How deep should a save discount go: the margin math#

Acceptance climbs with depth, then bends. In Churnkey's 2025 benchmark data, a 25% discount was accepted about 3.3% of the time, 40% off about 8.7%, and 50% off about 7.0% - deeper stopped helping before you reach half price. Meanwhile ProfitWell's research puts discounted customers at more than double the churn of full-price customers, with roughly 32% lower LTV. For most stores the workable band is 20-30% off for 2-4 cycles, and 50% off loses actual cash on any product whose costs exceed half its price.

This post assumes you have already decided a discount is the right offer for the cancel reason in front of you. If you are still deciding between offer types, that is a different question with different data, covered in pause vs discount. Here the question is narrower: the discount is chosen, so how deep, and for how long?

What the acceptance data shows#

The only public depth-vs-acceptance curve comes from Churnkey's voluntary churn benchmarks (published late 2025, built on 2 million cancellation survey responses across 5 million-plus sessions). Their acceptance rates by discount size:

Two caveats before you build on this. First, Churnkey does not fully label the chart's denominator, so read these as relative, not as your expected save rate. Second, their own conclusion is that "there is no right answer" on depth. But the shape is the finding: acceptance roughly doubles between 25% and 40%, then falls at 50%. Half off is not more persuasive than 40% off. Past a threshold, a deeper discount reads as desperation, or as confirmation that the full price was inflated all along.

For overall context, Chargebee's Q1 2024 cancel-flow data found discounts were the most effective save offer they track, at 17% acceptance overall and 21% in B2C, and found pricing was the top cancellation reason at 31% of exits. Discounts are the workhorse offer. That is exactly why depth mistakes are expensive: you make them at volume.

The margin math#

A discount is only as affordable as your contribution margin. The rule is one line: the moment discount depth exceeds gross margin, every retained cycle costs you cash. Not opportunity cost. Cash.

Take a $49/mo subscription in two archetypes. A digital product (SaaS, membership, content) with roughly 10% incremental cost per cycle, and a subscription box with $29.40 of product, packaging, and shipping per cycle - 60% of price, a normal number for physical goods.

The digital column degrades gracefully: even at 50% off, each retained cycle contributes $19.60. The box column hits zero at exactly 40% - the depth where Churnkey's acceptance curve peaks, uncomfortably - and goes negative at 50%. A box merchant running 50% off for 3 cycles pays $14.70 for the privilege of shipping three more boxes to someone who already tried to leave.

The break-even depth is just your gross margin percentage. Costs at 60% of price mean 40% is your ceiling, and your workable depth is well below the ceiling, because a save at zero margin is revenue theater: the P&L keeps the subscriber and none of the money.

The save is not the end of the story#

The depth decision looks different once you follow the cohort past the save. Two verified findings, and they point the same direction.

First, ProfitWell's discounting research (published via Paddle) compared companies using minimal discounts against aggressive discounters and found the discounted cohort's LTV came in upwards of 32.41% lower. Their related analysis found discounted customers churn at more than double the rate of full-price customers and value the product at least 12% below list price. Discounts recruit and retain the least committed slice of the base, and depth amplifies the selection.

Second, Churnkey's own discounting analysis found cancel-flow discount accepters stay 5.1 months longer on average, but only 11% remain subscribed a year later. A save discount mostly defers churn rather than reversing it. That is still often worth doing - 5 extra months of positive margin is real money - but it caps what a save is worth, and therefore what it is rational to spend on one.

Put the two together: the customer you save with 50% off is, on the data, likelier to be gone within a year than the one you save with 25% off, and pays half as much while they stay. Depth buys acceptance from exactly the customers least likely to make the discount back.

Duration is the other half of depth#

A 40% discount for 2 cycles costs less than a 20% discount for 6 cycles on the same price. Depth gets all the attention; duration moves the same money quietly. Honest note on the data: we found no public benchmark for how many cycles save discounts should run. Vendor guidance clusters around a few months, but nobody has published acceptance or retention by duration. What we can say from arithmetic and the deferral data:

• Time-box every save discount, and make reversion automatic. Per Churnkey, the median discount save is gone in a few months anyway. A discount that outlives the median save lifetime is depth you gave away for nothing.
• 2-4 cycles is the defensible default. Long enough to change the renewal decision twice, short enough that most accepters see a full-price renewal again while still subscribed.
• A permanent discount is not a discount. It is a price cut for one customer, forever, granted under pressure. If the sustainable number really is 40% below list forever, what you have is a customer on the wrong plan - that is a downgrade wearing a discount costume, and the downgrade offer handles it with a durable plan instead of a leaky exception.

When 50% off is actually right#

Deep discounts are not always wrong; they are wrong by default and right in specific corners:

• High-margin digital products with real switching costs. At 90% margin, 50% off retains $19.60 of monthly contribution. If your data shows saved annual-plan customers renewing at full price later, a one-time deep discount can clear the bar. Check the claim against your own cohorts, not against hope.
• A single cycle, framed as a credit. One deeply discounted renewal ("your next month is half price") costs a bounded amount and avoids retraining the customer's reference price for months.
• Where the alternative is negative anyway. If a canceller triggers a refund, a chargeback risk, or an unrecoverable acquisition cost, one expensive cycle can be the cheaper exit path.

The disqualifier is unchanged: physical COGS. If your costs are 50% of price or more, 50% off is not a retention offer. It is a donation with shipping.

The decision rule#

Depth is a solved equation once you write your own numbers into it:

1. Ceiling: gross margin. Never discount past it. Costs at 55% of price mean 45% is a hard wall, and staying 10-15 points under the wall keeps saves cash-positive.
2. Default: 20-30% for 2-4 cycles. It sits on the rising part of Churnkey's acceptance curve, stays affordable in both archetype columns above, and matches the sizing bands in pause vs discount.
3. Escalate depth only with evidence, never past 40%. The public acceptance data shows nothing to buy between 40% and 50% - acceptance drops while cost rises.
4. Price the save before you ship it. Expected value of a save offer = acceptance rate x retained cycles x per-cycle margin at the discounted price. Run your real numbers through the LTV lift calculator to see what a given depth is worth across your whole cancel volume - it is usually the fastest way to discover that a 25% offer beats a 45% offer on cash even at half the acceptance.
5. Measure the cohort, not the save. Track 6-month and 12-month survival of discount accepters against non-discounted retainers. If your accepters look like Churnkey's 11%, the discount budget may be better spent one tier down, or on a different offer entirely.

The depth question feels like psychology and is mostly accounting. Acceptance curves bend, margins are hard walls, and the save that keeps the customer while losing the money is the one mistake this offer type lets you make at scale.