Why is my ROAS (Return on Ad Spend) decreasing?

Alexandre Airvault
January 13, 2026

ROAS is decreasing: first confirm you’re looking at the same “ROAS” as before

ROAS is a ratio, so it drops for only two reasons

Return on ad spend is simply conversion value divided by cost. That means your ROAS can fall even if your business is stable, as long as either (a) reported conversion value goes down, (b) ad spend goes up, or (c) both happen at once. The fastest wins usually come from proving which side of the fraction moved first: tracking/value/reporting (the numerator) or auction costs/traffic mix (the denominator).

Make sure your reporting columns still match what your bidding is optimizing for

A very common “mystery ROAS drop” is actually a reporting mismatch. In many accounts, the Conversions and Conversion value columns only include the conversion actions currently set as primary for optimization, while secondary actions show up in All conv. and All conv. value. If someone recently changed which actions are primary (or changed campaign conversion goals), your reported conversion value can drop overnight even when real sales didn’t.

Exclude the conversion delay period when judging “recent” ROAS

If you evaluate ROAS including the most recent days where conversions are still “in flight,” you’ll almost always understate performance—especially for longer consideration cycles. This matters even more right after you change targets, budgets, or conversion setup. A practical approach is to compare two equivalent windows that both sit far enough in the past to be “fully baked” (for example, look at 14–28 days ago versus the prior 14-day window, rather than yesterday versus last week).

The most common reasons ROAS decreases (and how to diagnose each one)

1) Your conversion tracking or conversion value collection is incomplete

If fewer conversions are being captured—or if conversion events are firing without complete value/user-provided data—ROAS will fall even when traffic quality is unchanged. This can happen after site releases, checkout changes, consent implementations, tag refactors, or when only some conversion pages are correctly instrumented.

If you rely on enhanced conversions, confirm it’s transmitting properly on every page where the conversion tag fires, and use the platform diagnostics after the system has had time to process the new signals (it’s normal to need a few days before diagnostics fully populate). If the tag fires but key parameters are empty on some pages, you’ll often see a slow bleed in reported performance rather than an obvious “tracking broke” outage.

2) You changed conversion goals (primary vs secondary) and your ROAS “numerator” shrank

ROAS is only as good as the conversion actions included in the conversion value column you’re using. When teams add new actions, demote purchase actions to secondary, switch goals at the campaign level, or split goals by campaign type, you can unintentionally remove meaningful value from the primary columns. The result looks like declining ROAS, but it’s really a measurement scope change.

This also comes up when you begin including additional omnichannel actions (like store-related conversions) in the set of conversions used for optimization. Historical reporting won’t retroactively “rebuild” the past using your new setting, so trend lines can look jagged right at the change date.

3) You shortened conversion windows, so late conversions stopped counting

If your conversion window was reduced (for example, from 30 days to 7 days), conversions that happen after that window simply no longer appear in reports. That can instantly reduce recorded conversion value and make ROAS look worse—especially in higher-AOV, longer-cycle categories. Window changes only apply going forward, which creates a clear “break” in trend lines starting on the change date.

4) You changed attribution settings/models and the credit shifted away from the campaigns you’re watching

Attribution changes can move conversion value between campaigns without changing total sales. If you recently adjusted attribution model settings or started using different attribution reporting columns, some campaigns will “lose” credit and others will “gain” it. That’s not inherently bad—but if you judge ROAS at the campaign level without accounting for the shift, it can look like a real decline.

Also watch out for campaign-type-specific reporting columns. Certain campaign types offer alternative “platform comparable” style columns that intentionally use different methodologies than the standard conversion columns. Mixing these views when comparing against Search, Shopping, or Performance Max can create false ROAS conclusions.

5) Smart Bidding is reacting to changes faster than your conversions can be observed

When you change ROAS targets, budgets, conversion goals, or the conversion values themselves, the bidding system can start optimizing toward the new goal quickly, but performance measurement lags because conversions lag. In practice, you should avoid stacking multiple target changes inside a single conversion cycle (the typical time it takes clicks to convert). Doing so gives the system moving goalposts before it can “see” the outcome of the previous change.

