Defending Display Spend Against a Last-Click Verdict
A B2B software company generates demo requests through Display ($20,000/month), Search ($40,000/month), and LinkedIn (outside Google Ads). Google Ads conversion tracking counts a 'demo request' with a 30-day click window and a 1-day view-through window, and reporting uses last-click attribution. Display reports 40 conversions at $500 CPA; Search reports 200 at $200. The view-through report shows 90 additional view-through conversions for Display, 35% occurring within one day. Finance sees only the last-click numbers and wants to cut Display entirely. The sales cycle is 60–90 days, and the CRM holds opportunity values. Search volume for the brand has risen sharply, and Display impressions are concentrated on news sites. There is no holdout or geo experiment in place.
1. Assess the evidence Finance is using. What are the flaws in concluding that Display should be cut?
Read the worked answer
The comparison is not apples to apples. Last-click credits Display only when it is the final click, which systematically undervalues upper-funnel channels that create awareness before a branded search. Display's $500 CPA may therefore overstate its true cost per acquisition, while Search's $200 CPA includes conversions that Display or LinkedIn influenced. The view-through conversions cut the other way: 90 view-throughs at a 1-day window may include users who would have converted anyway, so they should not be added uncritically to Display's credit either.
Finance is also ignoring the sales cycle. Demo requests are not revenue; the CRM holds opportunity values that mature over 60–90 days, and a channel that produces fewer but larger opportunities can be more valuable. Rising brand search is consistent with Display working, but correlation is not proof. The honest conclusion is that last-click alone cannot settle the question - not that Display is either effective or wasteful.
2. Design a measurement approach that would let the company make a defensible keep-or-cut decision.
Read the worked answer
Run an incrementality test rather than another attribution report. The cleanest option is a geo holdout: select matched metro areas, pause Display in half for four to six weeks, and compare demo requests and CRM pipeline between test and control, not platform-reported conversions. If conversion lift studies are available for the account, use one; otherwise a well-matched geo test is the practical substitute. Predefine the primary metric as incremental qualified demos or pipeline value, and calculate the minimum detectable effect given $20,000/month and roughly 40 last-click conversions.
In parallel, improve the measurement stack: import CRM opportunity values as conversion values, use data-driven attribution alongside last-click, and report assisted conversions and path length. Segment by new versus existing customers and by deal size. Judge Display on cost per incremental opportunity and pipeline-to-spend ratio. If the holdout shows no lift after a properly powered test, cutting is defensible; if it shows lift, the budget decision changes.
3. What changes would you make to conversion tracking and attribution before the test, and what would you do about the news-site placements?
Read the worked answer
Separate conversion actions by funnel stage: demo request, marketing-qualified lead, and opportunity, each with a value, and import CRM values so bidding can optimize toward revenue rather than a flat demo count. Keep click and view-through windows deliberate - a 1-day view-through window is conservative and reasonable, but document it and avoid double counting when the same user sees Display and clicks Search. Enable data-driven attribution as a second lens, and deduplicate conversions across campaigns in reporting.
For the news-site placements, pull the placement report and exclude low-quality or brand-unsafe sites, and check whether view-through conversions are concentrated on placements with incidental exposure. Consider limiting Display to placements with observed assisted conversions, and add brand-safety exclusions. Do this before the holdout so the test measures a cleaned-up campaign, not a straw man. If the cleaned campaign still shows no incremental lift, the case for cutting is much stronger.