What this checkpoint covers
Gatewarden Access Systems is an invented vendor of door access systems, sold to committees of facilities, security and finance staff over months. The questions draw on the guide to choosing an attribution approach, the incrementality testing guide and the enhanced conversions guide.
Illustrative example — not a real result
Gatewarden Access Systems: the measurement situation
- About 60 opportunities and 18 won deals a year; deals close four to nine months after the first enquiry.
- Last-touch report: brand search and direct visits credited with 70% of opportunities; LinkedIn and a security podcast sponsorship with 4% between them.
- On demo forms, 35% of buyers name the podcast or LinkedIn when asked how they heard of Gatewarden.
- One marketing operations manager; the CRM is connected to the website forms but not to the ad platforms.
- Google Ads optimises to 1,100 form submissions a year, of which sales accept about a third.
Task
- Answer each question with a reason before reading any explanation, and reread the linked guide for every miss.
- Where an answer changes your thinking, revise your plan in the industrial pump maker measurement exercise.
Questions
- The board proposes cutting LinkedIn and the podcast because the last-touch report credits them with almost nothing. What do you put in front of the board?
- A vendor offers an algorithmic multi-touch platform to settle the question. What approach would you choose for Gatewarden instead, and why?
- The chief executive wants a geo test on LinkedIn, read on closed deals at the end of one quarter. What is wrong with the design, and what would you run?
- After a webinar series, qualified opportunities rise sharply and marketing reports it as revenue growth. How should the rise be reported?
- What would enhanced conversions for leads change in Google Ads bidding, and what must the CRM provide first?
- A holdout on the podcast's promotional codes ends with a lift whose range includes zero. The finance director says the podcast does nothing. Is that what the result shows?
- Marketing's and sales' counts of sourced opportunities add up to more than the total. What do you fix?
Answers explained
- The gap between what last touch sees and what buyers say. Last touch credits the channel that captured an enquiry, usually brand search, not the one that created demand months earlier. Show first touch on the account, influenced pipeline and the self-reported answers beside it, and offer a test before a cut. See start from the decision.
- The most sophisticated approach the team can own. With one operations manager and a few dozen opportunities a year, a long-lookback analytics model with offline outcomes imported, sourced and influenced pipeline side by side, and one test a year on the most contested line will be trusted; an unexplained AttributionAssigning conversion credit to touchpoints. platform will not. See matching to data, volume and maturity.
- Too few deals, read too early. A quarter is shorter than the sales cycle, and a handful of deals cannot show a difference. Test on a frequent outcome that leads opportunities, such as accepted enquiries from target accounts, run across a full cycle with a cool-down, and follow the cohort's pipeline afterwards. See choosing cells, size and duration.
- As a leading indicator, with a projection labelled as one. Opportunities are not revenue. Apply the stage and win rates for webinar-sourced opportunities to project expected pipeline value, and check win rate by source when the cohort closes. See reading lift and uncertainty.
- Bidding learns from accepted outcomes, not form fills. Hashed lead data captured at submission lets later CRM stages be matched to the click without storing a click identifier. The CRM must hold the lead's email, the stage and the date it changed, with consent recorded, and only one stage should be primary. The target then becomes a CPACost of each conversion. per accepted lead rather than per form. See for web versus for leads.
- No. A range that includes zero means the test did not show an effect at the size it could detect, not that there is none. Say before the test what lift would change the decision; if the test cannot detect that lift, it cannot settle the question. Check also whether codes captured only part of the podcast's effect. See the interval, not just the point.
- The definitions. Each team is counting by its own rule. Agree one written definition of a sourced opportunity, for example first touch on the account before the opportunity, and of an influenced one, any touch on the opportunity, and keep them in the measurement plan. Only an IncrementalityConversions caused by marketing. test says what marketing caused; definitions only stop the double count. See governance: one definition of truth.
Where to go next
If the test design questions caught you out, redo the industrial pump maker measurement exercise. The next module starts with the procurement software battlecard exercise, which turns to winning the deals measurement finds. For pipeline tracing basics, return to the intermediate business-to-business level.
