Design First-Touch vs Multi-Touch Attribution Rules

You are a marketing analytics consultant for digital marketing agencies. You translate attribution theory into operating rules executives can defend.
Agency: [AGENCY_NAME] — [SERVICES] | CRM: [CRM_TOOL] | Analytics stack: [ANALYTICS_STACK] | Channels: [CHANNELS] | Avg sales cycle: [SALES_CYCLE] | Avg touches before close: [TOUCH_COUNT]
Define when to use first-touch, last-touch, linear, and U-shaped attribution. Tie each model to a specific business decision (budget, hiring, commission, exec reporting).

- Each model must be tied to ONE primary decision, not 'general reporting'
- Self-reported attribution must reconcile with platform attribution — explain how
- No double-counting credit across models in the same report
- Rules must be simple enough to apply without a data analyst in the room

1) Decision -> Model matrix (Decision | Model | Why | Owner)
2) Reconciliation logic: when platform and self-reported disagree, which wins and why
3) One-page exec summary the founder can show to the board

• [AGENCY_NAME] — Your agency name
• [SERVICES] — Service lines
• [CRM_TOOL] — Your CRM
• [ANALYTICS_STACK] — GA4, HubSpot, Dreamdata, Plausible, etc.
• [CHANNELS] — Channels driving pipeline
• [SALES_CYCLE] — Average days from first touch to close
• [TOUCH_COUNT] — Average marketing touches before a deal closes

Agency: Northbeam Studio — SEO + paid social for DTC | CRM: HubSpot | Stack: HubSpot + GA4 + Dreamdata | Channels: SEO, podcast, LinkedIn outbound, partner | Cycle: 47 days | Avg touches: 11

DECISION -> MODEL
Budget allocation -> Linear (every channel that touched a closed deal gets fractional credit). Owner: Head of Growth.
Channel hiring -> First-touch (which channel originates pipeline). Owner: Founder.
Sales commission -> Last-touch outbound only (SDR who booked the meeting). Owner: Head of Sales.
Board reporting -> Self-reported + first-touch (one chart, two bars). Owner: CFO.

RECONCILIATION
If self-reported 'Heard from a friend' but UTM shows google/cpc — self-reported wins for source category, paid wins for assisted credit. Never both in the same revenue bar.

EXEC SUMMARY
We use 4 models, each for one decision. We never average them. If two models disagree, we report both and name the decision each is built for.

• Pin the decision-to-model matrix in your reporting dashboard so nobody asks 'why is paid bigger here than there' again
• Self-reported attribution beats platform attribution for source-of-truth — but only for deals over your AOV
• Never tie commission to a multi-touch model — reps will not trust math they cannot explain to their spouse