Ground-Truth Attribution: the model
Event attribution fails because the context layer — who you met, where, and what happened — never gets recorded. Fix the record and the rest follows.
For field reps & aes · field marketers & booth teams · sales managers · company admins · marketing ops · 4 min read · Updated July 10, 2026
Events are usually the biggest discretionary line in a B2B marketing budget, and the least accountable one. A booth at a mid-size trade show runs five figures before anyone books a flight. And when the quarter closes and someone asks what it produced, the answer is assembled from a badge-scan CSV, a stack of business cards, and whatever the reps still remember.
That’s not an analytics problem. It’s a record problem. You can’t attribute what you never wrote down.
The missing layer
B2B makes this worse than it looks. A deal that closes in March might trace back to three people from the same account, met at two different events, eight months apart. The buying committee is plural, the cycle is long, and the touchpoints that matter most — real conversations at real events — are precisely the ones that live only in someone’s head.
Between the event floor and the CRM there’s a layer of context that almost every team loses:
- Who you actually talked to (not who walked past the scanner)
- Where — which event, which dinner, which booth
- What happened — a demo? a brush-off? a promise to send pricing?
- What’s owed — the follow-up you committed to, out loud
Badge scans don’t carry it. Memory doesn’t keep it. The CRM never receives it. So attribution gets reconstructed — weeks later, from artifacts — instead of read from a record.
Ground truth, defined
Ground-Truth Attribution starts from one rule: record the interaction at the moment it happens, and make that record the source of everything downstream.
The unit of record is the interaction: one person, at one Event, with a label that says what kind of conversation it was, at a timestamp, with your note in your own words. Not a stage guessed at retroactively. Not an export enriched after the fact. A firsthand entry, made by the person who was there, while they still knew the answer.
Stack up enough interactions and attribution stops being an argument. “What did the show produce?” becomes a reading exercise: 43 conversations, 17 of them qualified, 12 follow-ups completed inside a week, 3 now in pipeline. Every number traceable to a named person on a named day.
The six-stage loop
The model runs as one loop per Event:
| Stage | What happens |
|---|---|
| 1 · Prep | Create or import the Event, set the lead goal and budget, and design the label questions you’ll qualify with. |
| 2 · Live | Capture every conversation as it happens — snap a card, add a contact, record a voice note, or let the Booth capture for you. |
| 3 · Wrap-up | Before you leave the venue, review every contact and answer each label question. Minutes, not hours. |
| 4 · Debrief | Talk through the day once, out loud. The recording becomes contacts, notes, and follow-ups. |
| 5 · Follow-through | Work the follow-ups while the conversations are still warm — the record tells you who’s owed what, by when. |
| 6 · Prove | Read the record: conversations against goal, cost against qualified conversations, follow-ups against pipeline. |
Each stage has its own article in this series and a step-by-step guide in the playbooks.
What the model assumes
Every methodology has load-bearing assumptions. These are ours, stated plainly:
- Capture must cost under thirty seconds. If recording an interaction takes longer than that, reps stop doing it by the second day of the show. Every capture path is built against this bar — it’s the reason the phone, not the laptop, is the instrument of record.
- The rep in the moment is the best qualifier. “Did they engage? How deeply? Who are they to us?” — the person who just had the conversation can answer in seconds. Three weeks later, nobody can. Qualification is captured as answers to short questions, not inferred later from behavior.
- A complete record beats a perfect one. One unlabeled contact isn’t a crisis; a hundred is. The wrap-up exists so that every contact gets reviewed once, cheaply, before memory fades — including the honest answer “didn’t really engage.”
- The CRM stays the system of record. Ground truth is only useful where your team already works, so the record syncs into HubSpot — contacts, labels, notes — rather than living in a silo. (Athel is the system of capture, not a CRM.)
If those assumptions hold for your team — you sell in the field, you work events, your pipeline runs through a CRM — the rest of this series shows you how to run the loop. Start with why capture happens in the moment or not at all.