Marketing attribution: crediting channels for outcomes
Marketing attribution is the problem of deciding which channels deserve credit for an outcome when a customer touched several before converting. It sounds like accounting, but it shapes where budget goes — and because journeys are tangled, no method answers it perfectly.
Last updated
The attribution problem
Customers rarely convert on a single touch. Someone might find you through a search, return via a social post, click an ad weeks later, and finally convert from an email. When the sale lands, which of those gets the credit? That is the attribution problem, and it has no clean answer.
It matters because credit drives budget. If you credit only the last touch, you may underfund the channels that started the relationship; if you credit only the first, you may starve the ones that closed it. How you assign credit quietly steers what you invest in.
Attribution models
Attribution models are different rules for splitting credit. Last-touch gives it all to the final interaction; first-touch to the initial one. Linear splits it evenly across every touch; others weight toward the first and last, or decay credit by how recent each touch was.
Each model tells a different story from the same journey, and each has a bias. Simple models are easy but crude; more elaborate ones capture nuance but are harder to trust and explain. There is no single correct model, only ones more or less suited to a particular question.
Living with imperfection
Attribution is inherently approximate. Some touches are invisible — offline conversations, word of mouth, an ad seen but not clicked — and privacy changes are shrinking what can be tracked at all. Any attribution picture is a partial reconstruction, not the truth.
The pragmatic stance is to use attribution as a directional guide rather than a verdict, to understand the bias of whatever model you rely on, and to sense-check it against outcomes you can see. Treated as one imperfect input among several, it informs decisions; treated as gospel, it misleads.
How Kirti approaches this
Kirti is being built to measure across the six surfaces it runs and to read attribution as directional evidence for where a brand’s effort pays off — informing the next plan rather than pretending to a precision the data does not support.
The agent is in active development and not yet generally available; the realities of attribution apply whatever analytics and models you use.