All about: Budget Optimization
Budget Optimization automatically sets a daily budget for each active experiment in a budget group, redistributing the group's budget toward the experiments that perform best against the metric you choose. This article explains how to turn it on, how the budget is split each time it runs, how new campaigns are handled while they ramp, and every optimization formula available.
Note: Budget Optimization applies a minimum daily budget of $10 for all active experiments, which may cause over-pacing in some situations. Avoid this by applying the recommendations made by Budget Control and getting to the Economy or Optimal state.
Note: Budget Optimization runs every 48 hours. Additionally, whenever big changes are made to budget groups (for example, launching or pausing a lot of experiments), Budget Optimization automatically runs soon after.
Enable Budget Optimization
Configure Budget Optimization to apply daily budgets based on the optimization formula.
- Go to the Budget Groups page.
- Click the budget group name.
- Toggle Budget Optimization to enable the feature.
How the budget is split
Each time Budget Optimization runs, it takes the budget group's total daily budget and divides it across the experiments that are currently active, in two steps:
- Every active experiment receives a base amount first so nothing is starved. Each experiment gets a minimum daily budget of $10 if the average daily budget per experiment is less than $51, or $25 if it's more than $51.
- The budget left over after the base amounts is distributed by performance. With Performance & Pacing, higher-performing experiments earn a larger share of the leftover. With Pacing Only, the leftover is split evenly and performance is ignored.
Performance is measured over the last 30 days that each experiment delivered, and suggested budgets are rounded to the nearest $5.
Note: If a group's total budget is too small for the number of experiments in it, every experiment falls back to the $10 minimum and no performance-based split happens (the Limited state). Raise the group's cap or reduce the number of active experiments so each one has enough budget to perform — this is what Budget Control's Economy / Optimal recommendations help you reach.
Pacing Only
Pacing Only ignores experiment performance and allocates the total daily budget equally across all active experiments.
For example, with a total daily budget of $400 split across four active experiments, each experiment receives $100 per day regardless of how it is performing:
| Experiment | Daily budget (Pacing Only) |
|---|---|
| Experiment A | $100 |
| Experiment B | $100 |
| Experiment C | $100 |
| Experiment D | $100 |
Performance & Pacing
Performance & Pacing calculates and assigns a daily budget for each experiment based on performance, so the total daily budget skews toward the experiments performing best on your chosen metric.
Note: Performance calculations only consider data for the last 30 days that the experiment was active.
Note: Each experiment receives a minimum daily budget of $10 if the average daily budget per experiment is less than $51, or $25 if the average daily budget per experiment is more than $51.
For example, with the same total daily budget of $400 allocated by CPL performance, the experiments with the strongest CPL receive more budget and the weakest receive less (the figures below are illustrative):
| Experiment | CPL performance | Daily budget (Performance & Pacing) |
|---|---|---|
| Experiment A | Best | $160 |
| Experiment B | Strong | $120 |
| Experiment C | Average | $80 |
| Experiment D | Weakest | $40 |
New campaigns and the ramp-up period
A brand-new campaign has no performance history yet, so on its own it would score near the bottom and could be pushed to the minimum daily budget before it ever had a chance to deliver. Budget Optimization protects new campaigns from that so they get a fair chance to ramp:
- A boosted starting budget. For its first days after launch, a newly launched campaign is given a higher base budget than an established one, so it has enough spend to start delivering and generating data.
- Treated as an average performer until it has data. Until a campaign builds up delivery history, the optimizer scores it as an average performer rather than a zero. It therefore receives an average-sized share of the group's daily budget while it ramps, instead of being crowded out by campaigns that already have a head start.
As the campaign accumulates delivery history (about 30 days of delivery), its real performance takes over and it is scored on its own results like any other experiment.
Note: This gives new campaigns a fair, non-starved share of the budget — it does not guarantee they will out-earn proven campaigns. While a campaign is still ramping, a safeguard keeps it from drawing more budget than the established, delivering experiments in the same group until it has enough data of its own.
How often the budget changes
Budget Optimization runs automatically about every 48 hours, and again soon after you make a big change to a group — such as launching or pausing a lot of experiments, or changing the cap or strategy. Between runs, the budget split stays as the last run set it.
This is why a change to a group's makeup may not be reflected in the split immediately. For example, if you pause several experiments, the paused ones are dropped from the split at the next optimization run, not the moment you pause them. Saving a change to the group prompts a fresh run.
Advanced Settings
Optimization Formula
Optimization Formula is used by Budget Optimization to score and proportionately distribute daily budget to active experiments in the budget group.
Note: Optimization Formula is only available for Lead Generation budget groups. Brand Awareness budget groups optimize using the CPC formula.
Tip: Start with CPL and change to MQL when most active experiments have one or more MQLs. Update MQL to influenced opportunities or triggered opportunities when enough opportunity data has accumulated.
