Every attribution model apportions credit using assumptions. First-touch assumes discovery matters most; last-touch assumes the final step does; linear assumes everything counts equally. None of these are true — they are choices, and the choice determines the answer.
Why this matters commercially
Budget moves on these numbers. If the model over-credits the last click, brand and demand work look like waste and get cut, which reduces the demand the last click was converting. The model does not merely describe performance; it shapes it.
What to instrument
- Source capture at first touch, stored on the record rather than inferred later.
- The connection between marketing source and CRM outcome, which is where most implementations break.
- Offline and assisted conversions, since a phone call is invisible to platform reporting.
- Time-to-close by source, because a cheap slow source and an expensive fast one are different businesses.
- Lead quality alongside volume, so an improving cost per lead cannot hide a falling close rate.
What attribution cannot do
It cannot measure the influence of things that produce no tracked interaction — a mention in an AI answer, a recommendation from a peer, a conference conversation. Those are real and they land as branded search or direct traffic, which most models credit to the last click. Knowing this is the difference between using attribution and being misled by it.
Why platform-reported numbers disagree
Each platform counts what it influenced by its own rules, with generous windows and view-through credit. Summed across platforms they routinely exceed total conversions. Reconciling them against CRM outcomes is not an anomaly to explain away; it is the core work.
What this means for your business
Build attribution before spend scales, not after — retrofitting means reconstructing history from data never collected for the purpose. Our analysis of CRM problems covers the data foundation this depends on.