Bundle Discount Margin: The Math Before You Run

Your bundle sold three hundred times last quarter. Revenue is up. Nobody in the business can tell you whether the store is better off, and the report that would settle it does not exist in your admin.

That is the normal state of bundle discount margin. The offer moves the number everyone watches, and quietly moves two nobody does: what each sale contributes, and how many of those sales you already had.

What follows is the arithmetic, in the order you need it. None of it takes a data team. All of it takes giving up percentages of price as the unit you think in.

What does a bundle discount actually cost you?

The headline discount is a share of the price. The cost is a share of your margin, and those two figures are never close.

Nothing on the cost side moves when you drop the price. The product lands at the same figure and the parcel weighs the same. Every unit you knock off comes out of contribution, whole.

Contribution margin per unit is the price minus everything that varies with the sale. Most stores get it wrong by leaving rows out.

That gap is set by your margin, so it differs in every catalogue. Substitute your own costs before anyone argues about the headline.

Which bundle sales were incremental and which were cannibalised?

A bundle sale is not automatically new revenue. Some share of the people who take it were buying both items regardless, and for them the offer is a gift with no change in behaviour attached.

Every taker arrives from one of four counterfactuals.

Rows one and two are the whole argument. Row three is real but uncommon, because a bundle rarely converts a shopper the product page could not, and row four is rarer still.

What separates them is your baseline attachment rate: the share of hero orders that already contained the partner item before any bundle existed. Measure it first, because after launch the offer has contaminated the behaviour you wanted to observe.

Assuming every baseline attacher takes the bundle is pessimistic, and close to true. A floor you can defend beats a forecast you cannot.

How do you set a margin floor before you pick a bundle price?

The floor is the bundle price at which blended contribution per order holds flat under a pessimistic view of how many takers are new.

• Stress-test the incremental share rather than forecasting it. Halve whatever you believe, then price so the bundle still clears.
• Never let a bundle contribute less than the hero sold alone. If it does, you are paying shoppers to take more of your inventory off you.
• Price the parcel, not the products. A second unit that tips the order into the next carrier weight band can cost more than the discount you spent a week debating.
• Reserve for returns before setting the discount. Bundles come back as parts, and the part that comes back is often the one that never sold alone.
• Write the floor down. One that lives in somebody's head gets renegotiated during a slow week, which is precisely when it should not be.

The pillar guide to product bundles covers which shape to build. This is the number that decides whether you can afford it.

What is the break-even attachment rate on a bundle discount?

You have both inputs already: the gain on an incremental sale and the loss on a cannibalised one. Break-even is the share of takers who must be new for those two to cancel out. Divide the loss by the sum of the loss and the gain.

Discount depth and risk do not trade off in a straight line. A deeper discount raises the cost of every cannibalised sale while shrinking the reward on every incremental one, so both halves of the ratio move against you at once.

So compute the break-even share at two or three candidate prices before committing to one. If the answer at your preferred discount sits above what your baseline leaves room for, the price is wrong, not the idea.

How do you measure blended margin per order without fooling yourself?

Blended contribution per order is the scorecard, but only if the denominator is honest. Count every order that could have carried the bundle, not only the ones that did.

1. Fix the baseline before launch. Record hero order count, attachment rate and contribution per order for the four weeks prior.
2. Compute contribution per unit from the full cost list, including the rows most stores leave out of the table above.
3. Tag bundle orders at line level, so a bundled unit is distinguishable from an ordinary one in your export.
4. Wait a full purchase cycle, and at least thirty days. A bundle only touches shoppers who reach one product, so a week is noise.
5. Recount the residual attachment rate on orders that did not take the bundle. If it collapsed, your cannibalisation estimate was right. If it held up, more takers were incremental than you assumed.
6. Divide total contribution by every eligible order, before and after, and compare.
7. Check order count moved with it. Contribution per order can rise because your cheapest orders stopped happening, which is a conversion problem, not a win.

What quietly breaks the arithmetic after launch?

The sums above are correct on the day you do them. Several inputs move afterwards, and none announce themselves.

• A site-wide code stacks on top. What you modelled becomes a deeper discount on the highest-volume weekend of the year, and the break-even share moves with it.
• A component goes on sale. Discount one item hard and the bundle's implied saving evaporates, so shoppers take the cheaper route and you keep the cost without the benefit.
• The parcel gets heavier. A supplier change to packaging can push a bundle over a weight step months after you priced it.
• Landed cost drifts. Freight, duty and currency all move. A bundle priced against last year's landed cost is thinner than the sheet claims.
• Returns arrive late. Month one looks profitable partly because month one's returns land in month two.

Re-run the contribution sum quarterly, and again before any promotional period where discounts could combine.

Does the same math work for mix and match and volume tiers?

The framework does. Two of the inputs change.

For mix and match, contribution varies by combination, so the floor has to hold for the cheapest permissible basket rather than the average one. Shoppers find the best-value combination faster than anyone expects, and it is the only one worth pricing against.

For a volume tier the counterfactual differs in kind. A cannibalised bundle sale costs you the discount once. A cannibalised quantity sale costs the discount and delays the next order, so the loss lands across two months and the second half is easy to miss. Our notes on setting up volume discounts cover the mechanics; the arithmetic here still applies, with the reorder gap added to the loss side.

The constant across all three is the counterfactual. You are never comparing the bundle against zero, but against what those shoppers would have done without it, which is the comparison no dashboard makes for you.

Where does Edge Bundles fit once the numbers are done?

None of the above needs an app. It needs an order export, your real landed costs, and an hour with a spreadsheet.

What an app can do is stop the arithmetic drifting after launch. Edge Bundles is our own app and it is new, so read this as a description rather than a claim about results. It draws stock down to component SKUs, so bundled and unbundled sales of the same item stay in one ledger and the export in step three is readable. It resolves which single offer applies to a line, so a stacked code cannot quietly move the break-even you calculated. There is a free plan.

The judgement calls stay yours. No app knows your landed cost, and none of them can tell you what a shopper would have done instead.