--- name: set-up-conversion-based-bidding description: >- Configures a campaign to run on Maximize Conversions or Target CPA and shepherds it through the learning period to stable performance. Use it when the strategy choice is already made and the goal is conversion volume: a new launch that should bid to conversions, or an existing campaign moving onto tCPA with a known target. It verifies the primary conversion action, checks the budget can support the target, applies the strategy with a conservative initial target, and defines the post-learning tuning loop. Do not use it to decide WHICH strategy fits (select-a-bidding-strategy), to compute the CPA number itself (calculate-bid-targets), for revenue/profit objectives (set-up-value-based-bidding), to move a Manual CPC campaign over safely (migrate-from-manual-to-smart-bidding), or to pool thin campaigns (set-up-portfolio-bid-strategies). --- # Configure Conversion-Volume Bidding (Maximize Conversions / Target CPA) ## Purpose Put a campaign on a conversion-volume bid strategy with the configuration details that actually determine success: the right primary conversion action, a budget that gives the algorithm room, and an initial target the campaign can realistically hit. Then protect the learning period and run a disciplined post-learning adjustment loop instead of reactive target-chasing. ## When to run - A strategy decision (from select-a-bidding-strategy) landed on Maximize Conversions or Target CPA and it now needs to be applied correctly. - A new campaign with a lead/signup/purchase-count objective is launching. - A campaign on Maximize Conversions has gathered 2-4 weeks of data and is ready to take on a CPA target. - A tCPA campaign was configured hastily and needs its conversion goal, budget, and target audited. ## When NOT to run - Still deciding between strategies -> select-a-bidding-strategy. - No validated CPA number yet -> calculate-bid-targets first. - The objective is revenue or profit efficiency (tROAS / profit-based) -> set-up-value-based-bidding. - The campaign is currently on Manual CPC or Maximize Clicks -> migrate-from-manual-to-smart-bidding (it adds baseline capture and rollback criteria this skill assumes are unnecessary). - The campaign converts fewer than ~15 times/month on its own -> set-up-portfolio-bid-strategies to pool data instead of forcing a single-campaign target. ## Prerequisites - Conversion tracking live and verified (set-up-google-ads-conversion-tracking if in doubt). - A calculated or historical CPA reference: either the output of calculate-bid-targets or a stable trailing 4-week average CPA. - Campaign structure complete (keywords, ads, targeting), this skill only touches bidding config. - Known conversion lag for the account (days from click to counted conversion). ## Procedure 1. **Session guardrail.** Before the first Google Ads write of this session, call `gads_policy_guardrail` and honor its output. 2. **Audit conversion actions.** Call `gads_list_conversion_actions`. Confirm the intended action exists, is enabled, and is the primary-for-goal action the campaign will optimize toward. Then run `gads_run_gaql_query` segmenting the campaign's last 30 days by `segments.conversion_action_name` to see which actions are actually accruing. If a micro-event (page view, scroll) is counted as primary, stop and fix that first, the algorithm optimizes whatever is primary. 3. **Pin the campaign goal (write).** If the campaign should ignore account-default goals, set a campaign-specific conversion goal with `gads_set_campaign_conversion_goal`. Preview first (validate_only=true default), show the user exactly which action becomes primary, apply only after explicit approval. 4. **Check budget headroom.** Call `gads_list_budgets` and `gads_get_campaign`. For Target CPA the daily budget should be at least 10x the target; below that the algorithm gets too few auctions per day to pace properly. If budget is short, route to allocate-budget-across-campaigns before applying a target. 5. **Establish the baseline.** Pull the trailing 4 weeks (excluding the last conversion-lag days) with `gads_run_gaql_query`: conversions, cost, CPA, impression share lost to budget (`metrics.search_budget_lost_impression_share`) and to rank (`metrics.search_rank_lost_impression_share`). Record these, every later evaluation compares against this snapshot. 6. **Choose the configuration** using the Decision rules below: Maximize Conversions with no target (data-gathering mode) vs Target CPA with an initial target. 7. **Apply the strategy (write).