Lead Quality Collapse After a Smart Bidding Rollout
Summit Home Services runs HVAC, plumbing, and electrical services across 14 locations. It spends $85,000 per month on Google Ads. Six weeks ago it moved from manual bidding to tROAS of 600% and broad match. Lead volume rose 40%, but booked jobs fell 12%. The call center reports more wrong-service and out-of-area leads. Call tracking is installed at only six locations. The conversion action counts every form submit and every call over 60 seconds equally. About 30% of form submissions are spam. Competitors bid on Summit's brand terms, and the team has not refreshed negative keywords in five months. The marketing director wants to return to manual bidding immediately. The agency argues that the tROAS algorithm needs more time and more conversion data.
1. Diagnose why lead volume rose while booked jobs fell. What is the most likely root cause?
Read the worked answer
The bidding system optimized to a conversion signal that is not correlated with booked revenue. If every form submit and 60-second call counts equally, tROAS will buy cheap, low-intent leads-wrong service, out-of-area, spam-because those are easy to generate. Broad match expanded query reach, and five months without negative keywords let irrelevant queries accumulate. The 40% volume increase and 12% booking decline are consistent with a lower lead-to-job rate, not a demand problem. Call tracking at only six locations means the other eight locations' calls are invisible or misattributed, so the algorithm learns from a biased sample. The 600% tROAS target may also be arbitrary if conversion values are not service-specific; a $300 drain cleaning and a $12,000 HVAC replacement are treated the same. Returning to manual bidding immediately would treat the symptom and lose data. The root cause is conversion-quality measurement, not bid strategy alone.
2. What would you change in the next 14 days without destroying the learning period?
Read the worked answer
Keep tROAS running but change the signal and guardrails. (1) Rebuild conversion actions: separate form, call, and booked job; assign values by service line and location; exclude spam via CAPTCHA, honeypot, and form-quality rules. (2) Implement offline conversion imports from the CRM so booked jobs feed back to Google, even if initially only for the six tracked locations. (3) Add location-specific call tracking or use a dynamic number insertion vendor for the remaining eight. (4) Split brand and non-brand campaigns; set a brand cap and separate tROAS. (5) Refresh negative keyword lists from search terms, especially out-of-area and wrong-service queries. (6) Set geo targeting to presence, not interest, and review radius overlap. Do not change everything at once; document each change and use drafts and experiments. Set a 14-day checkpoint on lead-to-book rate by campaign.
3. Design a 60-day test and the guardrails you would present to the marketing director and agency.
Read the worked answer
Run a geo-based holdout or campaign experiment. Test: tROAS with offline booked-job values vs tROAS on form or call values, while holding budget and geo constant. Primary metric: booked revenue per dollar, not CPL. Secondary: lead-to-book rate, wrong-service rate, and out-of-area rate. Guardrails: pause any campaign with booked-job rate below 5% after 30 days; cap spend on broad match to 30% of non-brand; require negative keyword review twice weekly; do not raise tROAS targets until offline conversion volume reaches statistical sufficiency. Use a 60-day window because home services have weekly seasonality and longer consideration. Report weekly to the director with a simple dashboard: spend, qualified leads, booked jobs, revenue, CAC, and tROAS. If the test shows no improvement, revert to manual or maximize conversions with value rules. The agency's more-time argument is only valid if the conversion signal is fixed first.