Another overlooked trigger: if you just started reporting conversion value (or changed the way values are sent), it takes time for value signals to come through consistently enough for value-based bidding to behave predictably. Expect a stabilization period before you treat the new ROAS as “truth.”

6) Your conversion values became less representative of profit (customer mix shifts)

ROAS can decrease even while conversions hold steady if average conversion value drops. Common causes include heavier discounting, selling lower-margin items, higher refund rates, or a surge of returning customers who buy smaller baskets. If you’re actively optimizing toward new customer acquisition, you may also be trading short-term ROAS for higher lifetime value—especially if your reporting doesn’t add incremental value for new customers in a structured way.

If you use conversion value rules to weight higher-value segments (by location, device, or audience), a configuration change can materially alter the value signal. Value rules are powerful, but they’re also easy to misapply if the “when/then” logic doesn’t match how your business actually values customers.

7) Post-purchase adjustments (returns/cancellations) are pulling value down

If you restate or retract conversion values to account for returns, partial refunds, or cancellations, reported conversion value will drop—often with a time delay. That can make ROAS look like it’s “randomly” falling days or weeks after a promotion. This is usually a good sign (your measurement is more honest), but it does require changing how you evaluate recent performance and how aggressively you set ROAS targets.

8) Costs rose due to auction pressure, budgets, or targeting changes

Sometimes ROAS is fine—the denominator moved. If cost per click rises because of competition, seasonality, broader match expansion, looser targeting, or more aggressive budget pacing, your spend can grow faster than conversion value. This is especially common after major structural edits (new campaign types, new bidding strategy, changes to locations/languages/schedules, audience expansion, or large creative refreshes) because traffic mix shifts before your site and funnel “catch up.”

A systematic fix: the fastest way to recover ROAS without guesswork

Step 1: Identify whether the drop is “value down” or “cost up”

  • Compare conversion value, cost, and conversion value/cost over two equal time ranges that exclude the newest conversion-delay days.
  • Check whether value per conversion fell (basket size/mix issue) or whether conversion rate fell (traffic/landing/funnel issue).
  • Confirm whether the set of conversions included in reporting changed (primary vs secondary, campaign goals, newly included actions).

Step 2: Audit tracking and value integrity (the “ROAS numerator” health check)

In mature accounts, most ROAS declines trace back to measurement integrity. Validate that your purchase conversion fires on the correct confirmation state, fires once per order, and consistently includes value and currency. If you use enhanced conversions, ensure it’s implemented across all relevant conversion pages and that diagnostics show healthy coverage after the system has had time to process recent changes.

If you recently changed conversion windows, attribution settings, or which conversions are primary, mark the exact change date and evaluate performance before/after using consistent definitions—otherwise you’ll end up “optimizing” against a reporting artifact.

Step 3: Stabilize Smart Bidding before making more changes

If you’re on value-based automated bidding, treat ROAS target edits like surgery, not a dial you spin daily. Make one meaningful change, then wait at least one to two conversion cycles to judge impact. When you do set a ROAS target, ground it in recent historical conversion value/cost from a period that excludes conversion delay, and avoid setting a target higher than what the account can realistically sustain, because overly aggressive targets often reduce traffic volume and can starve the system of the data it needs to improve.

Step 4: Improve value quality, not just volume

If the core issue is lower value per conversion, start by aligning the value signal to what the business truly wants. That may mean refining conversion values (so different purchases or customer types are valued appropriately), using value rules carefully where they genuinely reflect incremental value, and ensuring returns/cancellations are handled consistently so you’re not optimizing toward revenue that won’t be kept.

