by Trivas.aiMost CMOs walk into a board meeting with a slide that says "ROAS is up 12%" and hope nobody asks what that number is actually made of. Then someone on the board who used to run finance asks, "Up compared to what, and does that include the fees?" and the room gets quiet.
That's the moment this post is trying to help you avoid. Here's what CMO should know about ecommerce analytics before you're standing in front of the board with a dashboard you didn't build and can't fully defend.
The job has changed. CMOs used to get graded on campaign metrics: impressions, click-through rate, maybe a vague nod to brand awareness. Now the board wants revenue outcomes, and they want to know the marketing spend actually produced them. That shift means analytics literacy isn't a nice-to-have skill for a CMO anymore. It's baseline job requirement.
Here's the uncomfortable part. Most CMOs are working off dashboards built by the ad platforms themselves. Meta reports Meta's performance. Google reports Google's performance. Both are structurally incentivized to show inflated ROAS for their own channel, because that's what keeps budget flowing into them. Nobody at Meta is losing sleep over whether their attribution model is stealing credit from your email list or your organic search traffic.
This article isn't going to turn you into a data analyst. That's not the goal, and it's not realistic in one sitting. The goal is narrower: give you the handful of concepts that let you ask sharper questions of your data team or your analytics vendor, so you're not just nodding along to numbers you don't trust.
ROAS gets top billing in most marketing decks. It shouldn't. Contribution margin per order is the better headline metric because it actually accounts for COGS, shipping, and payment processing fees, not just what you spent on ads. A campaign can post a 4x ROAS and still lose money once you factor in a product with thin margins and a $12 shipping cost. Boards care about money left over, not spend multiples.
Second: stop blending your CAC. New customer CAC and blended CAC tell you two very different stories, and reporting only the blended number hides which one is true. Blended CAC folds in existing customers who were going to buy anyway, so it makes paid acquisition look more efficient than it is. If you want to know whether your paid channels are creating new demand or just harvesting demand that already existed, you need the new-customer number isolated.
Third: LTV:CAC ratio, and this one gets mishandled constantly. Most brands slap a 12-month LTV projection on top of a CAC number, but if the brand has less than 12 months of actual cohort data, that LTV figure is a guess dressed up as a metric. Use the cohort window you actually have data for. A confident 6-month LTV beats a fabricated 12-month one.
Fourth: marketing efficiency ratio, or MER (total revenue divided by total ad spend). It's blunt, but that's the point. When channel-level attribution is shaky, and it usually is, MER is the sanity check that keeps you from over-trusting any single platform's story.
Metric
What it tells you
Where it fails
ROAS
Revenue per dollar spent on a channel
Ignores margin, easily inflated by platform attribution
Contribution margin per order
Actual profit after COGS, shipping, fees
Requires clean cost data feeding into the dashboard
New customer CAC
Cost to acquire someone who wasn't already a customer
Needs first-time vs repeat order tagging
MER
Whole-business ad efficiency
Too blunt for channel-level decisions
Here's the mechanic that inflates your numbers. Meta's pixel claims a conversion under its own attribution window. Google Ads claims the same conversion under its own model. Add the two ROAS numbers together and you're crediting more revenue than you actually generated, sometimes by a wide margin. Nobody is lying to you exactly. Each platform is just reporting inside its own walled garden, and nobody's checking the totals against reality.
iOS 14.5 and the slow death of third-party cookies made this worse, not just fuzzier. It's not that platform-reported conversions got "a little less accurate." They got directionally weaker: platforms started modeling conversions they couldn't directly observe, which means the number on your dashboard might be describing a customer journey that didn't happen the way the report claims.
The fix is to stop treating ad platform dashboards as your source of truth for conversions. Build a revenue view from GA4 and your actual order data, Shopify or Amazon, and use the ad platforms only for spend data, not conversion credit. GA4 funnel data reconciled against real orders will always beat a platform's self-reported win.
