Spend is split across campaigns with too few purchases to judge

What you see

A material share of spend goes to small campaign units whose purchase results are too uncertain for the planned decision.

Illustration: two campaigns show $200 observed cost per first purchase, but sparse purchase counts don't support a reliable ranking. — Each campaign has an observed cost of $200 per first purchase.

Check this first

Compare spend by campaign with the amount and precision of mature purchase evidence.

Data needed

Campaign spend, eligible exposure, and mature purchase results
Campaign purpose, decision rule, and required precision
Dated campaign structure and separate platform learning status

Run the check

#1
State the budget decision and the difference that would change it.
#2
Report spend in units that can't yet distinguish that difference.
#3
Check whether similar units need separate delivery or reporting.
#4
Inspect platform learning status separately from statistical precision.

Calculate

For each decision unit, report spend, eligible exposure, mature purchases, cost, and uncertainty. The required precision comes from the size of the decision; there is no universal purchase-count cutoff.

Compare groups

Campaign or true optimization unit
Purpose: scale versus test
Mature purchase count
Decision limit and uncertainty

What a healthy result looks like

Routine campaign decisions use enough evidence for their economic consequence. Small tests have a stated purpose and decision plan.

When to investigate

Use decision-specific uncertainty or test power. Set the read window and minimum evidence for the planned decision.

Possible causes

Too many units divide the available purchase volume.
Similar audiences are separated without a useful economic distinction.

Rule out these explanations

Deliberate bounded experiments
Required regional or contractual separation
Late purchase reporting

What to do next

Consolidate only where the units share a goal and can be managed together. Keep necessary tests distinct and use an adequate evaluation plan.

What this check can tell you

A platform learning label and statistical uncertainty are different measures. Neither alone proves that merging campaigns will improve results.

Use this check with AI

Run a read-only check for: Spend is split across campaigns with too few purchases to judge.
First confirm the available sources, columns, row grain, date basis, currency, and customer definition. Use only authorized data.
Required inputs: Campaign spend, eligible exposure, and mature purchase results; Campaign purpose, decision rule, and required precision; Dated campaign structure and separate platform learning status.
Check: A material share of spend goes to small campaign units whose purchase results are too uncertain for the planned decision.
Calculate: For each decision unit, report spend, eligible exposure, mature purchases, cost, and uncertainty. The required precision comes from the size of the decision; there is no universal purchase-count cutoff.
Slice by: Campaign or true optimization unit; Purpose: scale versus test; Mature purchase count; Decision limit and uncertainty.
Use this comparison rule: Use decision-specific uncertainty or test power. Set the read window and minimum evidence for the planned decision.
Show the source totals, calculation, unknown groups, missing inputs, and result. Don't invent fields, thresholds, customer matches, or causal effects.
Rule out: Deliberate bounded experiments; Required regional or contractual separation; Late purchase reporting.
Describe the observed signal separately from possible explanations. If the check is incomplete, state the exact data needed.
Make no account or budget changes.

Sources and definitions

Google: how bidding learnsLearning depends on source conditions. Don't apply a universal event threshold.Read source
Google: conversion delayLate conversions can change a fixed period’s apparent acquisition efficiency.Read source