by Trivas.aiBy the time you spot a channel bleeding money in your weekly report, it's already cost you a week of budget. Knowing how to detect when a paid channel becomes unprofitable isn't about better reporting, it's about catching the shift while it's still small. Most teams find out three or four days too late, after the spend is already gone.
Most brands check ROAS or blended CAC once a week. Maybe twice if things feel shaky. But channel-level profitability doesn't wait for your Monday review. It can flip negative on a Tuesday and stay that way until someone finally opens the dashboard.
Here's the bigger issue: platform-reported ROAS lies to you, not maliciously, just incompletely. Meta and Google report revenue against ad spend, full stop. No COGS. No shipping cost. No discount codes. No returns. A channel can show 3x ROAS and still be losing money the second you factor in true margin.
So you get this gap between what the ad platform says and what your bank account says. And that gap is where the damage happens. By the time a marketer notices the discrepancy in a weekly pull, you've usually already spent five to ten days feeding a channel that stopped working.
That lag is the whole problem. Not the metric itself, the delay in seeing it move.
ROAS alone won't tell you when a channel goes underwater. You need a few other numbers doing the real work.
Contribution margin per order is the one that matters most. Revenue minus COGS, shipping, payment processing fees, and ad spend. This is the number that tells you if an order actually made money, not just if it generated more revenue than ad cost. A lot of "profitable" channels stop being profitable the moment you run this math honestly.
MER (marketing efficiency ratio) trending down over a rolling 7-day window is your early warning system. Total revenue divided by total spend, tracked over time. One bad day is noise. Three or four days of decline in a row is a pattern.
Blended CAC exceeding contribution margin per new customer is the hard stop. This is the actual breakeven point, not some arbitrary ROAS target. If it costs you more to acquire a customer than that customer contributes in margin, the channel is losing you money on every single acquisition, no matter what the platform dashboard says.
The gap between incremental ROAS and platform-attributed ROAS widening over time usually means a channel is coasting on attribution overlap, not driving real incremental sales. This shows up a lot with retargeting campaigns that are just claiming credit for purchases that would've happened anyway. If you haven't run an incrementality test recently, that gap is probably bigger than you think.
You can sanity check your breakeven math with the ROAS calculator before you build alerts around it. Get the baseline number right first.
You can't detect unprofitability in real time if you don't already know what "unprofitable" looks like for your business. Define it before spend goes sideways, not while you're staring at a bad number in a panic.
Start with breakeven ROAS calculated off real margin, not gut feel. If your margin is 40%, you need roughly 2.5x ROAS just to break even. Anything sustained below that for more than a couple days isn't a rough patch, it's a loss.
Use a rolling window, not single-day numbers. A 3-day or 7-day average smooths out normal daily volatility, weekends, and platform reporting delays. Reacting to one bad day is how you end up pausing a channel that was actually fine.
Set thresholds per channel, not one blanket number for everything. Meta prospecting, Meta retargeting, Google Shopping, and TikTok all carry different acceptable CAC ranges. Treating them the same way guarantees you'll misjudge at least one of them.
Write these thresholds down somewhere the whole team can see. A shared doc, not someone's head. When spend decisions get emotional (and they will, especially with a channel that used to perform), having an agreed-upon number removes the argument. You're not debating whether it's bad, you're checking it against a number everyone already signed off on.
Manual spreadsheet checks fail for a boring reason: someone has to pull data from three or four different places, every single day, and stitch it together correctly. Ad platform exports, Shopify order data, margin sheets. Miss a day and your detection window just doubled.
An actual alert system needs three data sources synced close to real time: ad spend by channel, order-level revenue and margin, and returns/refund data. Miss the returns piece and you'll overstate margin on categories with high return rates, which quietly masks the problem you're trying to catch.
This is the exact gap Trivas Wingman is built to close. It watches channel-level contribution margin against the thresholds you've set and flags drops automatically, instead of waiting for someone to open a report on a Monday. No manual export. No reconciling three spreadsheets before you even know if there's a problem.
The alert should fire on a trend break, not a single anomaly. Three consecutive days below threshold is a reasonable bar. One bad day gets ignored by design, because one bad day happens constantly and means nothing on its own.
Don't pause the channel yet. First, figure out what you're actually looking at.
Check whether it's a tracking issue before you assume it's a demand issue. Post-iOS 14.5, undercounted conversions are common enough that a real performance drop and a measurement gap can look identical from the outside. Cross-reference platform numbers against your actual order data before you draw conclusions.
If it's a real drop, segment before you pause. Look at campaign and ad set level, not just the channel total. Often one tired campaign or one bloated ad set is dragging the whole channel's average down while the rest of it is fine. Pausing the entire channel over one bad campaign is a bigger overcorrection than the problem calls for.
Try a spend reduction first. Cut 30-50% for 48 to 72 hours and watch whether CAC recovers at the lower volume. Sometimes a channel is only unprofitable because you scaled past its efficient ceiling, not because the channel itself is broken.
Last, check GA4 funnel data. If conversion rate dropped site-wide, not just on traffic from this channel, the problem might not be the channel at all. It might be your landing page, your checkout, or your site speed. This is a distinction that matters a lot for performance marketers trying to figure out where to actually spend the next hour of troubleshooting.
Alerts catch emergencies. They don't replace a standing review.
Put contribution margin by channel on the agenda every week, alongside whatever alerts fired. Alerts are for the sudden drops. The weekly review is for the slow drift, the channel that's been losing a little efficiency every week for two months without ever tripping a threshold.
Tie that review directly to budget decisions. Move spend from the declining channel to the improving one that same week, not at the end of the month. Monthly reallocation cycles are how brands keep funding a channel that's been quietly losing money for three weeks straight, just because nobody looked until the calendar said to.
Centralizing Amazon, Shopify, Meta, and Google data on one Redshift-backed dashboard removes the manual export-and-reconcile step that delays detection in the first place. If your team is still stitching together CSVs from four platforms before anyone can answer "are we profitable this week," that's the actual bottleneck, not the analysis itself. Forecasting and simulation tools can also help you model what a reallocation actually does to blended CAC before you pull the trigger.
Three things make this work: real margin math instead of platform-reported ROAS, thresholds set per channel before you're under pressure, and alerts that fire on a trend, not a single bad day.
None of this requires more hours in spreadsheets. It requires the right data in one place, checked automatically, against numbers you already agreed on.
If you're curious how Trivas Wingman surfaces these alerts across your ad accounts without the manual pull, it's worth a look, especially if you're still finding out about a bad channel a week after the damage was already done. Start a trial and see what it catches on your live ad account data.
explore more insights
3 min read
3 min read
3 min read
.d53b12e5.png&w=3840&q=75 "Trivas.ai")