This is also where infrastructure matters more than most CMOs realize. Reconciling GA4, Shopify, Amazon, and ad platform spend across all your channels isn't a spreadsheet job once you're past a handful of SKUs and campaigns. It's why Redshift-backed warehousing exists as the plumbing behind this kind of reconciliation, it's built to hold and cross-reference multiple large data sources instead of asking you to trust whichever platform shouts loudest.
Think of it in three layers.
Layer 1 is raw data integration: Shopify or Amazon order data, ad platform spend, and GA4 funnels, all landing in one place. Not five browser tabs and five different logins that someone has to manually stitch together every Monday.
Layer 2 is a unified dashboard layer. You should be able to open it yourself, whenever you want, without waiting for an analyst to build you a deck. If your team's answer to "can I see this before Thursday's meeting" is routinely no, that's a stack problem, not a people problem.
Layer 3 is the insights layer, an "AI Wingman" style function that flags anomalies for you. CAC spiked on a specific SKU. A channel's efficiency dropped 15% week over week. You shouldn't need a human staring at fifteen charts to catch that. This is what an insights layer is actually for, surfacing the thing you'd otherwise miss until it's already cost you a month of spend.
Here's the metric most teams never track: time-to-answer. If a simple question, "what was our new customer CAC on Meta last week," takes more than a day to answer, the stack is broken. That's not a knock on the analyst. That's a sign the BI reporting layer isn't doing its job.
Reporting tells you what happened. Forecasting tells you what's likely to happen next, and most marketing orgs are good at the first and bad at the second.
Most ecommerce brands are still forecasting in a spreadsheet with a manually typed growth-rate assumption sitting in one cell. "Assume 15% MoM growth." Where did 15% come from? Usually a gut feeling from last quarter, not a model trained on actual historical channel performance and seasonality.
Here's a concrete use case. Say you're considering shifting 20% of budget from Meta to TikTok. Instead of running that experiment live with real dollars and finding out in 30 days whether it worked, you can simulate it first against historical performance and seasonal patterns. Forecasting and simulation tools exist for exactly this, testing the shift on paper before it touches your actual budget.
This is also your leverage point with the CFO. "I want more budget" is a request. "Here's a forecast showing what an extra $50k in Q3 spend does to contribution margin, based on our historical seasonal curve" is a case. One of those gets approved in the room. The other gets tabled for next quarter.
A few things worth asking about directly, because they rarely surface on their own.
Cross-channel double counting. Same root issue as the attribution problem above, but it shows up again at the reporting rollup level. If nobody's deduplicating conversions across platforms before they hit the exec summary, your top-line revenue attribution is inflated twice over.
Return and refund lag. Margin numbers calculated in the first 30 days after a product launch often look better than they are, because returns haven't come back yet. That "70% margin" on a new SKU can quietly drop once the refund window closes. Ask whether your margin reporting accounts for this lag or just ignores it.
Amazon and Shopify living in separate silos. If your team can't blend order data across both, you can't calculate a real blended CAC or an honest LTV. You end up with two half-pictures instead of one whole one.
The single-analyst bottleneck. If one person is the only one who knows how to pull a custom report, that's not a reporting stack, that's a dependency risk. What happens when they're on vacation and the board wants a number by Friday?
You don't need to rebuild your entire analytics stack in a quarter. Start smaller.
Get one source-of-truth revenue dashboard working before you touch forecasting or AI-driven insights. Trying to layer predictive modeling on top of unreliable baseline data just gets you a confident-looking wrong answer faster.
A reasonable first step: connect your Shopify data and ad platform spend into a single view, then compare that against what Meta and Google are self-reporting. The gap between the two is usually the most convincing argument you'll ever make to your own team about why platform dashboards can't be the final word.
If you want to see what this actually looks like instead of just reading about it, a practical guide or a trial run through Trivas is a faster way to get there than another round of internal debate. Either way, the next board meeting is a good deadline to work backward from.
explore more insights
3 min read
3 min read
3 min read
.d53b12e5.png&w=3840&q=75 "Trivas.ai")