The research preview

Attribution that knows where every touchpoint belongs.

Most attribution models know what happened and when. Attribyte also models how the touchpoint sits inside the marketing system around it.

RESEARCH PREVIEW · HIERARCHY-PRESERVING ATTRIBUTIONSee the method

Why structure matters

A journey is more than a list of timestamps.

A paid-search keyword belongs to an ad group. The ad group belongs to a campaign. The campaign sits inside an account. Flatten that into rows and useful relationships disappear.

Typed marketing hierarchyAn account contains campaigns, ad groups, and keywords connected to a customer journey and outcome.AccountCampaignAd groupKeywordJourneyOutcomeCRM bridge

The method

Keep the system around the journey.

01

Build the marketing graph

Represent accounts, campaigns, ad groups, ads, keywords, audiences, creatives, landing pages, CRM stages, and outcomes as related entities.

02

Preserve the hierarchy

Place entities in a structure-aware representation designed for branching systems.

03

Place the journey inside structure and time

Each event retains where it belongs and when it happened.

04

Calculate contribution and conserve the total

Score event contribution, then roll it to selected hierarchy views without creating duplicate conversion value.

Designed to preserve

Relationships, time, bridges, and conservation.

A touchpoint can be neither first nor last and still connect two important branches. The model can measure that structural mediation and compare it with other contribution methods.

  • Parent-child · ancestor-descendant · sibling · cross-branch
  • Journey timing · bridge touchpoints · graph version
  • Credit conservation across keyword, ad-group, campaign, channel, or account views

Example outcome

$10,000

Persistent total

Keyword view

Different view, same outcome.

Ad-group view

Different view, same outcome.

Campaign view

Different view, same outcome.

One outcome. Multiple views. No duplicate credit.

What it is not: proof that a touchpoint caused the outcome, a guarantee of outperforming every model, or a repair for missing identity, revenue, or event data.

Validation

A research preview should survive scrutiny.

  • Hierarchy distortion
  • Stability under updates
  • Predictive / holdout performance
  • Model disagreement
  • Conservation invariants
  • Operator interpretability

Help validate a new attribution model.

Bring a complex journey, a real hierarchy, and outcomes you can verify.