figure and needs separating out before you draw any conclusions about the product.
- Median days between sign-up and first cancellation. A cluster on the day of the second charge means the pre-renewal notification is missing or not landing.
- Klaviyo performance on the welcome and pre-renewal flows, and whether recipients of the pre-renewal email churn at a lower rate than those who did not open it.
How to run a subscription CRO programme
Start with the data you already have rather than a test plan. Pull subscription share of orders, average cycles completed, cancellation reasons, and the point in the lifecycle where subscribers actually leave. That last number tells you which section of this post to work on first.
Fix the portal and cancel flow before touching the buy box. Acquisition tests are more fun and get more attention, but pouring new subscribers into a programme that loses them in cycle two is expensive.
Then work the buy box, the post-purchase offer and the frequency defaults, in that order. Run one change at a time, and hold the measurement window open for three billing cycles even when the early numbers look good.
This is the way we approach ecommerce CRO for the brands we build subscription programmes for, and the reason we do not separate the two into different workstreams. You can see how it plays out in our work with Bold Bean Co and Origin Coffee.
Subscription CRO questions we get asked
What is a good subscription conversion rate?
There is no single benchmark, because it depends heavily on category and consumption rate. Subscription share of orders is the more useful measure than a page-level conversion rate. Compare it against your own baseline over three billing cycles rather than an industry average, since replenishment categories like coffee behave nothing like quarterly beauty refills.
How is a subscription CRO audit different from a standard CRO audit?
A standard audit follows a session and ends at the order confirmation page. A subscription audit follows a subscriber across billing cycles and includes the portal, the cancel flow and the dunning sequence. It also uses different data, pulling cohort retention and cancellation reasons from Recharge or Skio alongside the usual GA4 and Shopify numbers.
Should the subscription option be selected by default?
In most replenishment categories, yes. Pre-selecting subscribe puts the effort on the shopper who wants a one-off, and those categories are bought repeatedly anyway. High price points and first-time flavour-led purchases are the exception. Whichever you choose, keep the one-time option clearly visible, because hiding it drives cancellations in the first cycle.
What discount should you offer on subscriptions?
Below 10% rarely changes behaviour. Above 20% starts to devalue the full price and makes future increases harder. The range between the two works for most DTC brands. Non-price benefits often perform better anyway: free delivery on subscription orders, early access to launches, or a member-only size protect margin while still giving a reason to commit.
How do you convert one-time buyers into subscribers?
Two moments work best. The first is immediately after checkout, while the customer has just committed and their payment details are already stored. The second is when their first order is nearly used up, triggered by real consumption data rather than a fixed delay. Both convert better than a product page widget because trust is already established.
Does a cancel flow reduce subscription churn?
A well-built one does, provided it answers the reason given rather than offering a blanket discount. Too much product needs a delay or longer interval, the wrong item needs a swap, and a holiday needs a pause with a set return date. Leading with money off teaches subscribers that cancelling is how they get a better price.
Is build-a-box worth building?
It works when the catalogue is broad enough for the choice to feel real, usually somewhere above eight or ten SKUs, and when a mixed box makes sense at your price points. Below that, a curated bundle converts better. Where it fits, build-a-box removes the two biggest objections to subscribing: boredom and being stuck with the wrong product.
The brands with the strongest subscription programmes are not the ones with the best app. They are the ones who stopped treating subscription as a checkbox on the product page and started treating every touchpoint from buy box to cancel flow as one conversion problem. If your subscription share has been flat for a year, the fix is almost never a different platform. It is the twenty small decisions nobody owns, and the reports nobody opens.