When Bundles Help Margin — and When They Just Cannibalize It

When Bundles Help Margin — and When They Just Cannibalize It

Bundles raise average order value — case studies across the industry report lifts in the 20–35% range. Which is exactly why you should distrust the AOV number. A bundle can raise AOV and destroy contribution at once, when it discounts two products the shopper would have bought anyway.

The two buyers a bundle creates

Any bundle splits its buyers into:

• Cannibalized: would have bought both items at full price anyway. The bundle just gave them a discount on a basket they were already building.
• Expanded: would have bought only one (or neither); the bundle pulled the second item in.

The bundle pays for itself only if the expanded buyers cover the margin lost on the cannibalized ones.

The break-even, worked out

Two products: A at $40, B at $30, 50% margin each.

• Separate basket (A+B): $70 → contribution $35.
• Bundle at 15% off: $59.50 → contribution $29.75.

Cannibalized buyer (would've bought A+B): contribution falls $35 → $29.75 = loss of $5.25.

Expanded buyer (would've bought only A):

• Without the bundle: buys A → $20 contribution.
• With the bundle: buys both → $29.75 contribution.
• Gain = $9.75.

Break-even: expanded × $9.75 > cannibalized × $5.25 → expanded ≥ 54% of cannibalized.

So at these numbers, the bundle wins if even a bit more than half of bundle buyers are being expanded. That's a lower bar than most merchants assume — and the reason bundles often do work. But it flips fast: raise the bundle discount to 25% and the cannibalized loss grows while the expanded gain shrinks, pushing the required ratio far higher.

Most stores never estimate the ratio. They see AOV rise and ship it.

How to actually tell

Holdout, as always — but the right one:

• Randomize who sees the bundle (not who buys it). Control shoppers see the two products separately.
• Compare contribution per eligible session, not AOV among buyers.
• Track mix: did the bundle pull revenue from high-margin singles, and did it change return rates?

Acceptor metrics ("bundle buyers spent $86 vs. $55!") are selection bias — of course they spent more to accept.

When bundles help

• Anchor + filler: a high-elasticity hero pulls the purchase; a high-margin, low-elasticity companion rides along. Discount lives on the anchor; margin lives on the filler.
• Discovery: a bundle that introduces a new SKU creates demand that didn't exist — sampling with a price.
• Threshold play: a bundle that pushes a basket over the free-shipping line earns the shipping economics too.

When they hurt

• Two heroes, one discount: bundling the two products your best buyers already bought together. Pure cannibalization with a UI.
• Discount without relevance: unrelated items where the only selling point is the percentage. No discovery, no behavior change.
• Split shipments: if the bundle ships as two parcels, the extra fulfillment eats the upside.

The inventory trap

Bundles that pair a fast mover with dead stock feel clever. But discounting the fast mover to move the slow one is usually the most expensive clearance channel you have. Price the bundle against what the slow SKU would clear for on its own — and check whether a threshold-gated gift wouldn't do it cheaper.

The takeaway

A bundle is a discount wearing a merchandising costume. It deserves the same test every discount deserves.

Hold out the bundle. Measure contribution per eligible session. And ship it only when the expanded buyers clearly outnumber the ones you just discounted for no reason.