Is your bundle eating your hero?

CMS display title: Is your bundle eating your hero?

Subtitle (deck):

Eight weeks after launch, your hero SKU is down 14% month over month, and four stakeholders are asking the same question in different words: is this bundle cannibalization, and should we keep the bundle? Here is a DTC bundle analysis framework that answers with contribution margin, not opinions.

Key takeaways

• Bundle cannibalization is not yes or no. Split revenue into substitution (hero buyers who switched), incrementality (net-new customers), and expansion (existing customers who would not have bought the hero in the same window).
• Track contribution margin, not unit counts alone. Bundles often look worse on units and better on dollars because bundle AOV is higher than the hero alone.
• Run three calculations in order: counterfactual hero sales → cohort decomposition → contribution margin sum.
• Flag five confounders (promo, channel mix, seasonality, stockouts, audience drift) before you present a single number to the CMO.

Is Your Bundle Cannibalizing Your Hero SKU?

Most DTC brands launch a bundle SKU at some point in their first three years, but rarely do they know whether the bundle is actually making money.

Now imagine you are a DTC data analyst at one of these brands and the bundle launched eight weeks ago. The hero SKU units are down 14% month over month and the brand manager wants to know if the bundle is the cause. The CMO has a hunch. The CFO is asking about contribution margin. The growth lead wants to know whether to keep running the Meta creative that has been promoting the bundle.

All four are asking the same question in different clothes: is the bundle cannibalizing the hero, and should we keep it?

This is the question that almost never gets answered in DTC at the depth it deserves. The hunches in the room are wrong about half the time, and somebody is going to have to make the actual call.

So let's break it down.

Why "Is the Bundle Cannibalizing the Hero?" Is the Wrong Question

Product bundle cannibalization is not a yes or no question. Treating it as such means losing the argument to whichever stakeholder has the strongest opinion in the room.

The real question has three parts, and that shape is what separates a hand wave at the QBR from a real bundle cannibalization analysis.

The Questions Every DTC Bundle Cannibalization Analysis Must Answer

First: How much of the bundle's revenue is coming at the expense of the hero SKU, how much is from customers who would not have purchased the hero anyway, and how much is from existing hero buyers who now spend more per transaction than they used to?

Second: What is the net contribution margin impact across the affected SKUs, not the unit count impact? Bundles almost always show worse looking unit cannibalization numbers and better looking margin numbers because bundle AOV is higher than the hero alone. Looking at units is what makes brand managers nervous about bundles that are actually working.

Third: What is the customer lifecycle impact? A bundle that pulls a customer into the brand earlier or moves them into a higher frequency consumption pattern creates value that will not show up in the first eight weeks of sales data. A bundle that becomes a one-and-done purchase replacing what would have been a long running subscription does the opposite.

These are the questions that lead to a data backed discussion instead of everyone defending their hunches.

Bundle cannibalization for subscription vs one-time purchase brands

For subscription brands, anchor on lifetime contribution margin per subscriber, not first purchase margin alone. A bundle that lifts AOV at the cost of churning the subscription cohort is worse than the same trade in a one time purchase brand. Run the same three calculations, but weight retention and subscriber base impact heavily.

What Data to Pull: 3 Slices for Hero SKU vs Bundle Analysis

If you want to track bundle performance on Shopify (or any commerce stack) with confidence, start with three clean data slices, not a single blended revenue line.

Slice 1 | Hero SKU sales trend

Six months pre-launch and eight weeks post-launch, normalized for total brand traffic and channel mix changes during the same window. Run this for the cohort of customers who would plausibly have bought the hero. If the hero is a subscription product, run it against the subscription base, not against new acquisitions alone.

Slice 2 | Bundle buyer audit

Of customers who bought the bundle: what percentage purchased the hero SKU in the prior 12 months? What percentage are net new to the brand? What percentage are existing customers who had not yet purchased the hero specifically? The first group is substitution risk. The second is pure incrementality. The third is expansion, often the most interesting case, because they may have been on a path to the hero anyway and the bundle pulled the purchase forward.

Slice 3 | Contribution margin per customer

The bundle's gross margin percentage is almost certainly lower than the hero's, because the second item in the bundle is usually a thinner margin companion product. Do the math on absolute dollars of contribution, not on margin rate. A bundle that compresses gross margin rate by six points while expanding AOV by 35% can still be a meaningfully positive trade.

5 Confounders That Skew Bundle Performance

Five things can mislead your Shopify bundle analytics read if you do not account for them:

1. Promo | If the bundle launched during a promo, promo lift confounds the results. Remove the promo period and re-run.
2. Channel mix shifts | A 30% increase in Meta spend during launch brings a different audience with different bundle vs hero preferences.
3. Seasonality | Hero SKUs often have natural peaks that do not match the bundle launch month. Use year over year comps unless the brand is in year one.
4. Inventory | If the hero was out of stock post-launch, some bundle buyers defaulted from a missing hero; penetration overstates cannibalization.
5. Audience drift on creative | New TikTok creative may pull a younger audience that simply prefers bundles. The read is then about audience, not SKU lineup. These belong in the footnotes of your final answer, not swept under the rug to tell an easier story.

