by Om RathodMonday morning, coffee in hand, a founder opens three tabs: Shopify, Meta Ads Manager, GA4. Each one tells a slightly different story. Shopify says sales are up. Meta says ROAS dipped. GA4 says conversion rate cratered. Which one is "the truth"?
None of them, individually. That's the trap.
The problem isn't a lack of data. Most brands have more dashboards than they know what to do with. The problem is nobody taught them how to read an ecommerce performance dashboard as a system instead of a collection of disconnected numbers. Each platform reports its own version of reality, using its own attribution logic, its own date boundaries, its own definitions of "conversion."
This post walks through the anatomy of a dashboard that actually makes sense: the layers it needs, the metrics worth your attention, and the specific ways numbers lie to you if you don't know what to check first.
A dashboard that only shows revenue is a vanity mirror. A useful one has four layers, stacked so you can trace cause and effect.
Revenue layer. Gross sales, net sales after returns and discounts, and average order value. Gross tells you activity. Net tells you what you actually kept.
Acquisition layer. Total ad spend, blended CAC, and channel-level ROAS for Meta, Google, TikTok, wherever you're spending. This layer answers "what did it cost to get these sales."
Conversion layer. Sessions, conversion rate, cart abandonment, pulled from GA4 or Shopify analytics. This is where you catch friction before it shows up in revenue.
Retention layer. Repeat purchase rate, LTV, subscription or reorder metrics if that applies to your model. This is the layer most dashboards skip, and it's usually the one that explains whether growth is real or rented.
Miss any one of these four, and you get a partial read. A dashboard heavy on acquisition but light on retention will make a leaky bucket look like a growth machine. For a deeper breakdown of how these layers should actually connect inside one view, Trivas's guide to dashboards and analytics is worth a look.
Gross revenue is the easiest number to feel good about and the least useful one to act on.
Net revenue subtracts returns and discounts. Contribution margin subtracts the cost of goods, shipping, and payment processing on top of that. Founders who only track gross tend to overestimate how profitable a "good week" actually was, especially during a promo.
Before comparing week over week or month over month, check two settings: the date range and the attribution window. A 7-day click window on Meta will show different numbers than a 1-day click window, and comparing periods with mismatched windows makes trend lines meaningless.
Here's a real trap: a revenue spike from a sitewide discount can hide a rising return rate underneath it. Revenue looks great. Margin quietly erodes. You won't see it unless you're looking at net revenue and return rate side by side, not just the top-line number.
One more thing that trips people up: Amazon and Shopify don't report revenue the same way. Amazon nets out fees at settlement, so what hits your Amazon report is already post-fee. Shopify shows raw order value before any of that. Blend the two without normalizing first, and your "total revenue" number is quietly wrong.
Platform-reported ROAS, blended ROAS, and MER (marketing efficiency ratio) all answer different questions, and using the wrong one is how founders convince themselves a campaign is working when it isn't.
Platform ROAS
Blended ROAS
MER
Platform ROAS is almost always inflated. Meta and Google use last-click or extended click-through windows that credit themselves for sales they only partially influenced. Read it in isolation and you'll overspend on channels that look efficient but aren't carrying the business.
A quick sanity check: take total ad spend for the period, divide it into total revenue for that same period, and compare that against what each platform is individually claiming. If the platforms' combined attributed revenue exceeds your actual total revenue, that's overlap, and it happens more than people realize.
Also watch CAC even when ROAS looks flat. A slowly climbing CAC is a leading indicator that something is degrading (creative fatigue, audience saturation, rising CPMs) well before it shows up as a ROAS problem you can't ignore.
The funnel is where you find out why traffic isn't converting, not just that it isn't.
GA4 breaks it into stages: sessions, product views, add-to-cart, checkout initiated, purchase. The drop-off between each stage tells a specific story. A big drop from add-to-cart to checkout usually points to shipping cost surprises or a payment friction issue, not ad quality.
Quick diagnostic worth running before you touch your ad accounts: if traffic is flat but conversion rate drops, check site speed or a recent theme/UX change first. It's tempting to blame the ads. Often the site changed and nobody flagged it.
Split the funnel by device before you draw any conclusions. Aggregate numbers frequently mask a mobile checkout problem that's dragging down the whole conversion rate while desktop performs fine. If you haven't set this up properly, Trivas's GA4 solution walks through getting funnel data structured in a way that actually separates device performance instead of blending it.
Most bad decisions don't come from bad data. They come from misreading good data.
Mismatched date ranges and timezones. Amazon reports in UTC. Shopify reports in your store's local timezone. Compare "yesterday" across both without adjusting, and you're comparing two different 24-hour windows.
Treating a spike as a trend. One good day (or bad day) isn't a pattern. Check a 7-day or 28-day rolling average before you decide anything changed.
Ignoring data latency. Ad platforms take 24 to 72 hours to finalize numbers. "Today's" ROAS is a rough draft, not a final answer. Reacting to it in real time is how you end up chasing noise.
Overlapping attribution across channels. If you're checking channel-level metrics without accounting for the fact that Meta and Google can both claim credit for the same sale, your "total" is inflated and you don't know by how much.
Metric names don't even mean the same thing across tools half the time. "Conversion rate" in Shopify isn't calculated the same way as "conversion rate" in GA4. When definitions get fuzzy, Trivas's data dictionary is a fast way to check what a given platform actually means by a term before you compare it to another one.
Tab-switching between Shopify, Amazon Seller Central, and two or three ad platforms works fine when you're doing a few hundred orders a month. Past that, it starts costing real hours, and the stitched-together spreadsheet becomes its own source of errors.
A unified dashboard built on a proper data warehouse (Redshift, in Trivas's case) solves the actual root problem: one source of truth, with consistent attribution logic applied across every channel, instead of five tools each grading their own homework.
Trivas also layers Wingman AI on top of that data, so instead of manually digging through every anomaly, you get a plain-language explanation of what shifted and why. That's not a replacement for understanding the framework in this post. It's what you reach for once applying that framework by hand every morning stops being a good use of your time. For founders juggling this alone, Trivas's resources for founders and CEOs cover what that transition typically looks like.
Apply the four-layer model, revenue, acquisition, conversion, retention, to any dashboard you're looking at, and most confusion disappears.
Before trusting a single number, check the date range, the attribution window, and whether the data has fully settled. "Today" is rarely final.
When metric definitions across tools don't match, don't guess. Check the source.
If you're past the point where checking five tabs every morning makes sense, Trivas's BI reporting unifies these layers automatically, so the framework in this post is already built in instead of something you're maintaining by hand.
explore more insights
3 min read
3 min read
3 min read
.d53b12e5.png&w=3840&q=75 "Trivas.ai")