Google Ads
Why Your Google Ads ROAS Is Dropping (And What to Actually Do About It)
In January your ROAS was fine. By March it wasn't. You checked everything you know to check. Nothing obvious had changed.
Same budget. Same targets. The campaigns you paused last year are still paused. You added a handful of negatives in February - the usual irrelevant queries. Someone on the team asked if anything was different. You said no, not really.
But the numbers keep moving. 4.8 in January. 4.1 in February. 3.6 in March. Budget review in two weeks.
This is what was actually happening.
A ROAS decline that unfolds over 60 to 90 days without a clear trigger is one of the most common situations in Google Ads - and one of the least well-handled. The instinct is to do something: adjust a target, restructure a campaign, pause a product group. Often that makes things worse, because the action is aimed at the symptom rather than the cause.
Before running through the six findings that account for most of these declines, there are three quick checks worth doing first. If any of them apply, the rest of this might not be relevant to you.
Seasonality. If you're in a category that runs soft at a particular time of year, compare the same period year-on-year rather than month-on-month. A decline that looks alarming in isolation can be completely normal in context. If revenue is also down proportionally, the problem may be the market, not the account.
Attribution changes. A GA4 migration, a change to your conversion tracking setup, or even a new attribution model applied in Google Ads can alter reported conversion numbers without changing actual sales. If ROAS is down in Google Ads but revenue in your eCommerce platform is holding, you have a measurement issue, not a performance issue.
Budget cuts disrupting Smart Bidding. Cutting budget by 20% or more reduces the number of auctions Smart Bidding participates in, which can trigger a recalibration period. If a budget reduction happened in the 30 to 60 days before the decline started, that's likely the cause. The fix is to hold the new budget stable and give the algorithm time to recalibrate - typically 4 to 6 weeks.
If none of those apply, you're looking at a structural issue. Here are the six diagnostic findings that account for the vast majority of unexplained ROAS declines.
Smart Bidding is not a static system. It updates continuously based on conversion signals, and when those signals shift - a change in product mix, a seasonal CVR swing, a price increase across the catalogue - the algorithm's model of what a conversion costs can lag behind reality.
The signal that shows up in the account: impressions stay roughly flat, click volume is similar, but cost per conversion is creeping up week over week. There's no event in the change history. Nothing was touched. The campaign just quietly got more expensive.
The instinct here is usually to raise the target ROAS to compensate. That is the wrong move. Raising your tROAS makes Smart Bidding more conservative - it bids less aggressively to protect the higher target, which reduces impression share and typically worsens the decline. If impression share lost to rank is rising and ROAS is already below target, the correct direction is to lower tROAS slightly, giving the algorithm more room to participate in auctions and gather fresh conversion data.
Never raise tROAS to fix impression share lost to rank. Raising tROAS makes Smart Bidding more conservative, not less. The right lever is lowering tROAS (to loosen the constraint) or improving Quality Score (to reduce CPCs). Raising tROAS when rank is the problem makes it worse.
The fix here is patience first: hold settings stable for 4 to 6 weeks and let the algorithm recalibrate against recent data before making any further adjustments.
Broad match and Smart Bidding work together to expand the queries your ads appear for. In most accounts this is a net positive. The system finds queries you wouldn't have thought to target and converts them profitably. But the same mechanism that finds good traffic can drift toward less commercially relevant queries over time - particularly when negative keyword reviews have been irregular.
The pattern is easy to miss because the top-line metrics don't change much. Impression share holds. Click volume is similar. The campaigns look healthy at the campaign level. What's actually happening is a slow substitution: the profitable queries that were driving revenue are being diluted by adjacent queries that get clicks but don't convert. The average quality of the traffic falls without any single query being the obvious culprit.
The diagnostic check is a search terms report filtered to the past 30 days, 10 or more clicks, zero conversions. Look for patterns rather than individual terms - category drift into adjacent product areas, informational queries entering product campaigns, competitor brand terms absorbing spend. That's where the ROAS is going.
The fix is a structured negative keyword review, followed by a monthly process to keep the query mix tight. Regular negative keyword reviews - at least monthly - are what keep the query mix tight.
Quality Score directly affects what you pay per click. Google rewards higher-quality ads with lower costs for the same position - a term at Quality Score 5 typically costs more per click than the same term at 8. When Quality Scores on your top converting keywords decline, your cost base rises - without any bid changes, without any campaign edits, without anything appearing in the change history.
This is the slowest-moving of the six findings and the hardest to notice. It shows up as a CPC trend that drifts upward over months, with impression share roughly flat, conversion rate roughly flat, but ROAS gradually compressing. The numbers look almost fine in any given week. Over a quarter, they've moved materially.
