Why Is My Google Ads ROAS Dropping? Diagnose It with AI

Google Ads ROAS falls when reported conversion value grows more slowly than ad cost. The account may be paying more for the same outcomes, converting fewer visitors, generating lower-value outcomes, or reporting values differently. Recent dates can also look weak before delayed conversions arrive.

Do not start by raising the target ROAS, cutting every expensive campaign, or reallocating budget from a blended account average. First compare two mature periods, confirm which conversion actions and values are included, decompose the movement, and find the campaigns or segments that contributed to it.

A useful diagnosis should end with one of five conclusions: act on traffic cost, act on conversion rate, act on value per conversion, fix measurement, or monitor because the apparent decline is not mature.

Google Ads reports conversion value per cost as the value of included conversions divided by ad-interaction cost. A ratio of 5.0 corresponds to a 500% ROAS: $5 of reported conversion value for every $1 of ad spend.

Reported ROAS = conversion value ÷ cost × 100%

For click-led campaigns, you can decompose that ratio further:

ROAS ≈ conversion rate × average conversion value ÷ average CPC × 100%

This is a diagnostic identity, not a bidding instruction. It shows the three economic levers behind the metric: what a click costs, how often clicks convert, and how much value each conversion contributes.

Suppose average CPC is $4, conversion rate is 5%, and average conversion value is $120. Expected value per click is $6, so ROAS is about 150%. If CPC rises to $5, ROAS falls to 120%. If conversion rate falls to 4% while CPC stays at $4, ROAS also falls to 120%. If average conversion value falls to $96, the result is again 120%. The same headline metric can require three different remedies.

Use this evidence table before changing the account:

Compare two complete periods in the Google Ads account’s time zone. Match weekdays where possible, separate promotions or unusual events, and exclude the newest dates if the normal conversion cycle is still open.

Google’s conversion-lag reporting exists because clicks and costs can arrive before their resulting conversions. That can make recent ROAS look artificially low. Compare mature windows first, then inspect a daily trend to see whether the decline began on a specific date.

Use both a percentage and a minimum value threshold. A 30% decline from 20% to 14% ROAS is mathematically large but commercially meaningless. A smaller percentage decline on a high-spend campaign may deserve immediate attention.

ROAS is only as trustworthy as the conversion values behind it. Confirm the conversion actions included in the Conversions column, their counting rules, value source, currency, attribution setup, and recent activity.

Look for changes such as:

Also compare Google Ads value with the ecommerce platform or CRM. A stable backend with a falling Google Ads value points toward measurement or attribution. A matching decline suggests the business outcome really changed.

Calculate five metrics for each period: cost, conversions, conversion value, average CPC, and conversion rate. Then add average value per conversion and ROAS.

Ask the questions in this order:

Keep the math additive where possible. Account ROAS is total conversion value divided by total cost. It is not the simple average of campaign ROAS values, because a $50 campaign should not carry the same weight as a $50,000 campaign.

For each campaign, compare cost, conversions, conversion value, ROAS, and average value per conversion. Rank campaigns by incremental spend, lost conversion value, or both. Then calculate their share of account cost in each period.

This catches mix shifts. Account ROAS can fall even when every campaign’s own ROAS is unchanged if spend moves from a 600% brand campaign toward a 200% prospecting campaign. That may be a deliberate growth decision—not a broken campaign.

Do not let one small campaign with a dramatic percentage change distract from the entities moving the account total. Show absolute value and spend beside every percentage.

Once you know which campaigns matter, split performance by the dimensions most likely to explain the movement: campaign type, device, geography, keyword, search term, and day.

Look for a segment that gained spend share while producing less value per dollar. Common patterns include broader queries, a weaker location, a mobile-heavy mix, a new campaign, or a budget shift toward prospecting. Treat these as evidence, not a verdict. A lower platform ROAS may still be acceptable if the segment creates higher lifetime value that Google Ads cannot see.

For Search, inspect the actual queries and match types. For Performance Max or Shopping, also check product or asset-group contribution where the account structure allows it. If the useful cut is not available in a standard report, an Agency user can run a read-only custom query.

