Attribution.
Where registrations came from, and how confidently you can say so. Source hygiene, unattributed records, multi-touch models, paid media measurement and incrementality testing.
31 articles · Acquisition and attribution
- The utm tagging conventions that survive five editions of one showutm tagging conventionsA locked utm tagging convention where source is the platform, medium is the mechanism and campaign carries the edition code, with the lookup that fixes old tags.
- How registration source tracking breaks between the ad click and the recordregistration source trackingRegistration source tracking fails at six identifiable hops between the ad click and the saved row. Test each one with a single tagged link and fix the worst.
- Capturing click ids on registration forms so paid credit survives the handoffcapturing click ids on registration formsCapturing click ids on registration forms puts a platform identifier on the registration row, and turns the gap against platform reported conversions into four causes.
- Cross domain tracking for registration when the form lives on another domaincross domain tracking for registrationCross domain tracking for registration stops a hosted form counting your own site as a referrer, and collapses the duplicate sessions that flatter your traffic.
- Build a channel grouping taxonomy before you argue about attribution modelschannel grouping taxonomyA channel grouping taxonomy turns raw source and medium pairs into eight reportable channels, with a first match wins order and a review queue for new values.
- Print and direct mail attribution without a click to followprint and direct mail attributionPrint and direct mail attribution needs vanity domains, one code per drop and a withheld control cell, plus the arithmetic that says how large that cell must be.
- What to do with the unattributed registrations sitting in your fileunattributed registrationsUnattributed registrations split into four causes with different fixes. Size each one over a single edition, then report the remainder as unknown rather than spreading it.
- Direct traffic registrations are four different problems wearing one labeldirect traffic registrationsDirect traffic registrations hide bookmarked returns, untagged email, referrer stripping apps and redirect loss. Three cross tabs separate them on your own file.
- Measuring dark social event marketing when the sharing happens in privatedark social event marketingDark social event marketing hides in private forwarding. Instrument it with share links, per person invite codes and a company domain cluster count on your own file.
- Adding self reported attribution to a registration form without wrecking conversionself reported attributionSelf reported attribution adds one open question after the commitment step, coded weekly into your channel taxonomy, and read beside tracked shares rather than instead.
- Attribution coverage rate is the number to fix before you trust a modelattribution coverage rateAttribution coverage rate is registrations carrying a usable source divided by all registrations. Measure it weekly and publish it beside every channel chart.
- Running multi touch attribution for events on a single registration filemulti touch attribution for eventsMulti touch attribution for events needs one touch table keyed to a person, then a weighting rule you can defend. Worked on one registration with six touches.
- First touch versus last touch on a show that sells for nine monthsfirst touch versus last touchFirst touch versus last touch answers two different questions on one registration file. Run both, and send each ranking to the meeting it belongs in.
- Why a time decay attribution model fits a deadline driven showtime decay attribution modelA time decay attribution model weights touches by how close they sit to the registration. Pick the half life from your own data and the arithmetic is reproducible.
- Markov chain attribution applied to a trade show registration journeymarkov chain attributionMarkov chain attribution derives channel credit from the removal effect rather than a rule. Worked on four states with numbers you can rebuild in a spreadsheet.
- Choosing an attribution window length for a show with a long buying cycleattribution window lengthAttribution window length decides which touches exist at all. Measure your own first touch distribution, set the window at the 90th percentile, and state what it discards.
- Measuring paid search for trade shows beyond the platform conversion countpaid search for trade showsPaid search for trade shows should be reported in three spend groups against registrations in the file, because brand, category and competitor terms behave nothing alike.
- What paid social for event registrations actually delivers once you match the filepaid social for event registrationsPaid social for event registrations sells two products at two prices. Split prospecting from retargeting, then match every platform claim to the registration file.
- Should view through conversions count towards your registration numbersview through conversionsView through conversions credit a registration to an ad nobody clicked. Set a viewability floor, a short window, and a de duplication rule that puts clicks ahead of views.
- Closing the platform reported conversions discrepancy against your registration fileplatform reported conversions discrepancyThe platform reported conversions discrepancy is four causes stacked on each other. Walk 1,480 claimed registrations down to 910 rows, one adjustment per line.
- Retargeting abandoned registrations from the point the form was leftretargeting abandoned registrationsRetargeting abandoned registrations works when the audience is built from the last completed step and recovery is counted as completions inside seven days.
- Suppression lists for registered attendees stop you paying twice for the same personsuppression lists for registered attendeesSuppression lists for registered attendees remove people who already registered from every paid audience daily, and the match rate on hashed uploads caps the saving.
- Design a retargeting incrementality test that survives a finance reviewretargeting incrementality testA retargeting incrementality test holds back a random slice of the audience, then compares registration rates with a two proportion test sized for the lift you need.
- Email attribution for event registrations that survives forwarding and shared inboxesemail attribution for event registrationsEmail attribution for event registrations holds up when links carry a send identifier, the join runs on normalised email, and forwarded clicks credit the send.
- Open rate inflation makes your event email reporting look better than it isopen rate inflationOpen rate inflation comes from mail privacy proxies fetching pixels automatically, so event email reporting needs click and registration rates over delivered instead.
- Reactivation email measurement for people who attended two editions agoreactivation email measurementReactivation email measurement needs the lapsed file split by last attended edition, a random tenth held back, and registration rates compared instead of opens.
- Media partner promo codes that actually tell you which title deliveredmedia partner promo codesMedia partner promo codes only measure a title when each placement gets its own code and the landing page carries a matching source tag, so code and click agree.
- Counting association partner registrations without double paying for the same membersassociation partner registrationsAssociation partner registrations overstate a deal when member lists overlap your own file, so measure the overlap first and report new to file registrations separately.
- Valuing a media partnership when no money changes handsvaluing a media partnershipValuing a media partnership means pricing both halves: what the space and passes really cost you, and what the inbound advertising would have cost at your own rates.
- Rules for double counted registrations across channels when three sources claim one persondouble counted registrations across channelsDouble counted registrations across channels need one precedence rule and two published counts, so partners see their claimed number while the total stays exact.
- Measuring exhibitor driven registrations without taking credit away from marketingexhibitor driven registrationsReport exhibitor driven registrations in three buckets, new to file, known but unengaged, and already registered, then work the split on 3,400 code redemptions.