How Tailkey's attribution modeling works
A look at how Tailkey's model turns raw visitor activity into content-level credit, and why it works differently from the attribution models you've used before.
4 sections · 6 minutes read
Many channels. One journey. Credit at content level.
The journey we follow all the way down this pageChannels are only how someone arrives — the same person can arrive through several of them. What matters is the journey underneath: the pages they move through and, inside those pages, the specific blocks of content that actually do the convincing. Every chapter below scores this same journey.
01 — The basics
Attribution modeling, in plain terms
Buyers rarely convert on the first visit. They arrive through search, come back from an ad, return through an email, and read several pages in between — sometimes across one afternoon, sometimes across months. Attribution modeling is simply the method used to decide how much credit each of those touchpoints deserves for the conversion that eventually happens.
Touchpoints
Every page, ad, email, or piece of content a visitor interacts with on the way to converting.
Paths
The full sequence of touchpoints a single visitor moves through, not just one moment looked at alone.
Credit
The share of the outcome assigned to each touchpoint, reflecting how much it actually contributed.
Touchpoints and paths are easy to record. Credit is where almost every attribution tool starts guessing.
02 — The problem
Most attribution models guess. Yours shouldn't have to.
Almost every model in wide use today shares the same flaw: it hands out credit using a rule that was decided in advance, based only on where a touchpoint sits in the sequence — never on what happened there. Switch between them below and watch the same journey get a completely different story each time.
None of the first five measure anything. They're rules, applied the same way regardless of what your visitors actually did — which is why the comparison page can look essential in one model and invisible in the next, without a single thing changing on your site.
Tailkey starts from a different place entirely — and it starts by changing what a touchpoint even is.
03 — The shift
Source, page, content block
Most attribution tools stop at the source — something like "organic search" or "paid social." Better ones reach the page. Neither tells you what actually convinced anyone. Tailkey resolves credit at all three levels, so a source becomes the pages that earned it, and a page becomes the specific blocks of content inside it. Step down a level at a time:
That's the what. Here's how an arrival from any channel becomes a score on a single paragraph.
04 — Under the hood
Every touchpoint, scored on real behavior.
Tailkey doesn't apply a fixed rule. It captures every touchpoint, reconstructs the full path, and scores each one by how much it actually moved the outcome. Four stages, in order:
We record every source, page, and block of content a visitor encounters, not just the first or last thing they saw — including the parts that never get a click, like a pricing FAQ read halfway down a page.
The model
Not one rule. A blended model, always learning.
Most attribution tools pick a single rule — first touch, last touch, or an even split — and apply it everywhere. Tailkey blends multiple established attribution techniques, including path-based modelling that compares converting and non-converting journeys and fair-credit allocation drawn from game theory, then layers in the further factors set out in our Technical Disclosure: measured engagement with each block, the shape of the journey around it, content type and position, and recurrence across other converting paths. Machine learning re-weights that blend continuously against your real outcomes, so the model you run next month is not the model you run today.
Multiple methods, one score
Established attribution techniques are combined rather than used alone, so no single method's blind spots dominate the result.
Real engagement, not just clicks
Dwell time, scroll depth, and on-page interaction feed the model, so a page that's actually read counts for more than one that's merely visited.
Re-weighted by ML, continuously
The blend is fit and re-fit on which journeys actually convert, so the weighting keeps adjusting as your content and audience change instead of staying fixed.
See it run on your own journeys.
Add Tailkey to your site and find out which pages and paragraphs are actually converting your visitors — scored down to the content block, within your first month.