Step 5: Control the “cost side” with tighter intent and cleaner traffic

If the core issue is cost rising faster than value, focus on tightening query quality and traffic mix. In practice, that usually means reducing wasted spend from low-intent queries/placements, aligning ad messaging to higher-intent segments, and ensuring landing pages match the promise of the ad so conversion rate doesn’t erode as you scale. When budgets are constrained or frequently changed, expect more volatility; stable budgets and fewer structural edits typically produce more stable ROAS because the system can learn and allocate efficiently.

Let AI handle
the Google Ads grunt work

Try now for free
Area What to check Why ROAS can drop How to diagnose in Google Ads Relevant Google Ads docs
ROAS definition & timeframes Confirm you’re looking at the same ROAS and date ranges as before. ROAS is conversion value ÷ cost, so it falls when reported value goes down, costs go up, or both. Comparing different “flavors” of ROAS or including days with incomplete conversion data makes it look like performance dropped when it didn’t.
  • Use identical date ranges and compare conversion value, cost, and conversion value / cost across two equal “fully baked” windows (for example, 14–28 days ago vs the prior 14 days).
  • Exclude the most recent days that are still within your typical conversion delay before judging ROAS trends.
Conversion tracking & value collection Check if all conversions and their values are still being recorded correctly. If tags stop firing, fire multiple times, or fire without value/currency or user data, reported conversion value drops even when traffic quality is unchanged.
  • In Tools > Conversions, review each key action’s Status, Category, Value, and Count settings.
  • Use tag diagnostics and tag assistant to confirm the purchase tag fires once per order on the correct confirmation state and always passes value and currency.
  • If you use enhanced conversions, verify implementation on every relevant page and wait long enough for diagnostics and impact data to populate.
Primary vs secondary conversions & goals Confirm which conversion actions are set as primary, which are secondary, and which goals each campaign is optimizing toward. Switching purchase actions to secondary, changing campaign conversion goals, or newly including/excluding omnichannel actions can remove value from the Conversions and Conversion value columns overnight, making ROAS appear to fall even when true sales are unchanged.
  • In Tools > Conversions, review each action’s Primary/Secondary setting and which conversion goal it belongs to.
  • On the campaign “Settings” page, check the campaign’s Conversion goals and whether they “use account default” or custom goals.
  • Compare trends using Conversions / Conversion value vs All conv. / All conv. value to see if the drop is only in primary metrics.
Conversion windows Review how long each conversion action’s window is and whether it was shortened recently. Reducing a conversion window (for example from 30 days to 7 days) stops counting conversions that occur after the new window, causing a step change down in reported conversion value and ROAS, especially for longer purchase cycles.
  • In Tools > Conversions, open each action and check the configured Click-through conversion window and any view-through window.
  • Overlay change history to identify the exact date when a window was modified and compare performance before/after using consistent windows.
Attribution models & reporting views Confirm whether attribution models or attribution reporting columns changed. Shifting from one attribution model to another, or using alternative attribution-style columns for certain campaign types, moves conversion value between campaigns without changing total sales. Some campaigns “lose” credit and show lower ROAS purely due to the model change.
  • In each key conversion action’s settings, check the Attribution model and whether it was recently changed (for example, to data-driven or last click).
  • Use attribution reports and the model comparison report to see how different models redistribute value across campaigns.
  • Avoid mixing standard conversion columns with “platform comparable” or campaign-type-specific columns when comparing ROAS across campaign types.