** Call `gads_set_campaign_bidding_strategy` with the chosen strategy and target. Preview first, present the before/after (old strategy, new strategy, target value, budget-to-target ratio), apply only after user approval. 8. **HUMAN STEP (outside VigilDog).** Brief the account owner: expect roughly two conversion cycles of volatile CPAs, no changes will be made during that window, and the first evaluation date is set now (learning end + one conversion cycle). 9. **Learning-period watch (read-only).** For ~2 conversion cycles, only observe: daily `gads_get_campaign` checks for anomalies (spend spiking past 2x the daily budget average, conversions flatlining to zero). Verify conversions keep recording via `gads_run_gaql_query`. Make no target, budget (>10%), or keyword changes. 10. **Post-learning evaluation.** Pull performance with `gads_run_gaql_query`, excluding both the learning window and the most recent conversion-lag days. Compare actual CPA and volume to target and goals; act per Decision rules. Every target change goes back through step 7's preview -> approval -> apply flow. 11. **Ongoing cadence.** Weekly: CPA vs target, volume, IS lost (budget/rank) via `gads_run_gaql_query`. Bi-weekly: `gads_get_search_terms_report` for query drift and CPC anomalies, `gads_get_keyword_performance` for keywords dragging the average. Quarterly: re-validate the target against unit economics (calculate-bid-targets). ## Decision rules - **Volume gates for conversion bidding:** ~15 conversions/month is the absolute floor (expect volatility), ~30/month is workable for tCPA, 50+/month is where it performs reliably. Below 15, do not apply a target, pool via portfolio or run Maximize Conversions to gather data. - **Start with vs without a target:** existing campaign with 50+ conversions/month and a validated CPA number -> go straight to Target CPA. New campaign or thin history -> Maximize Conversions (no target) for 2-4 weeks, then set the target from the average CPA actually achieved. - **Initial target value:** set at, or up to ~10% above, the trailing 4-week average CPA. Never launch at the aspirational number, an aggressive day-one target strangles volume and prolongs learning. Close the gap later in 10-15% steps. - **Budget rule:** daily budget >= 10x target CPA before applying tCPA. - **Post-learning matrix:** CPA within ~10% of target and volume on goal -> leave it alone. Actual CPA >20% over target -> first rule out tracking gaps and thin volume; if the target was simply aspirational, raise it 10-15% toward reality rather than fighting the auction. Actual CPA well under target -> tighten 10-15% for margin, or hold the target and raise budget to buy volume. Volume under goal with low IS lost to budget -> loosen the target 10-15%; with high IS lost to budget -> raise budget instead (handle-budget-limited-campaigns). - **Change discipline:** one variable at a time, max 10-15% per target move, one full conversion cycle between moves. A change >20% restarts learning. - **Evaluation hygiene:** always exclude the learning window and the last conversion-lag days from any comparison; including them makes every new strategy look broken. ## Common failure modes - **Optimizing the wrong action.** Account-default goals quietly include micro-conversions; the campaign then buys cheap page views. Always run the step 2 audit before touching the strategy. - **Aspirational day-one targets.** Setting tCPA at the number the business wants (rather than what the campaign does) collapses volume and gets the strategy blamed. Start at reality, walk down. - **Fiddling during learning.** Every mid-learning tweak resets the clock and extends the volatile window that caused the panic. The step 8 brief exists to pre-empt stakeholder pressure. - **Judging with incomplete conversions.** Evaluating the last 7 days when conversion lag is 7+ days always looks like a failure. Trim the window per the evaluation hygiene rule. - **Budget-starved tCPA.** A target with a sub-10x budget behaves erratically and is misread as a bad target. Fix budget first, then judge the target. ## Related skills - Before: select-a-bidding-strategy (choice), calculate-bid-targets (the number), set-up-google-ads-conversion-tracking (tracking). - Instead, when data is thin: set-up-portfolio-bid-strategies. - Instead, for Manual CPC campaigns: migrate-from-manual-to-smart-bidding. - After: monitor-and-maintain-bid-strategy-health (recurring checkups), scale-bids-and-budgets (pushing past stable performance), handle-budget-limited-campaigns (if IS lost to budget climbs).