Free PPC

Toolkit

The bid strategy you pick decides what the algorithm spends two months learning. Pick wrong and Smart Bidding optimizes confidently toward the wrong outcome, then resists every attempt to course-correct. This is the gate I run before flipping a strategy.

The framework

The strategy you can use is gated by your data. Pick from the volume row first, derive the target second, pay the learning tax third.

Volume gate

Target derivation

Targets come from unit economics, not feel. Set within 20% of historical performance.

Learning tax

Every strategy change costs 1 to 4 weeks of 10-30% worse performance. Pre-define rollback criteria in writing. Honor the wait or you waste the investment without learning anything.

Diagnostic by symptom

What this framework solves

Bidding is the economic engine of the account, but the platform’s recommendations push automation regardless of whether your data can support it. Switch to Smart Bidding too early and the algorithm optimizes on noise. Set aspirational targets and the algorithm under-delivers or finds junk. Change strategies frequently and you pay the learning tax over and over. This framework gates each decision on the data that has to be true for it to work.

The core idea

Smart Bidding is a machine learning system that needs statistical signal to function. Below 30 click-through conversions per month, the model cannot distinguish patterns from noise. This is a mathematical constraint, not a guideline. The bid strategy is only as good as the conversion signal it receives: optimize toward “form fill” without offline import and the algorithm finds the cheapest form fills, which are almost never the most qualified leads. Every strategy change resets a 1 to 4 week learning period during which performance is volatile and often worse than baseline. That cost is the learning tax, and it has to be paid before any new strategy can be evaluated.

How to apply it

1. Check click-through conversion volume against thresholds. Below 30/month: Manual CPC or Maximize Clicks. 30-50: gray zone, depends on consistency and risk tolerance. 50+: Target CPA or Maximize Conversions become eligible. 50+ with values: Target ROAS. 100+: automation consistently outperforms manual. Filter to click-through only; view-through conversions inflate apparent readiness.
2. Verify weekly conversion consistency, not just monthly totals. 35 conversions spread evenly is reliable signal; 40 in bursts confuses the model. Check the weekly distribution over 60 days before trusting automation.
3. Calculate targets from unit economics. In lead gen: (Average Deal Value × Repeat Rate × Margin) × Conversion Rate × Lead-to-Sale Rate gives target CPC, which backs into target CPA. In ecommerce: break-even ROAS = 1 ÷ gross margin, then add a 10-20 percent buffer. Aspirational targets (“$30 sounds right”) force the algorithm into auctions it cannot win.
4. Stay within 20 percent of historical performance when setting targets. Beyond that, the algorithm restricts spending, under-delivers, or chases low-quality traffic to hit the impossible number.
5. Match strategy to value structure. Uniform conversion value (lead gen, B2B demos): Target CPA. Variable order values (ecommerce): Target ROAS or Maximize Conversion Value. Brand campaigns with active competitor bidding: Target Impression Share at 90-100 percent absolute top.
6. Pay the learning tax explicitly. Expect 10-30 percent worse performance for 1-4 weeks. Set rollback criteria before switching (“revert if CPA exceeds target by 50 percent for 2+ weeks”) and honor the wait. Revert too early and you waste the investment without learning anything.
7. Change one variable at a time. Cutting budget while lowering target CPA destroys diagnostic clarity and often degrades performance further by shrinking the auction pool. Wait 14 days between changes for 30-50 conversion accounts, 7 days for 100+, before evaluating.
8. Diagnose the symptom before changing strategy. “Bidding stopped spending” is usually a too-aggressive target, not a bidding failure. “Lead quality collapsed but CPA stable” is a missing offline conversion signal, not a bid problem. “Performance dropped after 3 weeks” is tracking, competition, or market shift before it is bidding.

When to use it

Use this whenever picking a strategy for a new account, recommending a switch from manual to automated (or back), setting or revising target CPA or ROAS, and diagnosing erratic Smart Bidding behavior. It calibrates by vertical: ecommerce is the most automation-friendly because conversion signal is clean; lead gen requires offline conversion import to avoid optimizing for cheap junk; B2B SaaS usually stays manual because volume is below threshold and attribution lag is 60-120 days; local services needs call-duration filtering and tight geo bid adjustments; high-value verticals raise the readiness bar (50+ quality conversions, offline tracking, 3+ months consistency) because a $150 CPC means a learning period costs five-figure dollars. It does not override structural problems: a bid strategy operating against a 1 percent converting page or fragmented campaigns will not be rescued by switching from Manual CPC to Target CPA.