Review the current bidding strategy, target ROAS, budget, conversion goals, networks, locations, schedules, devices, and campaign status. Then line those settings up with change history.

A ROAS decline that begins immediately after a target, budget, conversion action, value rule, campaign composition, or landing-page change deserves investigation. Timing is not proof, but it gives you a testable hypothesis.

Do not assume the target ROAS should match the current result. Google notes that setting the target too high can restrict traffic. It also recommends evaluating ROAS outside the recent conversion-delay period. Diagnose the value and volume trade-off before editing the target.

Google Ads reports the value sent into its conversion system. That is not automatically revenue collected, gross margin, pipeline value, or profit.

Before making a large budget decision, reconcile the platform with the system of record:

If reported ROAS falls while profit improves, the platform metric may be incomplete. If reported ROAS rises while low-quality leads or refunded orders grow, the apparent improvement may be false comfort.

Use this as the starting brief. Replace the placeholders and keep the first pass read-only.

A useful answer should include the periods used, the conversion definition, an account-level driver summary, campaign contributors, segment evidence, settings or change-history evidence, missing business data, and a recommended next check. If a required report is partial, the answer should say so.

Review where CPC and spend increased, then test whether the extra traffic is commercially useful. Tighten poor query or location mix, improve relevance, or reconsider budget allocation where evidence supports it. Do not cut bids account-wide from a blended average.

Separate intent deterioration from experience deterioration. Compare queries, ads, devices, locations, and landing pages. Check tracking before rewriting ads or replacing a page. A bid change will not repair a broken form.

Check product, order, lead, and customer mix. Confirm that transaction values or offline values are arriving correctly. For lead generation, use CRM stages rather than assuming every form fill has equal value.

Repair the conversion action, value capture, or import process before optimizing bids to the distorted data. Document the correction date and allow enough time for a new mature comparison window.

A deliberate move into prospecting can lower short-term ROAS while increasing total value. Judge it against incremental profit, payback period, and the campaign’s role—not the brand campaign’s ROAS.

HireOtto can pull campaign reports for two date ranges with cost, clicks, conversions, conversion value, average CPC, and cost per conversion, then help calculate ROAS, conversion rate, and value per conversion. It can segment reports by day and retrieve supporting campaign, ad group, keyword, search-term, geographic, device, impression-share, and conversion-action evidence.

It can also inspect campaign settings—including bidding strategy, target ROAS, budgets, networks, locations, schedules, and devices—and review recent account changes. The detailed field-level change timeline is limited to the previous 30 days. Agency users can use the read-only weekly audit or a custom query when the standard reports do not provide the necessary cut.

Keep the first pass read-only. HireOtto can support approved changes to supported budgets, bidding settings, keywords, negatives, and other Google Ads entities, but the safer sequence is proposal → human review → narrow write → live read-back. It cannot determine profit, refund rates, CRM lead quality, inventory problems, or landing-page experience from Google Ads data alone. Supply those facts separately.

Start with the HireOtto reporting guide, verify current capabilities and change-history limits in the Google Ads MCP tools reference, review the weekly optimization audit when an Agency-level diagnostic helps, and connect Google Ads through the quickstart.

A good ROAS clears your economic threshold after gross margin, fulfillment, sales cost, refunds, and customer quality. A 400% ROAS can be excellent for one business and unprofitable for another. Build the threshold from unit economics, not an industry average.

Yes. Google explains that costs may already be present while some resulting conversions or values have not arrived, which can make recent ROAS look deflated. Use mature windows.

Spend may have risen, average value per conversion may have fallen, or the account may be counting a different mix of conversion actions. Compare cost and value per conversion, then inspect the conversion definition.

Not automatically. A higher target can reduce traffic, and changing the target does not repair missing values, weak conversion rate, or poor traffic mix. Diagnose the driver and business trade-off first.

AI can calculate drivers, rank contributors, and connect changes to timing. It usually cannot prove auction behavior, customer quality, margins, refunds, or website problems from Google Ads data alone. The useful output separates facts, inferences, and missing evidence.