Add the Quality Score column to your keywords view and sort by spend. Check the three sub-components - expected click-through rate, ad relevance, and landing page experience - for your top 20 terms by cost. If the account has had recent site changes without corresponding updates to landing pages or ad copy, landing page experience is a logical first place to look.
Branded search terms - searches for your business name, product names, or URLs - typically convert at a much higher rate than non-brand terms. They're also cheaper to win - your Quality Score for your own brand queries is typically higher than any competitor's, which lowers your cost per click for that traffic. The revenue they drive is real, but the work of acquiring it is minimal.
Performance Max will bid on branded queries unless you explicitly apply brand exclusions. When this happens, branded traffic gets absorbed into the Performance Max campaign - inflating its reported ROAS, because the high-converting, low-competition branded queries are mixed in with the actual acquisition work the campaign is supposed to be doing. The overall account ROAS looks reasonable because the brand is propping it up.
The problem surfaces when branded traffic volume fluctuates - seasonal dips, reduced organic visibility, a competitor starting to bid on your brand terms. Suddenly the Performance Max ROAS drops sharply, because the non-brand performance was never as strong as it appeared. The account-level ROAS follows it down.
The fix is brand exclusions applied to Performance Max, combined with a separate brand search campaign where you can see and manage that traffic independently. For more detail on how Performance Max cannibalisation works in practice, see the dedicated breakdown in the insights section.
ROAS is calculated from ad spend and conversion value. But the conversion happens on your website, not inside Google Ads. When landing page conversion rate drops, ROAS falls - regardless of how efficiently the campaign is performing in the auction. The ad account looks fine. The problem is elsewhere.
The causes are usually on the site side and easy to overlook from inside the ad platform: a checkout flow that broke on mobile after a site update, a page speed regression after a new plugin was added, a price increase that wasn't matched by updated ad copy, products going out of stock without corresponding ads being paused. None of these generate an alert in Google Ads. The account just slowly gets less efficient.
The diagnostic is in GA4, not Google Ads. Look at session-to-purchase conversion rate by landing page, broken down by device, for the period where ROAS started declining. If CVR is down and your Google Ads metrics - impressions, CTR, average CPC - look unchanged, the issue is on the site. Increasing ad spend into a broken landing page makes the ROAS problem worse, not better. Fix the site first.
You can do everything right in your account and still see ROAS decline if a well-funded competitor enters your key categories and drives up CPCs. You're now paying more per click for the same traffic, which compresses conversion value per dollar of spend. Nothing changed internally. The external environment changed around you.
The signal is in the numbers: impression share lost to rank increases, average CPC rises without changes to your bids or targets, and the Auction Insights report shows new or more prominent competitors on your core terms. It can look like a bidding problem from the inside, but the root cause is competitive pressure you can't directly control.
The response is to concentrate spend where you have the strongest advantage. Tighten targeting around your highest-Quality Score terms - where your lower CPCs give you a structural edge over new entrants who haven't built the same account history. Reduce broad match exposure on terms where you have no Quality Score differentiation. Over time, improving Quality Score is the most durable way to offset CPC pressure, because it lowers your cost floor regardless of what competitors are bidding.
The six findings above cover the vast majority of unexplained ROAS declines in eCommerce accounts. Most accounts showing a 60 to 90 day decline have one primary cause - with one or two secondary contributors compounding it.
The reason the decline is hard to catch isn't that the cause is hidden. It's that the standard in-platform dashboards show you campaign-level metrics, not account-level diagnostics. Impression share looks fine. Spend is on budget. The campaigns are active. Everything looks operational - while something underneath is quietly eroding efficiency.
The person looking at that January-to-March chart - watching the 4.8 become 3.6 - already has the data to find out why. It's in the search terms report, the Quality Score column, the Auction Insights tab, GA4. The problem is knowing which report to pull first, and what to look for when you get there.
That's the part that changes with a structured review.
The six findings above are diagnosable independently. But in most accounts showing a sustained ROAS decline, more than one is active - and they interact. A search term mix problem compounding a Quality Score problem. A Performance Max brand overlap hiding a weak non-brand acquisition result. A landing page CVR issue making a Smart Bidding recalibration look like a bidding failure.
The seventh finding is the relationship between the other six: which one is upstream, which ones are downstream, and which fixes are wasted effort until the upstream problem is resolved. That's the part that requires seeing the whole account at once - not campaign by campaign, but as a connected system where each layer affects the layers beneath it.
In most cases, a declining ROAS has a diagnosable root cause and a fixable path forward. Getting there faster is a question of where you look first.
The Google Ads Gap Analysis identifies the root cause and gives you a prioritised fix list - from $997.
Book here →