Why Price Testing and Bid Testing Can't Happen Simultaneously

The Testing Conflict

Price and bidding strategy are both direct inputs into conversion rate. A price increase reduces conversion rate (fewer people buy at higher prices). A bidding change alters which auctions you enter and which users see your ads. Both affect the same output metric.

When both variables change simultaneously, attribution becomes impossible. Your conversion rate dropped 15%. Was it the price increase that deterred price-sensitive shoppers? Or the new bid strategy that shifted traffic to less-qualified audiences? Or a compound effect of both? Without isolation, you can't know - and you can't learn.

Why Simultaneous Fails

Smart Bidding learns from conversion patterns. When you change prices, conversion patterns shift. The algorithm needs 1-2 weeks to recalibrate to the new conversion rate. During this recalibration period, it's bidding based on outdated conversion expectations.

If you simultaneously change the bid strategy, you've removed the algorithm's baseline entirely. It doesn't know what "normal" looks like because you've changed both the input (which users it targets) and the output (how likely those users are to convert at the new price).

This is the same principle behind why discounts destroy bidding efficiency. Any change to conversion rate destabilises Smart Bidding. Compounding that with a simultaneous strategy change makes recovery take 3-4 weeks instead of 1-2.

Sequencing Tests

The correct approach: sequence tests with adequate stabilisation periods between each. A practical framework:

• • Week 1-2: Implement price change, keep bidding stable
• • Week 3-4: Let Smart Bidding recalibrate to new conversion patterns
• • Week 5-6: Evaluate price test results using stable bidding data
• • Week 7-8: If price change is permanent, now test new bid strategy
• • Week 9-10: Evaluate bid strategy results against the new pricing baseline

Yes, this means testing takes 10 weeks instead of 2. But 10 weeks of reliable data beats 2 weeks of noise every time. Fast testing that produces unreliable results isn't actually fast - it just creates more tests.

Isolation Methods

If you must run concurrent tests (due to business pressure or seasonal constraints), use geographic isolation. Test pricing in one region and bid strategy in another. This prevents the variables from compounding while still generating parallel learning.

Campaign-level isolation works too: test pricing on Shopping campaigns (where bids are already automated) while testing bid strategies on Search campaigns (where pricing is less visible). The key is ensuring no single campaign is subject to both changes simultaneously.

Reading Results

When evaluating test results, use contribution margin per click - not ROAS or conversion rate alone. A price increase might reduce conversion rate but increase margin per order. A bid strategy change might increase CPC but improve conversion quality. Only contribution margin captures the full picture.

Document every test with start date, end date, variable changed, control group, and margin outcome. Without documentation, test learnings evaporate when team members change. Your testing log is an institutional asset - treat it like one.

Next Steps