by Trivas.aiMost brands can tell you their total revenue for last month. Fewer can tell you which specific SKU actually made them money once you subtract ad spend, returns, and COGS. That gap is exactly what product performance analysis is supposed to close: a real, SKU-level view of what's working, not a storewide number that hides the mess underneath.
Product performance analysis is the practice of tracking sales, margin, and ad efficiency at the individual SKU or listing level, instead of settling for storewide revenue. Not "how did the store do this month" but "how did this exact variant, in this exact size, do this month, after everything it cost to sell it."
That distinction matters more than it sounds like it should. Total revenue can climb 15% year over year while a third of your catalog quietly bleeds money. Nobody notices because the top-line number looks fine. Basic sales reporting will tell you revenue is up. It won't tell you that 30% of your SKUs are underwater once you factor in ad spend and refunds.
Here's a test: ask a founder to name their top 5 sellers by units. They'll rattle them off in seconds. Ask them to name the top 5 by contribution margin. Long pause. Usually a guess. That gap between "what sells" and "what's actually profitable" is the whole reason this kind of analysis exists.
Revenue per SKU and units sold are the starting point, but on their own they're almost decorative. What matters is weighting them against contribution margin after COGS, marketplace or payment fees, and ad spend. A SKU doing $50,000 a month at 8% margin is worse than one doing $18,000 at 40%.
A few metrics worth tracking on every SKU:
None of these are exotic. The problem is almost nobody looks at them by SKU consistently, because pulling them together by SKU, every week, across platforms is tedious enough that most teams give up and default to the top-line view instead.
The inputs for product performance analysis aren't hard to name. Shopify or Amazon Seller Central gives you unit sales and order data. Meta and Google Ads platforms give you spend and attribution at the product or campaign level. GA4 gives you on-site behavior: product page views, add-to-cart rate, where people drop off before checkout.
The problem isn't availability. It's that these four sources don't talk to each other. Shopify doesn't know your Meta ad spend per SKU. Google Ads doesn't know your COGS. GA4 doesn't know your refund rate. So somebody on the team ends up exporting CSVs from four platforms every week and stitching them together in a spreadsheet, matching SKU IDs by hand, praying the naming conventions match across systems.
This is where most teams lose real hours every week, not analyzing the data but assembling it. And margin data specifically tends to live nowhere centralized at all. It's in a supplier invoice, a separate COGS spreadsheet, sometimes just in someone's head. If you're running product performance analysis on Shopify and Amazon side by side, the spreadsheet approach breaks down fast once you're managing more than a couple dozen SKUs.
The biggest one: looking at units sold or revenue alone without netting out ad spend and COGS per SKU. A product can look like your best performer on a units-sold report and be your worst performer on a margin report. Both reports are technically correct. Only one of them tells you what to actually do.
Second mistake: averaging performance across a whole product category instead of isolating individual SKUs or variants. "Hoodies are performing well" hides the fact that the black medium sells out every restock while three other colorways sit in a warehouse for months. Category-level averages flatten exactly the signal you're trying to find.
Third: relying on last-click attribution only. This overcredits bottom-funnel retargeting ads for products customers actually discovered somewhere else entirely, whether that's organic search, a TikTok mention, or word of mouth. A retargeting ad "closing" a sale isn't the same as that ad being why the product sells. Confuse the two and you'll keep dumping budget into the wrong campaigns while starving the ones actually driving discovery.
Fourth: reviewing performance monthly instead of weekly. A SKU's sell-through can start declining in week one. If your review cadence is monthly, you find out three or four weeks late, after you've already reordered inventory or kept ad spend flowing to something that's already cooling off.
None of this matters if it doesn't change what you do next week. A few concrete ways to act on it:
Reallocate ad budget based on trend, not habit. If a SKU's CPA is climbing while a different SKU is selling well organically with minimal ad support, that's a signal to shift spend, not just watch it happen.
Set a real rule for discontinuing a SKU. Something specific, like: contribution margin below 10% for two consecutive months means it's cut or repriced. Vague rules like "if it stops selling well" never actually get enforced because nobody agrees on when "well" stopped.
Feed this into reorder and forecasting decisions. Gut feel about "what'll sell next quarter" is how brands end up with a warehouse full of last season's colorway. Sell-through and margin trends by SKU are a much better input than a merchandiser's hunch.
Let it inform new product development. If your top three SKUs by contribution margin all share an attribute (a scent, a size, a price point), that's not a coincidence worth ignoring. Double down on what the data already shows is working before you guess at what's next.
Product performance analysis is one layer of the full picture, not the whole thing. It sits alongside channel-level analysis (how is Meta doing versus Google) and funnel-level analysis (where are people dropping off before purchase). All three need to work together, because a SKU's poor performance might actually be a channel problem, or a funnel problem, not a product problem at all.
The honest catch is that product-level views only become reliable once Shopify, Amazon, and ad platform data are unified in one place, instead of pulled separately every time someone wants an answer. That's the difference between a report you trust and a report you double-check against three other tabs before you believe it. Tools like BI reporting built to unify these sources, or an insights layer that surfaces SKU-level trends automatically, exist specifically because the spreadsheet version of this doesn't scale past a certain SKU count.
If you're still stitching this together by hand every week, it might be worth seeing what a unified view actually looks like. Check out getting started for a sense of what that setup involves, and if you want more breakdowns like this one, keep an eye on future posts here.
explore more insights
3 min read
3 min read
3 min read
.d53b12e5.png&w=3840&q=75 "Trivas.ai")