Subscribe
Conversations

EP 278- Track What You Can, Let Go of the Rest: Dana DiTomaso on Attribution After GA4

Dana DiTomaso of Kick Point tells Greg Sterling and Mike Blumenthal why attribution was always a little fictional, how much of your traffic you can really track, and the GA4 fixes, brand-search signals and forecasting methods that still work.

Near Media, Mike Blumenthal

EP 278- Track What You Can, Let Go of the Rest: Dana DiTomaso on Attribution After GA4

Dana DiTomaso, principal and founder of Kick Point, returns to the Near Memo (her first visit was GA4's early days) to talk with Greg Sterling and Mike Blumenthal about what's left of attribution: how much of a site's traffic is trackable, why one AI-started purchase can show up as four different channels, and what a marketer can still measure with confidence.

Attribution was always partly fiction that Universal Analytics papered over, and GA4 and privacy law have simply exposed it — so as consent banners, ad blockers and AI referrals keep shrinking the trackable slice, the job shifts from reconciling exact numbers to sizing what you can see, watching directional signals like branded search, and explaining the gaps to clients before they find them.

Dana DiTomaso's Trackability Flowchart

Dana's main client tool is a flowchart (she builds hers in Figma) that follows a visitor from first touch to the CRM and shows where each one drops out of view (11:56–12:43, 32:13–34:12). Her rule for every stage is the same: "track the heck out of" the people you can see, and let go of the rest.

Stage What it answers Primary lever
Before the site Where did the journey actually start: an AI answer, a brand search, a shopping listing, a link a spouse texted? Walk through specific scenarios with the client. Add UTMs wherever you control the link. Watch branded-search impressions in Search Console.
Consent banner What share of visitors said no to tracking? Read the decline rate from the consent tool (she uses Cookiebot), and make sure "no" actually stops the tags.
Ad blockers and browser privacy How much of the consenting traffic still disappears? A cookieless backup (Plausible) for page views, and server log files, which nobody can hide from.
GA4 processing How does the visible traffic get bucketed? 14-month data retention, and custom channels for GBP, Bing, tab collectors and any small channel that matters.
CRM Did the lead close, and where did it really come from? Tag the sign-up source (in store or online) and check self-reported answers against GA4.

Takeaways

  • Attribution isn't dead, just not healthy. It is "not as dead as the parrot" in the Monty Python skit. Universal Analytics hid incomplete attribution, while GA4 hands you the data and lets you find the gaps. Privacy laws such as GDPR and California's CPRA arrived at the same time and made the gaps bigger. (02:03–02:40)

  • Your trackable share depends on who visits. In Texas, with a non-technical audience, few people go untracked. For some of Kick Point's European-facing clients, 80% decline tracking, and about half of the European visitors to Dana's own site say no. Ad blockers and Safari's Intelligent Tracking Prevention cut further into the rest. (04:48–05:52)

  • Most small businesses will never qualify for GA4 modeling. Consent Mode v2 sends cookieless pings, but GA4 only models the people who declined if you have roughly 1,000 consenting and 1,000 declining visitors a day on 7 of the past 28 days. "If you're never gonna qualify… just do something else instead." (05:57–06:56) Google's thresholds [1]

  • One AI-started purchase can show up as four different channels. A shopper asks ChatGPT for a Canadian lighting manufacturer. If the link carries no UTM, the click lands as direct. If the shopper then Googles the brand and clicks its shopping listing, Search Console records a branded search, Merchant Center records an impression, and GA4 credits Google Shopping. The tell is an srsltid parameter on the URL. (07:54–09:40) [2]

  • "Unassigned" is often tab collectors. A visitor who leaves your blog post open for two weeks and then books has had their session time out, so GA4 shows them as unassigned. Dana moves that traffic into a channel she names "session timeout reactivation" (or "tab collectors"), so nobody has to keep explaining it. (09:47–10:55) Kick Point Playbook

  • Size the slice before you make decisions. Start with 100 visitors, subtract everyone who said no, and then look at the channel mix. A channel that looks like half your traffic in GA4 can be a much smaller share of what really happened. She also warns against the opposite mistake: a Fortune 100 client spent more on meetings about 0.05% of its traffic than that traffic earned. (13:03–15:38)

  • "How did you hear about us?" undercounts what actually worked. Home buyers credit the "show home this way" sign, not the ad that put the builder in their head. One client's branded PPC leads say "recommended by a friend," which is no reason to cut branded search: it's cheap and it keeps competitors off your name. Rotate drop-down options, because people pick the first one. (15:48–16:50, 37:55–38:20)