Bundle Cannibalization Math: Counterfactual, Cohort Split, Margin Sum

Three calculations, done in sequence:

1. The counterfactual (maximum possible cannibalization) Project hero SKU sales using the prior 12 months of trend, adjusted for total brand growth rate and for the confounders above. Compare to actual hero SKU sales during the eight week post-launch window. The gap is the maximum possible cannibalization. Note "maximum possible" carefully, it is not the final answer yet. This is your counterfactual baseline.

2. The cohort decomposition (bundle cannibalization rate by bucket) Of the bundle's revenue, allocate each dollar to substitution, incrementality, or expansion. Substitution is the existing customer who bought the hero in the prior year and bought the bundle in place of the next hero refill. Incrementality is the new customer with no prior brand purchase. Expansion is the existing customer with prior brand purchase but no prior hero purchase. Use a 12-month attribution window for prior purchase, not the default 30 or 90 days.

3. The contribution margin sum Total dollar contribution from hero plus bundle in the post-launch window, compared to counterfactual contribution from hero alone. A positive number is net-positive cannibalization. A negative number means the bundle is losing the brand money even after accounting for new bundle revenue.

Done in that order, the three numbers tell a single story, and the CMO can challenge any of them on their own terms.

What a Good Bundle Analysis Looks Like (One Paragraph, Three Numbers)

The deliverable to the brand manager and the CMO is one paragraph, three numbers, and a footnotes section.

Paragraph:

"The bundle is contributing +$X in net contribution margin per month over the counterfactual where the bundle did not launch. Of bundle revenue, roughly 18% is substituting hero purchases, 32% is incremental net-new customer revenue, and 50% is expansion from existing customers who would not otherwise have bought the hero in the same window. Recommend keeping the bundle and watching for diminishing returns at month four."

Three numbers: substitution share, incrementality share, contribution margin delta.

Footnotes: the confounders, the assumptions, the windows used, and an honest read on which numbers are tighter than others. The substitution share is usually the noisiest. The contribution margin delta is usually the most robust.

That is the answer. It takes a focused analyst about a day and a half to assemble, once the data is reconciled and ready to query.

Why DTC Brand Managers Rarely Get This Answer

Most brand managers never get this answer at this depth. They get a hand wave at the QBR. They get "yes the bundle is fine" or "yes the bundle is hurting" depending on who has the strongest opinion in the room that day.

The actual math gets done at few DTC brands at this revenue scale, and gets done well at even fewer.

That gap is not a sophistication gap. The math is not difficult, it is a time gap. The analyst who could answer this question well is buried in reporting cycles and reconciliation work, and the deep questions get pushed to next month, every month. When the time opens up, this is the kind of work that quietly decides whether a brand's product line is set up right for the next 18 months.

Frequently asked questions about bundle cannibalization

Q1. How much post-launch data do you need for bundle cannibalization analysis?

A: Ideally eight weeks of post-launch data with a clean (non-promo, non-stockout) period of at least four weeks inside it. With fewer than six weeks, confounders eat the signal, and with 12 weeks or more the cohort decomposition becomes meaningfully more reliable. Flag anything under four weeks as directional only.

Q2. What if the bundle launched alongside a price change on the hero SKU?

A: You cannot isolate the bundle effect cleanly in that window. Build two counterfactuals, one with the price change alone and one with the bundle alone, and present both. State plainly that the two effects are confounded, and recommend re-running after a clean comparable period exists.

Q3. Can you run bundle cannibalization analysis without customer level cohort tracking?

A: Partially. The counterfactual and contribution margin calculations only need SKU level data. The substitution vs incrementality breakdown needs customer level repeat purchase history. Without it, you can still answer the contribution margin question with confidence, but label the bucket split as an estimate in the footnotes.

Q4. Is bundle cannibalization analysis different for subscription brands?

A: Yes. For subscription brands, the most important number is impact on the subscription base and on lifetime contribution margin per subscriber. A bundle that lifts AOV at the cost of churning subscribers is worse than the same trade in a one-time purchase brand. Anchor on lifetime contribution margin, not first-purchase margin alone.

Q5. What if the bundle was designed primarily as a new customer acquisition tool?

A: Weight the incrementality shares heavily and compare second purchase rate of bundle first buyers vs hero first buyers. If bundle first customers convert into hero buying repeat customers within ~10% of hero first customers, the bundle is doing its job even at a thin first-purchase margin.

Q6. How often should you re-run a DTC bundle cannibalization analysis?

A: At launch, every four weeks for the first quarter. After 90 days, quarterly. If the bundle is featured in a major promotion or new channel, refresh after that period closes. The cannibalization profile often shifts as the audience discovering the bundle changes.