Overview
Microsoft Advertising offers several automated bidding strategies — each thrives under specific data and tracking conditions. Your job is to pick the right one for your stage, feed it clean conversion data, and guide it with budgets, targets, and guardrails.
When to use which strategy
- Manual CPC / Enhanced CPC (short term): Use for brand-new accounts or low data. Good control, but labor intensive.
- Maximize Conversions (no target): Early automation; let the system learn after you have reliable conversion tracking.
- Target CPA: Once you have stable conversion volume (e.g., ~30+ conv in the past 30 days per campaign), set a CPA target near recent actuals.
- Target ROAS (for Shopping/Revenue): Requires accurate revenue tracking and enough transactions (e.g., ~30–50+/30d).
Data prerequisites
- UET installed site-wide + verified conversion goals (primary vs. secondary defined).
- Attribution consistent with your KPI (e.g., last-click vs. data-driven if available).
- Clean conversion events (no double-firing; no soft conversions marked as primary).
- Enough volume per campaign/ad group to let strategies learn.
Setting targets (CPA/ROAS)
Set targets based on recent actuals and your unit economics, then adjust gradually:
- Target CPA: Start at your last 30–60 days blended CPA for that campaign. Move in 10–15% steps every 1–2 weeks.
- Target ROAS: Start at actual ROAS −10% (slightly easier). Raise in 10–15% increments after stability.
Large swings reset learning and can tank volume. Be patient.
Learning & stabilization
- Expect a 7–14 day learning phase after strategy or target changes.
- Hold other variables steady (broad restructures, budgets, ads) during learning.
- Read blended KPIs (CPA/ROAS, conv. rate) rather than day-to-day noise.
Budget pacing
Use simple pacing logic to avoid end-of-month surprises:
- Daily budget vs. Month target: Track MTD spend vs. linear plan. If behind by >10%, increase budgets by a small step; if ahead, reduce slightly.
- Weekend/weekday seasonality: Apply schedule modifiers or separate campaigns if behavior diverges strongly.
- High-spend safeguards: Rule to cap bids/budget if CPA spikes or if no conversions after X clicks.
Rules & scripts
Low-risk automations that compound over time:
- Search term mining: Auto-flag queries for review if CTR < threshold or CPA > threshold; suggest negatives.
- Budget pacer: Adjust budgets ±5–15% to hit MTD target (respect learning phases).
- Ad rotation: Pause worst RSA variant after X impressions if underperforms by Y%.
- Placement controls: For audience/display extensions, exclude poor placements by rule.
Anomaly detection
Set simple alert thresholds to catch issues early:
- Daily conversions down >50% vs. trailing 7-day average.
- CPA up >40% day-over-day and conversion rate down >25%.
- UET goal fire rate drops suddenly (tracking or site issue).
Many anomalies are caused by tracking changes, feed issues, or landing page outages — check those first.
A/B experiments
Use campaign experiments or split structures to test:
- Max Conversions vs. Target CPA
- Different CPA/ROAS targets
- Keyword expansions (Phrase vs. Broad with strong negatives)
- Creative variants and assets
Run for at least 2 weeks (or until stat power is reasonable), then roll forward the winner.
Common pitfalls
- Targets too aggressive, too soon: Choking volume stalls learning.
- Dirty conversions: Soft goals marked as primary mislead bidding.
- Constant changes: Frequent edits never let the system stabilize.
- Single-campaign view: Look at portfolio performance and LTV, not just daily CPA.
Implementation checklist
- ✅ UET verified; primary conversions defined
- ✅ Start with ECPC or Max Conversions → collect volume
- ✅ Move to Target CPA/ROAS when stable (30–50+ conv/30d)
- ✅ Adjust targets in 10–15% steps every 1–2 weeks
- ✅ Add pacing rules & anomaly alerts
- ✅ Run controlled experiments; roll out winners
Next steps
Ready to scale? Pair your bidding strategy with robust audience layering and remarketing. Continue with Audiences & Remarketing That Actually Move CPA.