Labels are your qualification system
Every label group asks one question a rep can answer in the moment. Every answer becomes a timestamped interaction — the atoms your attribution is built from.
For field marketers & booth teams · company admins · marketing ops · field reps & aes · 4 min read · Updated July 10, 2026
Most teams qualify event leads the way they inherited from their CRM: a rep stares at a name three weeks later and picks a lifecycle stage. The stage is a guess, the guess is stale, and everyone downstream knows it — which is why nobody trusts event-sourced pipeline numbers.
Ground-Truth Attribution replaces the guess with a question — asked at the moment the rep actually knows the answer.
Groups ask questions, labels answer them
In Athel, interaction labels are organized into label groups, and every group poses exactly one plain-English question. Qualifying a contact isn’t “assign the right taxonomy value” — it’s “answer four short questions about a conversation you just had.”
The default set looks like this:
| Group | The question | Select | Example labels |
|---|---|---|---|
| Engagement Status | Did they engage? | one | Engaged · Brief hello · Didn’t engage |
| Engagement Depth | How deeply? | one | Full demo · Real conversation · Small talk |
| Funnel Intent | Who are they to us? | multiple | Decision maker · Influencer · Existing customer · Partner |
| Outcome | What happened? | one | Meeting booked · Send pricing · Not a fit · Follow up next quarter |
The labels above are examples — you define your own per group, and the templates section ships starter kits per event type. The questions are the load-bearing part: a rep who just finished a conversation can answer all four in under ten seconds, and their answers are worth more than any stage a database guesses at later.
Answers become interactions — the attribution atoms
Here’s the part that makes labels more than tags: every answer is stored as its own interaction record — this person, this Event, this label, this timestamp.
That structure is what makes B2B attribution readable later:
- The same person can hold different answers at different Events. Sarah was Brief hello at the spring show and Full demo · Decision maker at the fall one — both facts survive, in order.
- An account’s story assembles itself. Three people from one company across two Events isn’t a data-cleanup problem; it’s three timestamped threads that together explain why the deal exists.
- Nothing is overwritten. Qualification isn’t a single field that mutates; it’s an append-only history of what actually happened.
When people ask what “ground truth” means concretely — it’s this table of interactions. Everything in the attribution article is arithmetic over these rows.
Designing a good label system
A few rules, learned the hard way:
- One question per group, phrased as a rep would say it. If the question needs explaining at the booth, it’s wrong. “Did they engage?” works. “Assign MQL disposition” does not.
- Fewer labels, sharper edges. Five labels a rep can distinguish while standing beat twelve that need a rubric. If two labels are ever debated, merge them.
- Single-select by default. Multi-select is for genuinely plural facts (a person can be both Decision maker and Existing customer). Everything else should force a choice — forced choices are information.
- Always include the honest exit. “Didn’t engage” and “Not a fit” are some of the most valuable labels in the system. A no recorded on day one saves an hour of zombie follow-up later — and in the wrap-up flow, choosing a skip label like Didn’t engage ends the review for that contact so you’re never grinding through questions that no longer apply.
- Make labels mean something in your CRM. Labels can mirror HubSpot association labels, so “High intent at the fall show” isn’t an Athel-only fact — it’s a segment your marketing team can build lists and reports on. The HubSpot playbooks cover the mechanics.
The assumption, stated plainly
This system assumes reps answer honestly and in the moment. Both hold better than you’d expect, for one design reason: the questions are about what happened, not about forecast credit. “Did they engage — yes or no?” has no incentive game in it, unlike “is this an SQL?” And answering at capture or at wrap-up — hours, not weeks, after the conversation — means the answer comes from memory that still exists.
Design the questions once, before the Event, as part of prep. Then the qualification takes care of itself, one tap at a time — and the review pass that guarantees every contact gets those taps is the wrap-up ritual.