by Trivas.aiPicture a skincare brand running $200k a month in one category on Amazon. Looks great on paper. Except ad spend to revenue on that category is sitting at 38%. Meanwhile the same category on Shopify is doing less total revenue, but ad spend to revenue is 12%.
Rank these by top-line numbers and Amazon wins easily. Rank them by what's actually left after costs, and Shopify is the one funding payroll.
This is the trap most brands fall into. They sort categories by revenue or units sold, because that's what's sitting right there on the dashboard. But revenue doesn't tell you which categories are profitable, it tells you which ones sell. Two very different things.
Here's the problem underneath it: the same SKU can carry wildly different economics depending on where it's sold. Fees differ. Ad costs differ. Discount depth differs. Return rates differ. A hoodie that returns at 8% on Shopify might return at 18% on a marketplace with looser return policies. None of that shows up in a revenue ranking.
Figuring out how to identify the most profitable product category by channel means building a view that strips out revenue vanity and gets to what's actually left in the bank. That starts with knowing which metrics matter, and which ones lie to you.
Revenue and units sold are surface metrics. They tell you activity, not health. Four numbers actually matter here.
Contribution margin per category. Revenue minus COGS, minus channel fees, minus shipping, minus the ad spend allocated to that specific category. Not store-wide. Per category. This is the number that answers "does this category make money," and it's the one most brands skip because it takes real allocation work.
Blended CAC by category, not blended CAC for the whole store. A store-wide CAC of $22 hides the fact that your accessories category might be acquiring at $8 while your bundles are acquiring at $45. Averaging those together erases the signal.
Return and refund rate by category. Apparel returns nothing like electronics. Consumables barely return at all. If you're not netting out returns per category, you're overstating margin on exactly the categories that need scrutiny most.
Channel-specific costs. These vary more than people expect:
Get these four right and you've got a real basis for BI reporting that reflects actual channel economics instead of a vanity leaderboard.
Same product, same cost to make, wildly different profitability depending on where it sells. This isn't a fluke, it's structural.
Amazon eats profitability through advertising cost of sales (ACOS) and FBA fees. Bulky or heavy items get hit twice: once on referral fees, again on storage and fulfillment. A category can look fine on gross revenue and still be barely breaking even once ACOS climbs past 25-30%.
Shopify profitability lives or dies on different levers: discount code usage, email and SMS-driven repeat purchase rate, and payment processing costs. A category with strong repeat purchase behavior can carry thinner ad spend and still come out ahead, because you're not paying to reacquire the same customer every 30 days.
Meta and Google introduce volatility that neither of the other two channels has in quite the same way. CPMs move with the market, audience saturation builds over a campaign's lifecycle, and a category that looked strong in March can look mediocre in April with zero change to the product itself.
Take skincare sets as an example. On Shopify, a set with a built-in subscribe-and-save option can be highly profitable because repeat purchases carry almost no incremental ad cost. On Amazon, that same set might be marginal, because ranking for the right keywords requires sustained ad spend, and the referral fee structure doesn't reward the loyalty piece the same way. Same product. Same margin on paper. Completely different outcome once channel costs get applied.
Here's the actual workflow, step by step.
Step 1: Pull unified sales data by category and channel into one place. Comparing Amazon Seller Central against Shopify's admin against Meta Ads Manager side by side doesn't work. The definitions don't match, the date ranges don't align, and you'll spend more time reconciling than analyzing.
Step 2: Allocate ad spend to categories, not to the store as a whole. Use UTM tagging on Shopify campans and channel-level category breakdowns on Amazon and marketplace ads. A flat blended CAC applied evenly across categories will always understate winners and overstate losers.
Step 3: Subtract channel-specific fees per category. Referral fees, FBA costs, payment processing, fulfillment, whatever applies on that specific channel for that specific category.
Step 4: Calculate contribution margin per category per channel. This is your real ranking. Not revenue, not units, contribution margin.
Step 5: Re-run this monthly. CPMs shift. ACOS shifts. Amazon changes fee structures without much warning. A category that ranked first in January can rank fourth by April, and if you're not re-checking, you won't know until the P&L tells you the hard way.
A few habits quietly wreck this analysis, even for teams that are otherwise diligent.
Using store-wide blended ROAS instead of category-level ROAS. Blended ROAS is a comfort metric. It hides the categories that are losing money by burying them inside categories that are winning big. You need the category-level number, full stop.
Ignoring return rates when calculating margin. This one hurts apparel brands specifically. A category that looks like it's running a healthy 40% margin can lose 15-20% of that straight back to returns. If returns aren't in the model, the model is wrong.
Comparing raw revenue across channels without normalizing for fee structure. A dollar of Amazon revenue and a dollar of Shopify revenue are not the same dollar once fees are applied. Comparing them raw is comparing two different currencies and calling it apples to apples.
Treating a one-month spike as a trend. A category that pops for four weeks because of a seasonal push, an influencer mention, or a CPM dip isn't necessarily your new top performer. Check 90 days before you reallocate budget based on it.
Doing this by hand works, until it doesn't. Manually blending Amazon exports, Shopify exports, and ad platform exports into a spreadsheet takes hours, and the whole thing breaks the moment Amazon tweaks a fee structure or your ad platform changes its attribution window.
Trivas pulls Amazon, Shopify, Meta, Google Ads, and GA4 data into one Redshift-backed dashboard, so category profitability by channel becomes a live view instead of a spreadsheet you rebuild every month. If you're running both channels, solutions for Amazon and solutions for Shopify sit in the same system, which is the whole point: no toggling between platform-native dashboards that were never built to talk to each other.
The Wingman AI layer sits on top of that and flags when a category's margin drops below a threshold on a specific channel, so you find out the week it happens instead of at the next quarterly review. And the forecasting layer shows which categories are trending toward higher or lower profitability before the quarter closes, which is a very different thing than finding out after the fact. That kind of surfaced insight is the difference between reactive reporting and actual insights you can act on before the number moves against you.
None of this matters if it just sits in a dashboard. Once you've got real contribution margin by category and channel, there are three moves worth making.
Reallocate ad budget away from categories that look strong on revenue but are running weak contribution margin once channel fees are applied. That budget is almost always better spent on a category that's quietly outperforming.
Get specific with channel merchandising. If a category is genuinely more profitable on Shopify than Amazon, push it there. Don't spread merchandising effort evenly just because it feels fair, it isn't a profitability strategy.
And revisit pricing or bundling on channels where a category's margin is thin because of fee structure, not because of weak demand. Those are two different problems with two different fixes, and conflating them leads to the wrong decision.
If you want to see this breakdown running against your own store's data instead of a hypothetical, a Trivas trial will show you exactly where each category stands, channel by channel, before you make the next budget call.
explore more insights
3 min read
3 min read
3 min read
.d53b12e5.png&w=3840&q=75 "Trivas.ai")