Confidence level indicators

The Budget Optimiser uses campaign confidence signals to explain how much trust can be placed in each campaign’s recommended budget changes.

These signals are not a judgement of whether a campaign is “good” or “bad”. Instead, they describe how much reliable evidence is available to model the relationship between spend and performance.

Each campaign receives an overall confidence level supported by several underlying individual indicators. Together, these indicators help answer questions such as:

• Was there enough recent campaign activity?

• Did spend vary enough for the model to learn from it?

• Did historical data show a clear relationship between spend and results?

• Were there enough conversions to trust the estimated business value?

A high confidence level means the model had stronger evidence to work with. A lower confidence level means recommendations should be interpreted more cautiously, or that the campaign may require more data before it can be optimised reliably.

The Budget Optimiser evaluates four key indicators to assess the reliability of a campaign’s optimisation model. These indicators are combined into a single overall confidence level shown in Alvie.

Campaigns are more likely to receive a high confidence level when:

• the modelled curve closely matches historical performance,

• the campaign has enough recent activity,

• spend changes provide clear evidence of performance response,

• and conversion activity is strong enough to support value estimation.

The following indicators directly influence the overall confidence level and help explain why confidence is low, medium, or high.

Fit quality measures how closely the modelled curve matches the campaign’s historical performance.

The Budget Optimiser models how performance results change as spend changes. When historical performance follows a consistent and predictable pattern, fit quality is higher. When performance is noisy or inconsistent, fit quality is lower.

Fit quality may be lower when:

• clicks or results fluctuate for reasons unrelated to spend,

• the campaign changed significantly during the look-back period,

• unusual events occurred, such as promotions, outages, budget shocks, or tracking issues,

• the campaign lacks enough stable historical data,

• spend changed but performance did not respond consistently.

Ways to improve fit quality:

• change look-back windows or allow more time for the campaign to gather more stable historical data,

• avoid making multiple major changes simultaneously,

• investigate unusual days or tracking issues,

• ensure campaigns are grouped and tracked correctly.

Data sufficiency measures whether the campaign had enough recent activity for reliable modelling.

Campaigns with activity on most days provide a stronger basis for modelling. Campaigns that are new, paused, or only active intermittently provide less evidence.

Data sufficiency may be lower when:

• the campaign is newly launched,

• the campaign was paused during the look-back period,

• spend levels were too low to generate regular traffic,

• platform data is missing for some days,

• the campaign only receives sporadic clicks.

Ways to improve data sufficiency:

• maintain activity on more days within the look-back window,

• avoid long pauses when optimisation is expected,

• ensure platform data is complete and available,

• maintain enough activity to generate regular traffic.

Spend variation measures whether campaign spend changed enough for the model to learn how performance responds at different spend levels.

If spend remains almost identical every day, it becomes difficult to estimate the impact of increasing or decreasing budget. Natural spend variation provides the model with more evidence about performance behaviour.

For search campaigns, the Budget Optimiser also accounts for changes in available demand to avoid confusing fluctuations in search volume with intentional spend variation.

Spend variation may be lower when:

• daily spend remains nearly flat,

• budgets are tightly constrained,

• the campaign spends similar amounts every day,

• overall activity is low.

Ways to improve spend variation:

• allow some budget flexibility over time,

• collect data across different spend levels,

• avoid maintaining exactly the same spend level throughout the look-back period.

Conversion volume measures whether there are enough conversions to reliably estimate business value.

While the optimisation curve is primarily based on spend and clicks, optimisation decisions depend on conversions or revenue. When conversion volume is very low, conversion rate and revenue estimates can become unstable.

For campaigns using attribution, this indicator is based on the conversion credit assigned to the campaign, not only platform reported conversions.

Conversion volume may be lower when:

• the campaign generated very few conversions,

• the selected conversion action rarely occurs,

• attribution assigns limited conversion credit to the campaign,

• conversion tracking is incomplete.

Ways to improve conversion volume:

• collect more conversion history,

• verify that conversion tracking is working correctly,

• review whether the selected conversion action is appropriate,

• consider optimising at a broader grouping level if campaign conversion volume is too low.