KEY TAKEAWAYS:
Understand why cannibalization hides inside campaigns that report strong returns.
Learn why last-click measurement credits the effort that costs you margin.
Identify the four places independent retailers compete against themselves.
Apply three tests that prove whether your revenue is incremental.
Recently, I stopped into Macy’s to purchase something I wanted. I had already decided to buy the item, and it turned out that only Macy’s had what I was looking for. During the checkout process, the person said, “I’ve applied a 15% promotion we’re running to your purchase.” I was pleased because I saved money, but it has not made me a more frequent Macy’s shopper.
Their system records the sale as coupon-driven. The promotion looks successful. And they just paid 15% to discount a purchase that was going to happen anyway. That is cannibalization.
Last week, we examined fragmentation: marketing efforts spread so thin across so many channels that none of them ever reach full effectiveness. This time, the focus is on a different issue: where your channels are strong enough to perform, but two of them are competing for the same sale.
Your Best Performing Campaign Might Be Your Most Expensive One
Return on ad spend (ROAS) measures credit, not cause. It tells you which effort was standing closest to the sale when it happened. It does not tell you whether the sale needed that effort at all.
These are different questions, and only one affects your bank account. A campaign might show a good return but still be purely a cost, because the revenue it reports was already yours. The real question is: would this sale have happened anyway?
The reason it hasn’t been asked is that your tools don’t provide an answer. They are designed to allocate credit among visible touchpoints, not to identify the cause-and-effect relationship. As a result, they respond to a question you didn’t ask, and they do so with certainty.
Why Your Measurement Points at the Wrong Channel
This is how the mechanism works.
Last-click attribution gives 100% of the credit to the final interaction before a purchase. That final interaction is often the cannibalizing one, because the efforts closest to the finish line are the ones that catch customers who were already on their way.
The retargeting ad shown to a customer holding a full cart. The branded search ad served to someone who typed your store name. The coupon on the sidewalk. Each of these appears at the end of a journey it did not create, takes the credit, and reports a strong return.
Then the problem compounds. Strong reported performance justifies more budget. More budget goes to the effort that is capturing existing demand rather than creating new demand. Your measurement system is not merely failing to catch the problem. It is actively recommending that you make it worse.
The industry knows this. According to an EMARKETER survey conducted with Snap, only 21.5% of marketers believe last-click measurement reasonably reflects a platform’s long-term impact on their business, and 74.5% are moving away from it or want to. Those respondents were large advertisers, not independent retailers, but the acknowledgment is industry-wide, and the mechanism is identical at any size.
4 Areas Retailers Compete Against Themselves
1. Paid Media
You already rank first organically for your store name, and you are also bidding on it. Automated campaign types compete with your manual campaigns for the same inventory. Broad match can pull one ad group into another’s territory. In each case, you are running an auction against yourself and paying the winner.
2. Promotional
A sitewide sale discounts inventory that was already selling at full price. Loyalty rewards stack on top of a promotion, doubling the discount for a customer who needed neither. Run promotions too often, and you train your best customers to wait, which permanently cannibalizes full-price demand.
3. Channel
Your store and your website compete for the same local customer, particularly when one carries free shipping or a code the other will not honor. Marketplace listings undercut your own site after fees are counted, so you win the sale and lose the margin.
4. Assortment
A new budget-friendly line causes current customers to switch to a lower tier rather than attract new ones. The line remains profitable. Units sold increase. Meanwhile, the average transaction value gradually decreases.
3 Tests That Prove Whether Revenue Is Incremental
Part 1 measured thresholds. These tests measure cause.
Test One: The Pause Test
Disable one effort for a specified period (minimum two weeks, depending on your business traffic and purchase frequency) and focus on the total revenue instead of that channel’s revenue. Begin with your highest-spend suspect.
Healthy result: Total revenue drops measurably when the effort stops. The spend was buying something.
Problem result: The channel’s numbers decrease, and your overall total remains almost unchanged. The demand has been directed elsewhere. This is the most valuable test discussed in this article, and 52% of US brand and agency marketers now employ some form of incrementality testing because dashboards cannot provide an answer to this question.
Test Two: The Branded Search Test
Look at how much you spend on ads triggered by your own store name, then check whether you already rank first organically for that term and on the map.
Healthy result: You are defending against a competitor bidding on your name, or you do not yet rank well in organic search.
Problem result: You rank first organically, but you still pay for clicks. Stella’s benchmarks, based on 225 controlled tests from August 2024 to December 2025, showed that branded search had a median incremental return of 0.70x—lower than other channels and below breakeven. Overall, the difference between platform-reported return and actual incremental return can be two to three times greater, with branded search and retargeting exhibiting the largest discrepancies.
Test Three: The Tier Migration Test
Pull your average transaction value and unit count for the last twelve months, then compare them to the twelve months before your newest lower-priced line launched.
Healthy result: Units rose, average transaction value held, and a meaningful share of buyers on the new line are first-time customers.
Problem result: Units rose, average transaction value fell, and buyers were mostly people who used to purchase at a higher price point. You did not expand your customer base. You moved it down.
Cannibalize Yourself Before Someone Else Does
Not all cannibalization is a mistake. Some of it is strategy.
Planned cannibalization occurs when you intentionally select the timing, retain the customer, and accept a different margin for that transaction in exchange for maintaining the relationship. Accidental cannibalization happens when it occurs unexpectedly, and you’ve essentially paid the price for it. The key difference is about control, not the result.
A clear example in jewelry today is lab-grown diamonds. When a retailer offers lab-grown options, they are likely to attract customers who might have purchased natural diamonds instead, but opt for the lower-priced lab-grown versions. This is essentially cannibalization, with an important detail: lab-grown diamonds usually have higher profit margins than natural ones. Therefore, even if the total sales value decreases, the profit on each sale can increase. Looking only at the average transaction value might suggest a loss, but the decision could still be right, especially if the alternative is losing the customer entirely to another seller. The key difference lies in whether the retailer made this choice intentionally, understanding its impact on sales and profit margins, and whether they analyze tier migration results rather than assuming unit growth signals overall expansion.
“A campaign that reports a strong return and a campaign that grew your business are not the same thing, and your dashboard cannot tell them apart.”
– Jennifer Shaheen
President and Founder, Technology Therapy® Group
The Business Case, and Where to Start
Fragmentation costs you the money that produces nothing. Cannibalization costs you the money that produces what you already had.
The promotional side of this is measurable at scale. Klaviyo’s analysis of thousands of brands found that those with low discount rates grew GMV 12% year over year from Q1 2025 to Q1 2026, while deep discounters grew 6%. The margin picture is wider still: low discounters posted 8% margin growth, while brands running eleven or more promotional events a year saw profit margins fall 11%. That is a nineteen-point margin gap between businesses selling similar products to similar customers. The only real difference is how often things went on sale.
The gap exists because of the reason this entire article is written: each quarter reports a successful sale on the dashboard, but the books don’t match annually.
Begin by conducting the pause test on your top-spending effort, allowing it two full weeks. Compare your branded search spend to your organic ranking, which takes about ten minutes. Next, review your average transaction value alongside your unit counts to determine if growth results from expansion or migration. If a test shows that revenue was not incremental, reallocate that budget to channels that generate demand instead of merely capturing it. Finally, evaluate the overall impact, not just individual line items, as this is the same principle that led to issues in Part 1 and explains their continued invisibility.
Ready to Find Out What Your Marketing Is Actually Buying?
The Clarity Program gives you senior-level oversight of where your budget goes and what it returns, so you can tell the difference between revenue you earned and revenue you paid twice for.
