ATTRIBUTION
Seven attribution models. One set of data.
A model is an assumption, not a truth. That is why Metrunio sets two models against each other for the same week and shows all seven in one table, with and without post-view. Six of them are rules we disclose. The seventh reads the rule from your own paths. Model comparison and post-view come with Pro and above; on the free plan Metrunio works with “Last non-direct click”.
01 / MODELS
The seven models.
MODEL 01
First touch
The first measured touchpoint receives the entire share.
MODEL 02
Last touch
The last touchpoint before the conversion receives the entire share.
MODEL 03
Last non-direct click
A direct visit does not override a known contact, and a viewable impression never receives a share in this model, not even with post-view.
MODEL 04
Linear
All touchpoints share equally.
MODEL 05
Position-based
First and last touchpoint 40 percent each, the rest share 20 percent.
MODEL 06
Click before view
Clicks take precedence. Without a click, a visible impression counts.
MODEL 07
Data-driven
The rule is read from your own paths, including those without a conversion. It appears only once there are enough paths; below that nothing is distributed evenly as a substitute.
If a model finds matching touchpoints, it distributes the conversion’s full value across them. If it finds none, the conversion stays “unattributed” and is shown separately; under “Last non-direct click”, for example, that happens after viewable impressions only. Attributed and unattributed together add up to all conversions it calculates; that is verified automatically, not asserted. A visit is one touchpoint, however many pages are viewed in it. The same visit only adds another with a new campaign signal.
Estimate how far the platform figures diverge from your orders or leads
02 / POST-VIEW
What post-view can do and where it ends.
Post-view credits a conversion with ads that were seen but not clicked, measured with your own impression and viewability tag. Each view carries the website the ad ran on: passed by the ad server, otherwise taken from the referrer.
That works only where a tag of your own may be served, and only the viewable impression counts. A viewable impression can only be linked to the purchase in browsers that keep its identifier: Chrome and Edge, not Safari, Firefox and Brave. The linkable share is shown next to it as post-view coverage. Where a signal is missing, no contact is filled in, and attribution describes assignment, not cause. All limits of measurement
DECISION POINT
Compare click-only against post-view before you move budget. The difference is the actual finding.
03 / EFFECT
What attribution cannot answer.
No attribution model tells you whether the conversion would have happened without the advertising. A suitably designed trial can examine that; how robust its result is depends on the design and on the uncertainty.
Two designs
Alternating periods (on, off, on) or two campaign groups side by side. Which design fits depends on what can be cleanly separated in your advertising mix.
An interval, not a single figure
The result is a range with its uncertainty. If it contains zero, the verdict is “no measurable difference”, and nothing is extrapolated.
Measured, not modelled
Other tools estimate effect from a regression over past weeks. Here you switch something off, and Metrunio measures what happens.
How it connects to attribution
Attribution shows which campaign looks suspect. The experiment tests whether the suspicion holds.
The switchback trial comes with Growth, two campaign groups with Scale and the agency plan.
No geo experiment: Metrunio does not know the region of a visit and does not guess it. If you steer geographically, you can compare two campaign groups. That tells you which did better, not what would have happened without advertising.
AND THEN?
In the portal your campaigns are sorted by what is known about them: first those an experiment says something measured about, then those a model credits a share to, then the open ones. For an attributed campaign the next step is not “move budget” but “start experiment”. That is the whole difference from a tool that outputs a daily model figure.