The Bid Strategy That Ate the Budget
A software company launches a campaign with a single line item using an automated bid strategy optimizing toward conversions, with a $5,000 daily budget and a target CPA of $80. Conversion tracking was set up three days before launch and shows 12 conversions in the first week, but the advertiser's analytics platform reports 40 sign-ups from the same period. By day 10, the line item has spent the full $5,000 daily budget every day, reported CPA is $410, and the automated strategy has pushed average CPMs up 60% versus the advertiser's historical benchmark. The advertiser wants to switch to a manual bid of $3.00 CPM to 'stop the bleeding.'
1. Explain why the automated strategy is overspending and reporting an inflated CPA, distinguishing between a tracking problem and a bidding problem.
Read the worked answer
Two issues are entangled. First, a tracking gap: the DSP sees 12 conversions while analytics sees 40 sign-ups, meaning roughly 70% of true conversions are invisible to the DSP's optimization signal. An automated conversion-optimizing strategy learns from the conversions it can see; with sparse, incomplete signal it cannot accurately estimate value and may overbid for the users it does observe, driving CPMs up 60%. Second, budget and target mismatch: a $5,000 daily budget against a $80 target CPA implies ~62 conversions/day, far beyond the observed conversion volume, so the strategy spends aggressively chasing a target it cannot measure. The reported $410 CPA is partly an artifact of undercounting conversions-true CPA against 40 sign-ups would be far lower. Switching to a flat $3.00 manual CPM treats a measurement problem as a pricing problem and would likely starve delivery of the conversions the strategy was supposed to find.
2. Recommend a remediation plan ordered by priority, and state what to monitor before re-enabling automated bidding.
Read the worked answer
Priority order:
- Fix conversion tracking first. Reconcile the DSP pixel/event with the analytics platform, confirm deduplication and attribution windows, and validate that sign-up events fire on the DSP tag. Do not tune bids against a broken signal.
- Right-size the budget and target to observed conversion volume. If true conversions are ~40/week, a $5,000 daily budget is wildly oversized; set budget and target CPA consistent with measured volume (e.g., start near historical CPA and scale gradually).
- Hold automated bidding in a learning-friendly configuration rather than abandoning it: keep the strategy but constrain budget so it cannot overspend while signal stabilizes, and avoid frequent target changes that reset learning.
- Only then consider manual bidding if automated still underperforms after tracking is fixed-manual $3.00 CPM is a fallback, not a first response. Monitor: conversion event volume vs. analytics, CPA against the corrected count, CPM trend, and delivery pacing for at least one full learning cycle before scaling. The reasoning: automated strategies are only as good as their signal; fix measurement, then economics, then strategy.