Channel cannibalization is one of those ecommerce problems that sounds obvious in hindsight and feels impossible in the weekly meeting. Amazon revenue is up. Shopify revenue is down. bol.com conversion improved, but the same SKU suddenly needs a discount to move direct. Google Shopping looks efficient, Amazon Ads looks efficient, and finance is quietly wondering why total contribution margin did not move.
The named mistake I see with multi-channel brand owners is treating every channel win as incremental. A team sees Amazon.de add €18,000 in monthly revenue after increasing Sponsored Products budget. Lovely. But Shopify branded search fell by €7,500, bol.com lost €4,200 on the same product family, and the blended return rate increased because the Amazon offer attracted more trial buyers. The channel dashboard says “growth”. The business dashboard says “we paid to move the same demand through a more expensive route”.
My stance: channel cannibalization is not a reason to avoid marketplaces. That is the lazy conclusion. Amazon, bol.com, Walmart, TikTok Shop and Mirakl retailers can all create real demand. The better question is: which part of channel growth is new demand, and which part is demand you already owned?
This guide is for brand owners selling across Amazon, bol.com, Shopify, Walmart, TikTok Shop, Kaufland or Mirakl retailers, typically from around €1.5K monthly ad spend or 1,000 orders per month. At that stage, cannibalization stops being a theory. It becomes a weekly allocation problem: where should stock go, where should ads scale, where should discounts be protected, and where is a marketplace simply taking a toll on an order that would have happened anyway?
What the current advice gets right
The research landscape is useful, but it often stops one level too early. DataHawk makes the strongest analytics argument: marketplace data is scattered across Amazon, Walmart, Shopify, ads and spreadsheets, so teams need unified dashboards, alerts and product-level profit overlays instead of debating exports. Their advertising analytics content also points out that ROAS and TACoS are hard to compare across channels unless attribution windows, currencies and cost inputs are normalised.
MerchantSpring explains the operational pain well. Marketplace metrics become decentralised as sellers add countries and platforms. Impressions, CTR, sales performance, seller health, inventory and profitability live in separate places, which turns analysis into a time sink. Their point is simple and correct: professional sellers need centralised marketplace reporting to make better decisions.
Jungle Scout, Helium 10, sellerboard and SellerApp all cover Amazon profitability from different angles. Jungle Scout focuses on sales analytics, profit overview, product-level PPC costs and P&L statements. Helium 10 positions profit tracking as a dashboard for revenue, net profit, refunds and product decisions. sellerboard is particularly strong on hidden Amazon costs such as returns, low-inventory fees, reimbursements and missed profit from stockouts. SellerApp covers Amazon analytics, PPC optimisation, keyword performance and SKU-level profit visibility.
The channel-cannibalization articles and videos add another useful layer. They warn that Amazon, Shopify, Google, Meta and marketplace ads can compete for the same customer. They talk about overlap, branded search, Amazon off-site placements, assortment differences and pricing architecture. One good point from the broader industry conversation is that pulling out of Amazon rarely fixes the problem. It usually just creates reseller risk and data loss.
So the foundation is there: unify the data, calculate profit properly, assign channels a role, and stop looking at revenue in isolation.
What most of the advice misses
Most articles still frame cannibalization as a marketing overlap problem: two channels targeted the same shopper, so spend was wasted. That is true, but incomplete.
The expensive version of cannibalization happens when a SKU’s operating constraints move with the order. The order does not only carry revenue. It carries fulfilment cost, marketplace commission, return probability, ad cost, cash timing, review impact, stock pressure and future customer access. If the same demand moves from Shopify to Amazon, the margin profile changes. If it moves from bol.com to a Mirakl retailer with stricter service levels, the operational risk changes. If it moves from Amazon.nl to Amazon.de, the VAT, fulfilment and return economics may change again.
That is why a channel-cannibalization dashboard needs to answer three questions, not one:
- Demand: did total product-family orders grow, or did the channel merely capture existing demand?
- Economics: did retained contribution margin improve after ads, fees, fulfilment, returns and discounts?
- Capacity: did the channel consume stock, cash or operational attention that would have earned more elsewhere?
If your dashboard cannot answer all three, it will over-reward the channel with the loudest attribution model. Tiny violin for the spreadsheet, but that is how profit leaks are born.
