Campaign Attribution Readout

Turn a pasted campaign results export into an honest readout — what influenced pipeline, where the in-market accounts actually came from, and what to do next.

Demand Gen - Attribution

What surfaced after the campaign, what was already in-market, and what showed no intent

$420k
Influenced pipeline
58%
First observed post-campaign
Credit typePipeline
First observed post-campaign (no pre-campaign signal observed)$244k (58%)
Pre-existing intent$134k (32%)
Touched, no intent$42k (10%)
Why it lands
The campaign claims $420k influenced. The honest read: 32% of that pipeline sat with accounts already researching observability before the campaign touched them, and 10% never showed intent. $244k came from accounts with no pre-campaign signal observed that showed intent after it: coincident with the campaign, not caused by it, since the earlier signal may simply have been below HG's visibility threshold.

Overview

Take a pasted campaign results export and produce an attribution readout that cross-checks the campaign's influenced accounts against HG intent — splitting them into pre-existing intent, intent first observed post-campaign (no pre-campaign signal observed) and no intent — so the readout credits the campaign honestly, without claiming causation from timing.

Use cases

  • Attribution leadership believes

    Influenced-pipeline numbers get discounted because everyone knows they're inflated. Splitting influenced pipeline into pre-existing intent, intent first observed post-campaign (no pre-campaign signal observed) and no intent gives a credible number that survives scrutiny, without claiming timing proves causation.

  • Invest where the campaign actually works

    Knowing which segments showed intent first observed post-campaign — not just touched — tells you where to test next quarter's budget.

View workflow prompt
# Campaign Attribution Readout

## Parameters

- `{{campaign_results}}` *(required)* — Pasted campaign results (campaign, accounts touched, MQLs, influenced pipeline $, account domains). Example: `Q2 Observability Webinar - 184 accounts - 31 MQLs - $420k influenced`
- `{{category}}` *(required)* — Category whose intent contextualizes attribution. Example: `observability`

## Purpose
Produce an honest attribution readout from {{campaign_results}} by cross-checking influenced accounts against {{category}} intent — so the team can tell which accounts only showed intent after the campaign versus which were already shopping. Timing is not causation: intent that appears after the campaign is coincident with it, not proof the campaign produced it.

## Process
1. **Parse results** — read {{campaign_results}} (accounts, MQLs, influenced pipeline).
2. **Intent timing check** — resolve {{category}} to topic IDs with `list_intent_topics` (short keyword in `name`); optionally pre-screen all influenced domains in one `search_companies` call (`company_identifiers.domains: { ids: [<influenced domains>], inclusion_method: "ANY_PRESENT" }` + `intent.topics: { ids, inclusion_method: "ANY_PRESENT" }`); then, per influenced account, two `company_intent` calls with `topic_ids`, `fields: ["topics"]` and `granularity: "global"`: one with `signal_date` ending the day before the campaign start (date from {{campaign_results}}; the windows must not share a day), one with `signal_date: { from: <campaign start>, to: <end of attribution window, else today> }`.
3. **Honest credit** — classify each influenced account into one of three buckets: 'pre-existing' (the pre-campaign window returned topic rows), 'first observed post-campaign (no pre-campaign signal observed)' (the pre-campaign window returned no rows, `topics.count` 0, and the post-campaign window returned rows), 'touched-no-intent' (no rows in either window and no `search_companies` pre-screen match). If a `company_intent` call errors (as opposed to returning no rows), report the account as 'not classified'; the same goes for an account that matches the `search_companies` pre-screen but has rows in neither window. Split influenced pipeline across the three buckets, list not-classified accounts and their pipeline separately outside that split, and state once: absence before the campaign can also mean the signal was below HG's visibility threshold, so the post-campaign bucket shows timing, not causation, and its pipeline is coincident with the campaign, never caused by it.
4. **Segment performance** — `company_firmographic` to see which segments the campaign worked best in.
5. **Next-action** — recommend what to double down on and what to cut based on the honest read.

## Output Format
Markdown with:
- `# Campaign Attribution Readout`
- `## Headline Numbers` (from the export)
- `## Pre-existing vs. First Observed Post-Campaign (No Pre-Campaign Signal Observed) vs. No Intent` (three-way intent-timing split, stating the visibility-threshold caveat once and that timing is not causation)
- `## Segment Performance` (where it worked)
- `## Next Actions` (double down / cut)
- `## Citations`

## Quality Checklist
- The three-way split cites both `company_intent` windows; only accounts with no pre-campaign rows and post-campaign rows are counted as first observed post-campaign, labelled "no pre-campaign signal observed", with the visibility-threshold caveat stated once
- The readout credits the campaign without overclaiming pre-existing intent, and calls post-campaign pipeline coincident with the campaign, never caused by it
- Segment performance cites `company_firmographic`
- Next actions follow from the honest read