← BlogGuide

How to Detect Sudden Changes in Ad Performance

Spend, traffic and conversion data in Google Ads and Meta Ads campaigns can shift unexpectedly. During November sales, heavier competition, changing user behavior and budget movements make performance analysis even harder. Comparing metrics like CPC, CPA, ROAS and conversion rate against past periods helps separate normal fluctuations from problems that need action. With AdsLuma's Sentinel Findings and Luma AI, you can investigate sudden campaign changes, explore likely causes and decide on actions for issues that hurt budget efficiency, so you can monitor ad performance more consistently and respond to important changes in time.

16 Nov 2026·9 min read
How to Detect Sudden Changes in Ad Performance

In Google Ads and Meta Ads campaigns, spend, traffic and conversion results can move outside their expected levels in a short time. During November sales in particular, rising ad costs and shifting shopping behavior make it harder to tell normal fluctuations from genuine performance problems.

To detect sudden changes in ad performance, you need to look at more than daily percentage differences: the campaign's historical results, conversion volume and current conditions should be reviewed too. This data shows which changes stay within the normal performance range and which campaigns deserve a closer look.

How do you detect sudden changes in ad performance?

Sudden changes in ad performance are detected by comparing metrics such as spend, clicks, conversions, CPA and ROAS against an expected performance range built from historical data. During anomaly analysis, you examine the size and duration of the change and which campaigns it affects. The findings then help you judge whether there is a problem in tracking, budget or the conversion process that requires intervention.

What Is an Ad Performance Anomaly?

When a campaign moves outside the range of change considered normal based on its past performance, it is defined as an ad performance anomaly. Examples include spend rising unexpectedly, conversions dropping, or traffic costs drifting away from their usual levels.

For a campaign that spends an average of 5,000 TRY per day, a day with 5,500 TRY in spend may be consistent with similar past variations. If the same campaign starts spending 9,000 TRY without any budget change, that is a signal worth investigating.

User demand, auction competition and the campaign calendar all affect ad results. In a campaign that gets two conversions a day, a single missing sale can create a large percentage difference in CPA. When evaluating an anomaly, data volume, the historical performance range and how long the change persists should be considered together.

Which Metric Changes Signal an Anomaly?

When you investigate the source of unexpected movements in campaigns, the relationships between metrics point the way. Because a spend increase, a conversion loss and a ROAS change can each have different causes, it helps to evaluate each one alongside related indicators.

  • CPC increase with a stable conversion rate: Higher click costs can put pressure on CPA. Check auction costs and traffic distribution.

  • Stable CPC with a declining conversion rate: The rate at which traffic turns into sales or leads has changed. Landing pages, traffic quality and tracking checks take priority.

  • Rising spend with an unchanged number of conversions: CPA is going up. Investigate the campaigns where the spend increase is concentrated.

  • Stable sales volume with declining ROAS: Compare the revenue attributed to ads, average order value and changes in spend.

These relationships narrow down the likely source of the change. To compare results over longer periods, you can also use the methods described in ad performance comparison.

How Do You Detect Anomalies in Ad Campaigns?

To detect sudden changes, you first build a reference that represents the campaign's normal performance. Current results are then compared against this reference to identify deviations worth investigating.

1. Build the Expected Performance Range from Historical Data

Using the campaign's past results, you can compare spend, traffic and conversion values for similar days of the week. In accounts with enough historical data, methods such as the median and the interquartile range can be used to set normal performance limits.

For campaigns with low data volume, longer comparison periods may be preferable. For special campaign days such as November sales, it makes more sense to build a separate reference based on the results of previous promotions, the current discount conditions and the planned budget. Budget allocation and seasonal differences for the November period are explained in detail in November sales ad budget planning.

2. Set Alert Thresholds Based on Data Volume

For automated anomaly detection, you can use fixed alert thresholds or thresholds that adapt to past performance. For an initial setup, you might define a review condition such as CPA rising 30% above its expected level. This rate should be tested against the campaign's historical results and updated when necessary.

In campaigns with few conversions, small numerical differences can produce large percentage changes. Conditions such as a minimum volume of clicks, spend or conversions help reduce unnecessary alerts.

For campaigns that keep spending without recording conversions, traffic and budget consumption can be checked separately. The duration of the change and its potential financial impact also determine the priority of the alert.

3. Trace the Source of the Change Through Campaign Breakdowns

A performance change seen across the account may come from a single ad group, device or landing page. The investigation starts by identifying when the change began. Then you compare device, location and traffic results for the relevant campaigns.

Using the change history in Google Ads, you can check budget, targeting and bidding strategy updates made during the same period. The platform's Explanations feature helps you investigate the possible causes of performance movements in supported campaigns.

Conversion data from recent dates may be subject to reporting delays. When interpreting same-day CPA and ROAS results, the completeness of the data should also be taken into account.

Example of a Sudden Performance Change During November Sales

Consider an e-commerce brand running a Google Ads campaign for 11.11 sales. Suppose the business has used past sale results to set a performance expectation for the 10:00–14:00 window.

The data below is fictional. The order figures represent the site's total sales records and are evaluated separately from the conversions attributed to Google Ads.

MetricExpected LevelActual
Google Ads spend12,000 TRY12,000 TRY
Google Ads clicks600600
Site-wide orders46–5230
Mobile orders30–3414
Desktop orders16–1816

Spend and ad clicks are in line with expectations. Site-wide orders, however, fall below the reference range built from historical data. The device breakdown shows that desktop sales are on track, while the shortfall is concentrated in mobile orders.

This result calls for an investigation of the mobile purchase experience. To understand whether the problem comes from Google Ads, you should also look at ad traffic quality, organic visitors and changes in other channels.

How Do You Find the Cause of the Change in Mobile Orders?

As a first step, you can check the rate at which mobile users reach the checkout page, whether discount codes work and whether product stock is up to date. The distribution of mobile traffic sources over the same period should also be compared.

Suppose that in this fictional review a technical error in the payment step is confirmed. Instead of changing the ad budget, fixing the problem in the purchase process becomes the priority.

To evaluate the impact on ad performance, you should also wait for Google Ads conversions to be fully reported. Suppose the expectation set in advance for this campaign was 40 conversions and 60,000 TRY in ad-attributed revenue. Once conversion reports are complete, 24 conversions and 36,000 TRY in revenue are recorded.

With 12,000 TRY in spend, the target CPA is 300 TRY and the target ROAS is 5. In the actual results, CPA rises to 500 TRY and ROAS falls to 3. Both metrics show that the campaign has deviated from its targeted cost and revenue performance.

After the technical issue is fixed, re-examining site orders and the conversions attributed to the ad platform helps you evaluate the results of the intervention.

Can Anomalies in Ad Performance Be Monitored Automatically?

Automated performance alerts compare campaign results against defined conditions or expectations based on historical data, helping you spot unusual movements early. They make it easier to prioritize what to review, especially in accounts with large budgets and many campaigns.

In GA4, the Analytics Insights feature can be used to identify unexpected changes in your data. With Custom Insights, you can define hourly, daily, weekly or monthly conditions. Hourly evaluation is available for web data.

Data collection delays, low conversion volume and planned campaign changes can also trigger alerts. Findings flagged by automated systems should be verified against the relevant data sources and campaign conditions.

How Do You Analyze Sudden Ad Performance Changes with AdsLuma?

In AdsLuma, Google Ads and Meta Ads data can be tracked on a shared dashboard. Ad performance analysis