Discounting Strategy
Date: 2026-08-16
Deciding when and how much to discount, rather than reaching for it whenever a number is soft. The arithmetic is brutal and the behavioural cost is worse — customers learn, and a trained buyer never pays full price again.
What it is
Discounting strategy is the deliberate framework governing depth, frequency, targeting and duration of price reductions.
Most businesses don’t have one. They have a habit, which is different, and it compounds.
The two costs
Immediate: margin. A 10% discount removes a third of contribution and needs a 49% volume lift to break even — Discount Impact on Margin.
Cumulative: expectation. Customers learn your pattern. Once they know a sale is coming, full-price demand converts into discounted demand permanently. The margin doesn’t come back when the promotion ends, because the behaviour has changed.
In plain terms: the first discount costs margin on that sale. The tenth discount costs margin on every future sale, because nobody buys at full price any more.
That second cost never appears in a campaign report, which is why discounting looks better than it is every single time.
The hierarchy
Ordered from least to most damaging:
| Approach | Cannibalisation | Notes |
|---|---|---|
| Targeted at lapsed customers | None — they weren’t buying | The clear best case — Winback Campaigns |
| First-order only | None on existing customers | But raises CAC — paid to every new customer, incremental or not — and lengthens payback — Payback Period |
| Threshold-based | Low — requires incremental spend | ”Spend £60, save £10” — Shipping Thresholds |
| Clearance on dead stock | Low | The alternative is a write-off |
| Category promotion | Moderate | Shifts mix, sometimes deliberately |
| Site-wide, announced | High | Most demand would have converted anyway |
| Predictable calendar sale | Highest | Trains customers to wait — Promotional Cadence |
The rule underneath the table: discount narrowly, to people who weren’t going to buy. Every widening of the audience adds cannibalisation and subtracts contribution.
Before running one
Four questions, and the first is normally skipped:
- What volume uplift breaks even? Compute it. Above +50% and it’s very unlikely to pay
- How much is incremental? Hold out a random slice and find out — Incrementality Testing
- What does it do to mix? Discounting a low-margin category is worse than the headline — Basket Composition
- What does it teach? If this is the third quarter running, you’re building an expectation
Alternatives that don’t train the behaviour
- Added value rather than price off. Free delivery, a sample, extended returns. Costs less than the equivalent discount and doesn’t reset the reference price
- Bundles priced above the sum of individual contribution — Bundling
- Member pricing, which requires an exchange rather than being universal
- Early access, which costs nothing at all and works on your best customers — where a discount would be pure margin loss
- Fixing the actual problem. Persistent discounting to hit a number is usually compensating for a demand, range or pricing problem, and the discount postpones the diagnosis
Where CRO is complicit
Worth naming, because it’s the connection most often missed. Conversion optimisation reaches for discounts because they reliably lift conversion rate — the metric being optimised.
A test that lifts conversion 8% with a 10% discount has lost money, and a conversion-rate-only readout will call it a win. Which is why contribution per session belongs as a guardrail on any test touching price or promotion — Guardrail Metrics, Contribution Margin.