Smart Bidding changes vs conversion lag Look at recent changes to ROAS targets, budgets, conversion goals, or value definitions versus your typical conversion delay. The bidding system can adjust bids immediately based on new targets or value signals, but conversions may take days or weeks to materialize. Stacking multiple target or goal changes within a single conversion cycle creates noisy, misleading short-term ROAS trends.
  • Use Change history to map edits to ROAS targets, bid strategies, budgets, and conversion settings.
  • Compare performance in windows that start after each major change and extend beyond your typical conversion delay.
  • Let Smart Bidding run for at least one to two full conversion cycles after a meaningful target change before judging ROAS.
Average order value & customer mix Check if average conversion value or profit per conversion has fallen even when conversion count is steady. More discounted orders, lower-margin products, higher refund rates, or a shift toward smaller repeat purchases reduce average conversion value. ROAS drops even if the number of conversions and traffic quality are stable.
  • Compare conversion value / conversion over time at campaign, product, and audience levels.
  • Align your on-site pricing, discounting, and product mix data to match ROAS trends.
  • Ensure conversion values and any value rules reflect true incremental value for higher‑value customers or orders.
Returns, cancellations, and value adjustments Review how post-purchase events (returns, partial refunds, cancellations) are fed back into Google Ads. If you restate or retract conversion values to account for returns and cancellations, reported conversion value for past days will decrease and historical ROAS will drift down, especially after promotions.
  • Check whether you’re importing adjusted offline conversions or sending updated values through the tag or API.
  • Segment by conversion date vs adjustment date (where available) to understand the lag between purchase and value restatement.
  • When evaluating recent performance or setting ROAS targets, factor in the expected rate and timing of returns.
Auction pressure, budgets, and targeting Check whether cost per click, impression share, or traffic mix changed around the time ROAS began to fall. ROAS can decline when the denominator (cost) rises faster than conversion value due to auction competition, broader match or audience expansion, relaxed location/language settings, or aggressive budget increases.
  • Compare CPC, cost, and conversion rate across time ranges and segments (device, location, audience, search terms, placements).
  • Review changes in campaign types, bid strategies, match types, audience expansion, and budget levels using change history.
  • Trim low‑intent queries/placements and align ad copy and landing pages to the highest‑intent segments to protect conversion rate as you scale.
Systematic ROAS recovery – measurement Step 1–2: Separate “value down” vs “cost up” and audit measurement integrity. Most “mysterious” ROAS drops trace back to measurement changes or inconsistencies rather than true demand shifts. Fixing tracking and value inputs often restores ROAS without major bidding overhauls.
  • Compare value, cost, and ROAS over equal, fully delayed windows to see which side of the ratio moved first.
  • Validate that purchase conversions fire once, on the correct state, and consistently include accurate value and currency.
  • Document the exact dates for any changes to conversion windows, attribution models, primary/secondary status, or goals, and analyze performance using pre‑/post windows with consistent definitions.
Systematic ROAS recovery – bidding & value strategy Step 3–5: Stabilize Smart Bidding, improve value quality, and tighten traffic. Frequent target tweaks, misaligned value signals, and loose targeting make ROAS volatile. Grounding targets in realistic recent data, sending high‑quality value signals, and improving traffic intent stabilizes and improves ROAS.
  • Make deliberate, infrequent ROAS target edits and wait at least one to two conversion cycles before further changes.
  • Refine conversion values and value rules so they reflect true incremental value (for example, higher weights for new customers or high‑margin products).
  • Reduce wasted spend by tightening keyword and audience intent, improving ad relevance, and ensuring landing pages align with the ad promise.