Lead
Budget Optimization suggests daily budgets for active experiments with one or more leads based on lead volume and CPL data. Experiments generating the most leads for the lowest cost in the shortest duration receive the highest score and the most budget.
Note: CPL is available to budget groups with an optimization goal of lead generation.
MQL
Budget Optimization suggests daily budgets for active experiments with one or more leads based on lead volume, CPL, MQL volume and cpMQL data. Experiments generating the most MQLs for the lowest cost in the shortest duration receive the highest score and the most budget.
Note: MQL is available to budget groups with an optimization goal of lead generation. Budget Optimization reverts to CPL if no MQLs have been generated.
Triggered Opportunity
Budget Optimization suggests daily budgets for active experiments with one or more leads based on lead volume, CPL, MQL volume, cpMQL, triggered opportunity volume and cost per triggered opportunity data. Experiments generating the most triggered opportunities for the lowest cost in the shortest duration receive the highest score and the most budget.
Note: Triggered opportunity is available to budget groups with an optimization goal of lead generation. Budget Optimization reverts to using MQL if there are no triggered opportunities.
Influenced Opportunity
Budget Optimization suggests daily budgets for active experiments with one or more leads based on lead volume, CPL, MQL volume, cpMQL, influenced opportunity volume and cost per influenced opportunity data. Experiments generating the most influenced opportunities for the lowest cost in the shortest duration receive the highest score and the most budget.
Note: Influenced opportunity is available to budget groups with an optimization goal of lead generation. Budget Optimization reverts to using MQL if there are no influenced opportunities.
Custom KPIs
Prerequisite: Set up custom metrics so that you can optimize towards them.
Budget Optimization suggests daily budgets for active experiments with one or more leads based on lead volume, CPL and custom KPI data. Experiments generating the most custom KPIs for the lowest cost in the shortest duration receive the highest score and the most budget.
Note: Custom KPIs are available to budget groups with an optimization goal of lead generation. Budget Optimization reverts to using CPL if there is no custom KPI data.
CTR
Budget Optimization suggests daily budgets for active experiments based on CTR data. Experiments generating the most clicks with the best CTR in the shortest duration receive the highest score and the most budget.
Note: CTR is available to budget groups with an optimization goal of brand awareness.
CPC
Budget Optimization suggests daily budgets for active experiments based on click and CPC data. Experiments generating the most clicks for the lowest cost in the shortest duration receive the highest score and the most budget.
Note: CPC is available to budget groups with an optimization goal of brand awareness.
Frequently asked questions
I just launched a new campaign in a group. Will the optimizer starve it?
No. New campaigns are protected during their ramp-up. A recently launched campaign gets a boosted starting budget, and until it builds up delivery history the optimizer treats it as an average performer rather than a zero — so it receives a fair, average-sized share of the group's daily budget while it gathers data. After about 30 days of delivery it's scored on its own performance. See New campaigns and the ramp-up period.
I paused some experiments, but the optimizer is still splitting budget across all of them. Why?
Budget Optimization recalculates every 48 hours and soon after a big change to the group. Paused experiments are removed from the split at the next run, not the instant you pause them. To force a fresh split immediately, open the group and save it — that triggers a new optimization run.
How often does the budget change?
Every 48 hours, plus an extra run soon after you make a big change to the group (launching or pausing a lot of experiments, or changing the cap or strategy).
What's the smallest daily budget a campaign can get?
Each active experiment gets a base amount before performance is considered: $25 when the average daily budget per experiment is more than $51, or $10 when it's less. If the group's total budget is too small for the number of experiments, every experiment drops to the $10 minimum and the performance split is skipped. Raise the cap or reduce the number of active experiments so each has enough to perform.
What does Enable Booster do, and does it help new campaigns?
Booster is a group-level pacing feature: when a group is underspending against its budget, Booster puts the unused budget to work by temporarily raising the group's daily spend so the group stays on pace. It changes the group's total daily budget, not how that total is divided among individual campaigns — so it's separate from the new-campaign ramp-up described above.
What's the difference between auto-pause and Budget Optimization?
They're independent. Budget Optimization decides how the group's daily budget is split across active experiments. Auto-pause is a separate rule set that pauses experiments — for example when a monthly cap is reached, or when an experiment underperforms against your auto-pause rules. A group can use either, both, or neither.
What's the difference between Performance & Pacing and Pacing Only?
Performance & Pacing gives higher-performing experiments a larger share of the budget, balanced against even pacing. Pacing Only ignores performance and splits the daily budget evenly across all active experiments. Choose Pacing Only when you deliberately want an even spread rather than concentrating spend on the best performer.
Does the group's Target Metric / Goal change how budget is distributed?
No. The Optimize For setting (the optimization formula) is what drives the split. The Target Metric and Goal Value under Group Goal are only a benchmark for rating the group on the Groups page — they don't change how budget is allocated. For creating and managing the group itself, see the Create Budget Groups article; for capping and pacing spend, see All about: Budget Control.