The cannibalization score: a practical model
Start with a product family, not a channel. A shopper does not think in SKUs, ASINs and EANs. They think “the blue lunchbox”, “the magnesium gummies”, “the cordless mini vacuum”. Your analytics should group the marketplace listings, Shopify variant, bundle and country versions into one commercial family.
Then score every SKU-channel combination on four signals.
1. Baseline demand
Build a four-to-eight-week baseline for each product family before the channel change. Use orders, revenue and contribution margin, not only sessions or clicks. The baseline should include seasonality if you have it. A sunscreen brand in June needs a different baseline from a kitchenware brand in February.
2. Transfer rate
Measure whether growth in one channel coincides with decline in another for the same product family. This is not perfect attribution. It is an operator signal. If Amazon gains 300 units and Shopify loses 220 units while total family demand barely changes, assume some transfer until proven otherwise.
3. Margin delta
Compare contribution margin per order by channel. Include marketplace commission, fulfilment, payment costs, return cost, ad cost, discounts and any channel-specific service fees. A €40 order on Shopify and a €40 order on Amazon are not the same animal. One may leave €17.20 contribution margin; the other may leave €10.80.
4. Capacity cost
Give stock and cash a score. If a channel consumes the last 600 units of a product that would have sold profitably on another marketplace, the channel’s reported ROAS is not enough. It used scarce inventory. Scarce inventory deserves a higher decision standard.
FiveX is useful here because the platform connects marketplace sales, advertising, product profitability and inventory into one view. Instead of checking Amazon Ads, bol reports, Shopify exports and stock sheets separately, you can see whether the SKU-channel combination has commercial permission to scale today.
Scenario 1: North Sea Hydration and the Amazon “win”
Imagine North Sea Hydration sells electrolyte sticks through Shopify, Amazon.de and bol.com. In May, the team increases Amazon Sponsored Products from €2,500 to €5,000 because branded and category campaigns show a 4.1 ROAS. Amazon monthly revenue rises from €42,000 to €61,000. Champagne? Maybe one cautious glass.
The product-family view tells a colder story:
- Amazon revenue: +€19,000
- Shopify revenue: -€8,400
- bol.com revenue: -€3,600
- Total product-family revenue: +€7,000
- Amazon contribution margin per order: €9.20
- Shopify contribution margin per order: €15.80
- Amazon return/refund impact: 4.5% of revenue versus 2.1% on Shopify
The Amazon dashboard says the extra €2,500 ad spend created €19,000 revenue. The family-level view says roughly €12,000 of that revenue may have shifted from higher-margin channels. Worse, the incremental €7,000 came through a lower-margin route. After ads and returns, retained contribution increased by only €820.
The right action is not “turn Amazon off”. Amazon is still a discovery channel. The right action is to split campaign roles. Keep non-branded category campaigns live for acquisition, cap branded defence, exclude terms where Shopify already converts profitably, and reserve stock for the direct subscription pack. In FiveX, this is where marketplace ad spend, SKU margin and inventory guardrails should sit in the same decision view. Amazon gets budget where it adds new buyers. It does not get paid twice for demand Shopify already earned.
Scenario 2: LumaLift Home and the bol discount trap
LumaLift Home sells a cordless desk lamp on Shopify, bol.com and a French Mirakl retailer. The lamp retails at €49.95. Shopify usually sells 900 units per month at a €16.40 contribution margin. bol.com sells 650 units at €11.70 contribution margin. The Mirakl retailer sells 220 units at €9.90 contribution margin but opens access to a new B2B audience.
In week one of a back-to-school campaign, bol.com pushes the team into a €44.95 promo. bol orders jump from 160 to 310 units. Looks good. But Shopify orders fall from 225 to 150 units, and the Mirakl retailer slows because shared warehouse stock drops below the replenishment threshold.
- bol incremental units: +150
- Shopify lost units: -75
- Mirakl lost units: -35
- Net new units: +40
- Promo margin on bol: €7.10 per order
- Lost Shopify margin: 75 × €16.40 = €1,230
- Lost Mirakl margin: 35 × €9.90 = €346.50
- New bol margin: 150 × €7.10 = €1,065
The campaign added orders and still lost contribution margin. This is the classic discount trap: the marketplace celebrates velocity, while the product family loses margin and stock cover. The fix is a channel permission rule: bol promotions are only allowed when net stock cover is above 45 days, Shopify is not running a full-price email sequence, and the expected margin delta stays positive after displaced orders.