If your ROAS is decreasing, it usually comes down to a shift in one side of the ratio (conversion value down, costs up, or both) or a reporting/measurement change that makes performance look worse than it really is—common culprits include incomplete conversion value tracking, primary vs. secondary conversion or goal changes, shortened conversion windows, attribution model updates, Smart Bidding reacting faster than your conversion lag, changes in average order value (discounting, product mix, refunds), or higher auction pressure and broader targeting driving up CPCs. If you want a structured way to spot what changed and what to do next without living in spreadsheets and change history, Blobr connects to your Google Ads account and continuously analyzes performance, then turns best practices into concrete recommendations, with specialized AI agents that can help tighten wasted traffic, align keywords to the right landing pages (Keyword Landing Optimizer), and refresh underperforming ad assets (Headlines Enhancer) while you stay in control of what gets applied.

ROAS is decreasing: first confirm you’re looking at the same “ROAS” as before

ROAS is a ratio, so it drops for only two reasons

Return on ad spend is simply conversion value divided by cost. That means your ROAS can fall even if your business is stable, as long as either (a) reported conversion value goes down, (b) ad spend goes up, or (c) both happen at once. The fastest wins usually come from proving which side of the fraction moved first: tracking/value/reporting (the numerator) or auction costs/traffic mix (the denominator).

Make sure your reporting columns still match what your bidding is optimizing for

A very common “mystery ROAS drop” is actually a reporting mismatch. In many accounts, the Conversions and Conversion value columns only include the conversion actions currently set as primary for optimization, while secondary actions show up in All conv. and All conv. value. If someone recently changed which actions are primary (or changed campaign conversion goals), your reported conversion value can drop overnight even when real sales didn’t.

Exclude the conversion delay period when judging “recent” ROAS

If you evaluate ROAS including the most recent days where conversions are still “in flight,” you’ll almost always understate performance—especially for longer consideration cycles. This matters even more right after you change targets, budgets, or conversion setup. A practical approach is to compare two equivalent windows that both sit far enough in the past to be “fully baked” (for example, look at 14–28 days ago versus the prior 14-day window, rather than yesterday versus last week).

The most common reasons ROAS decreases (and how to diagnose each one)

1) Your conversion tracking or conversion value collection is incomplete

If fewer conversions are being captured—or if conversion events are firing without complete value/user-provided data—ROAS will fall even when traffic quality is unchanged. This can happen after site releases, checkout changes, consent implementations, tag refactors, or when only some conversion pages are correctly instrumented.

If you rely on enhanced conversions, confirm it’s transmitting properly on every page where the conversion tag fires, and use the platform diagnostics after the system has had time to process the new signals (it’s normal to need a few days before diagnostics fully populate). If the tag fires but key parameters are empty on some pages, you’ll often see a slow bleed in reported performance rather than an obvious “tracking broke” outage.

2) You changed conversion goals (primary vs secondary) and your ROAS “numerator” shrank

ROAS is only as good as the conversion actions included in the conversion value column you’re using. When teams add new actions, demote purchase actions to secondary, switch goals at the campaign level, or split goals by campaign type, you can unintentionally remove meaningful value from the primary columns. The result looks like declining ROAS, but it’s really a measurement scope change.

This also comes up when you begin including additional omnichannel actions (like store-related conversions) in the set of conversions used for optimization. Historical reporting won’t retroactively “rebuild” the past using your new setting, so trend lines can look jagged right at the change date.

3) You shortened conversion windows, so late conversions stopped counting

If your conversion window was reduced (for example, from 30 days to 7 days), conversions that happen after that window simply no longer appear in reports. That can instantly reduce recorded conversion value and make ROAS look worse—especially in higher-AOV, longer-cycle categories. Window changes only apply going forward, which creates a clear “break” in trend lines starting on the change date.

4) You changed attribution settings/models and the credit shifted away from the campaigns you’re watching

Attribution changes can move conversion value between campaigns without changing total sales. If you recently adjusted attribution model settings or started using different attribution reporting columns, some campaigns will “lose” credit and others will “gain” it. That’s not inherently bad—but if you judge ROAS at the campaign level without accounting for the shift, it can look like a real decline.

Also watch out for campaign-type-specific reporting columns. Certain campaign types offer alternative “platform comparable” style columns that intentionally use different methodologies than the standard conversion columns. Mixing these views when comparing against Search, Shopping, or Performance Max can create false ROAS conclusions.

5) Smart Bidding is reacting to changes faster than your conversions can be observed

When you change ROAS targets, budgets, conversion goals, or the conversion values themselves, the bidding system can start optimizing toward the new goal quickly, but performance measurement lags because conversions lag. In practice, you should avoid stacking multiple target changes inside a single conversion cycle (the typical time it takes clicks to convert). Doing so gives the system moving goalposts before it can “see” the outcome of the previous change.

Another overlooked trigger: if you just started reporting conversion value (or changed the way values are sent), it takes time for value signals to come through consistently enough for value-based bidding to behave predictably. Expect a stabilization period before you treat the new ROAS as “truth.”

