Attribution is an incomplete reconstruction
Customers move between devices, reject tracking, ask colleagues and return through branded search. No analytics model observes the entire journey. Attribution is therefore a structured estimate that should make decisions better without claiming impossible precision.
Start by defining what counts as a lead, opportunity and customer, and how sources will be named. Inconsistent campaign parameters and overwritten CRM fields create more uncertainty than the choice between first-click and last-click models.
Create a dependable evidence layer
Capture landing page, campaign identifiers and referrer with consent-aware methods, then preserve original acquisition context alongside the most recent meaningful interaction. Connect form, call and offline outcomes under stable records where practical.
Ask new customers how they heard about the business and record the answer in a controlled field. Self-reported data has bias, but often reveals referrals, podcasts, communities or remembered channels that browser tracking cannot see.
- Standardise UTMs and channel definitions.
- Store original and recent acquisition context.
- Connect marketing data to qualified CRM stages.
- Add self-reported attribution to key conversions.
Use models to guide, not dictate
Compare several views: first touch for discovery, opportunity creation for demand capture and assisted paths for influence. Evaluate incrementality through geographic, audience or timing tests when decisions are large enough to justify them.
Report assumptions and unknowns. Attribution should support a portfolio decision about where to learn and invest, not allocate every pound or dollar of revenue to one final click.