  • Use regression to forecast staffing four weeks out. Dana regressed weekly GA4 sessions against booked appointments in Google Sheets and found a moderately strong relationship. A pacing chart shows whether sessions will reach the appointment goal before the month runs out, early enough that a client doesn't have to lay anyone off. After the client redesigned its booking flow, the line turned "almost straight up." (20:22–22:20) Kick Point Playbook

  • Branded search is your best sentinel metric. Dana tracks Search Console impressions for her own name ("however they decide to spell it"). For any client running billboards, radio, CTV or anything else outside PPC, she treats branded-search volume as table stakes, and in the AI era it is "probably gonna be the best signal that you've got." (24:44–26:30)

  • Day-of-week spikes are probably rank trackers. Leave your own brand name out of your rank tracker. If impressions jump on the same weekday, such as every Monday, it's probably a competitor's tracker, unless you work in personal injury or bankruptcy, where Monday demand is real. (26:58–27:36)

  • AI visibility tools aren't rank trackers. Position in a tool's synthetic prompt won't match any real user's answer. Use the tools to learn whether AI knows you exist and what it says when asked about you, then correct whatever is outdated. Dana discloses that she sits on SE Ranking's customer advisory board and uses SE Visible. She also likes Knowatoa, Gumshoe and Rand Fishkin's Alertmouse, and says nobody needs to spend "thousands of dollars on it a month." (26:16–26:45, 28:52–31:36)

  • Two GA4 settings to change today. Set event data retention to 14 months. The default is two, which silently empties Explorations. Then split organic search, often 40–50% of traffic, into Google organic, Google Business Profile, Bing and other organic. Small channels that matter get their own line too, like one client's Yelp ads channel. Finally, check your country report: an Edmonton realtor's traffic was about 80% "from China," which Dana believes is LLM bots, and bots also inflate direct traffic. (40:06–41:42, 45:41–46:35)

Practitioner Notes

  • Self-reported source and GA4 have to be read together. Dana's finding that branded-PPC leads describe themselves as "recommended by a friend" is the same tension that runs through Near Media's monthly lead-source attribution work for law firms. Neither the intake answer nor the analytics channel is the truth on its own. The useful report puts the two side by side, as Dana does when she matches web sessions to phone calls.
  • Greg's AI Overview observation makes branded search an AIO metric. In Near Media's user research, searchers who see firms named in an AI Overview often follow up with a brand search (27:36). Search Console won't link the two queries, but Dana's sentinel metric means a rise in branded impressions after a firm starts appearing in AIOs is a usable, if indirect, signal. It pairs with the brand-awareness findings in EP 270.
  • A dedicated GBP channel should be standard for local clients. Dana's split of "organic search" into Google organic, Google Business Profile and Bing is the GA4 counterpart to tagging the GBP website link with UTMs. Without both, a local business can't tell how much of its "organic" traffic is really its profile, and it can't see how that share shifts as AI Overviews and Ask Maps change the local SERP.

Concepts

  • Directional trends — reading analytics for direction and rate of change, not exact counts, because no tool sees everyone.
  • The trackable slice — the share of real visitors left after consent declines, ad blockers and browser privacy. Every channel percentage is a share of this slice, not of all traffic.
  • Tab collectors (session timeout reactivation) — visitors who return to a long-open tab after their session has expired. GA4 loses their source and files them as "unassigned."
  • Cookieless pings — the signal-only hits Consent Mode v2 sends for visitors who decline tracking. Like Plausible, each one counts as a brand-new visitor.
  • Effectiveness ratio — total impressions across every channel compared with closed sales or leads. If the ratio is flat or rising, keep going. If it falls, something you're doing has lost its punch.
  • Pacing chart — a regression-based forecast of whether current sessions will reach the month's appointment or sales goal before the month runs out.
  • Sentinel metric — a "canary in a coal mine" measure, such as branded-search impressions, that shows shifts no single channel report can.
  • Visibility, not position — the right question for an AI visibility tool: does the model know you, and is what it says accurate?

Pick your starting point:

00:00 Welcome and Dana DiTomaso
01:58 Is attribution dead?
04:42 How much of your traffic can you track?
05:52 GA4 modeling thresholds
07:45 One AI search, four attribution channels
09:47 Tab collectors and "unassigned" traffic
13:28 Sizing the trackable slice
15:38 The trouble with "How did you hear about us?"
19:29 The effectiveness ratio
20:22 Forecasting bookings with regression
24:44 Branded search as a sentinel metric
28:37 Are AI visibility tools worth it?
31:36 The way forward: consent, flowcharts and trust
39:43 Two GA4 settings to change right now
43:23 The best GA4 alternatives
45:07 LLM bot spam in GA4

Full Transcript --> on-line

Interested in sponsoring this podcast or our newsletters please reach out to mblumenthal@nearmedia.co

E-mail Mike