6) Your conversion values became less representative of profit (customer mix shifts)

ROAS can decrease even while conversions hold steady if average conversion value drops. Common causes include heavier discounting, selling lower-margin items, higher refund rates, or a surge of returning customers who buy smaller baskets. If you’re actively optimizing toward new customer acquisition, you may also be trading short-term ROAS for higher lifetime value—especially if your reporting doesn’t add incremental value for new customers in a structured way.

If you use conversion value rules to weight higher-value segments (by location, device, or audience), a configuration change can materially alter the value signal. Value rules are powerful, but they’re also easy to misapply if the “when/then” logic doesn’t match how your business actually values customers.

7) Post-purchase adjustments (returns/cancellations) are pulling value down

If you restate or retract conversion values to account for returns, partial refunds, or cancellations, reported conversion value will drop—often with a time delay. That can make ROAS look like it’s “randomly” falling days or weeks after a promotion. This is usually a good sign (your measurement is more honest), but it does require changing how you evaluate recent performance and how aggressively you set ROAS targets.

8) Costs rose due to auction pressure, budgets, or targeting changes

Sometimes ROAS is fine—the denominator moved. If cost per click rises because of competition, seasonality, broader match expansion, looser targeting, or more aggressive budget pacing, your spend can grow faster than conversion value. This is especially common after major structural edits (new campaign types, new bidding strategy, changes to locations/languages/schedules, audience expansion, or large creative refreshes) because traffic mix shifts before your site and funnel “catch up.”

A systematic fix: the fastest way to recover ROAS without guesswork

Step 1: Identify whether the drop is “value down” or “cost up”

  • Compare conversion value, cost, and conversion value/cost over two equal time ranges that exclude the newest conversion-delay days.
  • Check whether value per conversion fell (basket size/mix issue) or whether conversion rate fell (traffic/landing/funnel issue).
  • Confirm whether the set of conversions included in reporting changed (primary vs secondary, campaign goals, newly included actions).

Step 2: Audit tracking and value integrity (the “ROAS numerator” health check)

In mature accounts, most ROAS declines trace back to measurement integrity. Validate that your purchase conversion fires on the correct confirmation state, fires once per order, and consistently includes value and currency. If you use enhanced conversions, ensure it’s implemented across all relevant conversion pages and that diagnostics show healthy coverage after the system has had time to process recent changes.

If you recently changed conversion windows, attribution settings, or which conversions are primary, mark the exact change date and evaluate performance before/after using consistent definitions—otherwise you’ll end up “optimizing” against a reporting artifact.

Step 3: Stabilize Smart Bidding before making more changes

If you’re on value-based automated bidding, treat ROAS target edits like surgery, not a dial you spin daily. Make one meaningful change, then wait at least one to two conversion cycles to judge impact. When you do set a ROAS target, ground it in recent historical conversion value/cost from a period that excludes conversion delay, and avoid setting a target higher than what the account can realistically sustain, because overly aggressive targets often reduce traffic volume and can starve the system of the data it needs to improve.

Step 4: Improve value quality, not just volume

If the core issue is lower value per conversion, start by aligning the value signal to what the business truly wants. That may mean refining conversion values (so different purchases or customer types are valued appropriately), using value rules carefully where they genuinely reflect incremental value, and ensuring returns/cancellations are handled consistently so you’re not optimizing toward revenue that won’t be kept.

Step 5: Control the “cost side” with tighter intent and cleaner traffic

If the core issue is cost rising faster than value, focus on tightening query quality and traffic mix. In practice, that usually means reducing wasted spend from low-intent queries/placements, aligning ad messaging to higher-intent segments, and ensuring landing pages match the promise of the ad so conversion rate doesn’t erode as you scale. When budgets are constrained or frequently changed, expect more volatility; stable budgets and fewer structural edits typically produce more stable ROAS because the system can learn and allocate efficiently.