B2B Demo Campaign With Geo and Placement Leakage
A B2B SaaS company selling a project-management tool launches its first Display campaign to generate demo requests in the US and Canada, with English as the only language. Budget is $6,000 over 30 days, bidding on Target CPA $90, and the campaign uses a single ad group with one custom intent audience (people researching named competitors) layered on the default 'presence or interest' geo setting. After three weeks: 1.1M impressions, CTR 0.19%, 4 conversions, CPA $215. The placements report shows 58% of spend on mobile app categories, mostly games, and 31% of impressions come from outside the US and Canada. The audience report shows the custom intent segment has only about 40,000 active users. The marketing lead wants to add 15 more niche audiences and raise the tCPA to $200 to 'buy volume'. Sales says lead quality is acceptable but volume is far below the 25 demos promised.
1. Diagnose which campaign settings most plausibly explain the impression distribution and the CPA, and specify what you would change before touching bids.
Read the worked answer
Two settings stand out. First, 'presence or interest' geo targeting deliberately reaches people outside the selected countries who show interest in them, which explains the 31% out-of-geo impressions; switching to 'presence' only is the first fix, and the location report should confirm which countries are leaking. Second, Display campaigns are eligible for mobile app inventory by default, and the placement report shows 58% of spend there. B2B demo requests rarely originate from game apps, so exclude the app category or the specific app placements, then check device conversion rates rather than assuming all mobile is bad.
The CPA of $215 with only 4 conversions also reflects a statistical problem: Target CPA needs roughly 30 conversions in 30 days to stabilize, and 4 conversions over three weeks is far below that. Spend is about $41/day against a $200/day budget, so budget is not the constraint - auction eligibility is. Fixing geo and app exclusions raises the quality of what remains; broadening reach with in-market or affinity segments (not just more competitor-research segments) addresses the 40,000-user ceiling. Change settings first, then reassess bids with clean data.
2. Evaluate the proposal to raise tCPA to $200 and add 15 narrow custom intent audiences. Propose an alternative structure that could reach 25 demos within the same budget.
Read the worked answer
Raising tCPA from $90 to $200 does not create reach; it raises the ceiling on what the system will pay for the same small pool of eligible users, so the likely outcome is a higher CPA on roughly the same volume, plus a learning reset. Adding 15 niche custom intent segments fragments an already thin audience; combined reach may grow, but each segment is small and the system cannot learn efficiently across them.
A better structure: one prospecting campaign with a broad but relevant audience mix (one in-market segment plus one custom intent segment covering the category, not 15 competitor names), app category exclusions, presence-only geo, and a tCPA around $110–$130 to reflect the actual market clearing price. Keep the $6,000 monthly budget but split it into two ad groups by message theme, and run a separate remarketing campaign later. Set the tCPA from observed data after the targeting fixes, allow two weeks of learning, and only then adjust. If volume still stalls, test Maximize Conversions without a target to discover the achievable CPA before reimposing a target.
3. How would you verify that mobile app placements are actually harming performance rather than merely absorbing impressions?
Read the worked answer
Segment the campaign by placement, device, and network in the report and compare conversion rate and CPA, not just CTR. Look at whether conversions attributed to app clicks are real demo requests or misattributed sessions that convert later on desktop; check the conversion path and the click-to-conversion time. Then run a controlled test: exclude the top app categories for 14–21 days while holding bids and budget constant, and compare CPA and conversion volume before and after. Given only 4 conversions, a two-week test may not reach significance, so pre-plan the sample and treat results directionally. Also check the placement report for 'app' versus 'mobile web' and verify that app conversions are not duplicates from the same users. The goal is to establish whether app inventory is cheap but non-converting or merely a scapegoat for a too-narrow audience.