FiveX helps by combining product profitability, channel sales and stock analytics. The point is not to make bol.com look bad. The point is to know when bol.com should receive a promo SKU, when Shopify should hold the premium bundle, and when Mirakl should keep stock because the B2B customer is genuinely incremental.
Scenario 3: Atlas Kitchen and the false TikTok spike
Atlas Kitchen launches a TikTok Shop creator push for a compact blender. One creator video sends 480 orders in four days. The TikTok dashboard shows €23,040 revenue, a 2.9 ROAS and a creator commission of 12%. Everyone is excited because the channel finally “worked”.
Then the multi-channel analytics view lands. Amazon organic rank for the same blender did not improve. Shopify paid social conversions dropped by 95 orders during the same period. Amazon Sponsored Products stayed live and spent €1,100 while stock was being pulled into TikTok fulfilment. The blended return rate on TikTok is projected at 11%, compared with 5% on Amazon.
Here the cannibalization is not only customer overlap. It is inventory and operational cannibalization. TikTok did create some new demand, but it also consumed stock, distracted fulfilment, and left profitable Amazon ads running against a weakening availability position.
The smarter rule: creator spikes should trigger an inventory permission check before paid amplification scales. If stock cover falls below 21 days, TikTok budget pauses, Amazon defence stays live only for high-margin terms, and replenishment priority is recalculated by contribution margin per available unit. That is exactly the kind of operating logic multi-channel analytics should support.
How to build the dashboard
A useful channel-cannibalization dashboard does not need to be huge. It needs the right grain. Build it around SKU family × channel × week.
- Revenue and orders: by marketplace, country and fulfilment route.
- Contribution margin: revenue minus marketplace fees, fulfilment, payment fees, COGS, returns, discounts and ad spend.
- Ad role: branded defence, generic discovery, competitor conquesting, retargeting or promo support.
- Customer signal: new-to-brand where available, first order versus repeat order, subscription or returning customer behaviour.
- Stock pressure: days of supply, inbound inventory, stockout risk and channel reservation.
- Transfer warning: one channel up, another channel down, total family margin flat or worse.
Then create three simple labels:
- Incremental: total family contribution grows and stock pressure remains acceptable.
- Transfer: channel revenue grows, but another channel declines and total family margin barely moves.
- Destructive: channel revenue grows while total family contribution, stock cover or cash timing gets worse.
This is where FiveX should feel practical rather than theoretical. The platform can connect your marketplace data, ad performance, product profitability and inventory insights so the team sees the same commercial truth. The ad manager sees which campaigns deserve spend. The operator sees which channel deserves stock. Finance sees whether revenue became retained contribution margin. Everyone argues less. A tiny miracle.
The operator rule: do not punish channels, assign jobs
The goal is not to crown one perfect channel. Shopify is not always better because margins are higher. Amazon is not always worse because fees are higher. bol.com is not automatically cannibalizing direct sales because it runs a promotion. The job is to assign a role and measure the channel against that role.
Use this simple operating model:
- Amazon: acquire high-intent shoppers, defend core listings, prove category demand, but do not overpay for branded demand already captured elsewhere.
- Shopify: own customer relationships, subscriptions, bundles, email profit and premium positioning.
- bol.com: capture Benelux marketplace demand, test promo elasticity, but protect full-price direct moments and stock cover.
- Mirakl retailers: open retailer-specific audiences, but monitor service levels, payout timing and operational complexity.
- TikTok Shop: create demand spikes and social proof, but only with inventory and return guardrails.
Once channels have jobs, cannibalization becomes easier to discuss. You are no longer asking whether Amazon “stole” sales. You are asking whether Amazon performed its assigned job at a contribution margin the business can accept.
Final thought
Multi-channel growth gets messy because the customer does not respect your reporting structure. They search on Amazon, compare on Google, click a TikTok video, open an email, and buy wherever the offer feels easiest. Your dashboards need to be more grown-up than the journey.
Channel cannibalization analytics is not about fear. It is about honesty. Some marketplace growth is truly incremental. Some is a useful trade-off. Some is just expensive order relocation wearing a nice ROAS hat.
If you sell across marketplaces and your weekly meeting still ranks channels by revenue, start with one product family. Map every SKU, calculate contribution margin by channel, add stock pressure, then mark growth as incremental, transfer or destructive. Once you can do that for one family, you can do it for the whole catalog.
That is the difference between “we sell everywhere” and “we know where the next profitable order should happen”. FiveX exists to make that second sentence easier to run every week.