Close-up of server racks in a data center highlighting modern technology infrastructure.
Photo by panumas nikhomkhai on Pexels

Bidding & Pacing

Part of Programmatic experimentation

Comparing bidding strategies with matched campaign conditions

Set up a bidding-strategy test with aligned targeting, budgets and measurement, then judge outcomes alongside spend and auction delivery.

This is an editorial pen name used by the publication, not a claim of an independently verified personal identity. Original editorial credit: Trading Desk.

Compare bidding strategies by splitting eligible traffic between two buys. The planned difference between them is the bidding rule. Judge the completed campaign outcome and what each strategy managed to buy.

Equal budgets do not guarantee equal spend, impressions or placement mix. Those differences may be part of the result. In Display & Video 360, the Experiments page runs this as an A/B experiment that compares insertion orders or line items, with bid strategies among the variables it can test.

Choose the strategies and decision

Name the baseline and variant exactly. For example, a campaign might compare a fixed CPM bid in the baseline line item with the automated strategy “Maximize conversions, while prioritizing a target KPI” in the variant. Define the completed enquiry, attribution settings, total cost basis and the improvement needed to justify a switch.

Check that both strategies are available for the intended line-item type and objective. Under the “while prioritizing a target KPI” option, meeting the performance goal takes priority over spending the budget, so an aggressive goal can leave the variant underspent. Treat underspend in that arm as part of the result, not as a setup fault.

Record the bid settings used in each arm, including any average CPM ceiling.

Pros and Cons of Automated vs Fixed Bidding in Australia

  • Automated: Maximize Conversions (Target KPI)Pro: Optimises for conversions within target KPI constraints. Con: May underspend if KPI is too aggressive.
  • Fixed CPMPro: Predictable spend and budget control. Con: Less responsive to performance variation across audiences.

Match the other conditions

Align dates, Australian geography, inventory routes or deals, formats, creative, landing page, audience rules, frequency controls and required exclusions. Use the same outcome definition and reporting cut-off. Set each arm's authorised budget in proportion to its experiment split and match pacing; a budget out of proportion to the audience split becomes another experiment variable.

To set up the test, create an A/B experiment on the Experiments page. Choose the arms, set the audience split, pick the research goal, such as conversions or clicks, and choose a 95% or 90% confidence interval. Duplicate the baseline line item to build the variant so only the bid strategy differs.

Before launch, use the Diff tab to confirm that the bid strategy is the only difference between the arms. The audience split cannot be changed after the experiment starts, and the earliest start date is 24 hours after setup. Retain a dated configuration record.

Setting Up a Bidding Strategy A/B Test in Display & Video 360

  1. Define Baseline and VariantSet baseline as fixed CPM; variant as 'Maximize conversions, while prioritising target KPI'
  2. Match All Campaign ConditionsAlign audience rules, geography (Australia), inventory deals, creatives, landing page, frequency caps, and exclusions
  3. Verify Differences Only in Bid StrategyUse Diff tab to confirm only bid strategy differs between arms

Examine buying and outcomes together

Where reporting permits, compare incoming opportunities, bid responses, wins, actual spend, bought impressions and completed outcomes. Put counts beside rates and normalised values, which scale an arm’s actual result up to a 100% audience split so that arms with unequal splits can be compared.

A strategy that spends little may show an attractive cost per action from too few actions to guide a larger allocation. A strategy that spends quickly may buy more outcomes at an unacceptable cost.

Display & Video 360 Help includes troubleshooting for deals and line items. Check whether targeting or creative filtering, rather than bidding, is affecting delivery. A creative rejection or unequal exclusion means the comparison no longer tests bidding alone. Record the fault, its timing and any correction applied to both arms.

Allow for automated learning and delayed outcomes while investigating serving faults promptly. Recent conversion figures can also change. Set the review window for the campaign's actual outcome and label an early read provisional.

Call a winner only when the confidence interval for the difference excludes no change and the p-value is low; a high p-value means the result may have occurred by chance. Limited reach, such as deal inventory, can produce wide confidence intervals.

Adopt a strategy only when its mature outcome and spending trade-off meet the pre-agreed rule under a valid comparison. If an arm lacks sufficient eligible delivery, report that deliverability finding separately from an inconclusive effectiveness comparison.

Further reading: Experiments overview Stay organized with collections Save and categorize content based on your preferences. (developers.google.com).

More from Bidding & Pacing