Los artículos de ayuda existen en alemán e inglés. Aquí los lees en inglés.
Atribución
How do I read the model comparison?
The page compares two models on the same conversions: per channel, the share under model A (your default model) and under model B, plus the change in percentage points. The differences come from the distribution rule and are no evidence of additional advertising effect.
Seven models distribute the same conversions differently. How do I read the comparison?
Note
The model comparison is available from the Pro plan. On the Start plan it is replaced by one model, plus a sentence that names, from your data, which channel would shift most under another model.
Two models side by side
- Model A is your default model from “Settings”; model B defaults to “Click before view”, or to “First touch” without post-view, because the two click models then calculate alike. You can choose both.
- Share A and share B refer to all conversions the model distributed, including the unattributed ones. The “Unattributed” row appears as soon as either model has any.
- The change is share B minus share A in percentage points: 20 per cent and 25 per cent give +5.0 percentage points, not +25 per cent.
- “View touchpoints” opens the journeys with a touchpoint via that channel, in the same period.
- Period, conversion type, both models and the post-view setting are stated in one line above the table. The choice is part of the address and can be saved as a view; it does not change the default model.
Caution
The differences come from the distribution rule. They are no evidence of additional advertising effect, and the page derives no budget recommendation from them. Whether a channel causes additional purchases is a question for an experiment.
“Unattributed” means conversions the model could not attribute to any touchpoint: none was measured, all measured ones lay outside the attribution windows, or the model does not consider the ones there are, such as viewable impressions under “Last non-direct click”. The same set carries the same name in the overview, the conversion list and the model comparison, and always refers to the selected model.
All models side by side are one click away under “All models and method”: the same table as before, with post-view and click only, the basis per model and the removal effect. The last column, “Range”, gives the smallest and the largest share per channel across the models in the table, such as “12.0–18.5”. The wider it is, the more the channel’s figure depends on the choice of model; it does not prove any effect. Shares have one decimal place because a model splits a conversion across several touchpoints. What follows applies to the table.
| Observation | What it says about the channel |
|---|---|
| Strong on “first touch”, weak on “last touch” | often stands at the start of paths |
| Weak on “first touch”, strong on “last touch” | often stands just before completion |
| Similar across all models | its figure hardly depends on the model choice |
| Visible only with post-view | the conversions are preceded mainly by viewable impressions without a click; whether they had an effect only an experiment shows |
One model is rarely enough for a budget decision. The more useful question is: does the model choice change my decision? If not, the decision is robust. If it does, the decision is an assumption and should be named as one.
Caution
Choosing a model afterwards because it delivers the desired result is not analysis. Fix the model before the comparison and document the choice.
If all models return exactly the same value, that is not a calculation error. It happens when no journey holds more than one attributable touchpoint: a model with nothing to distribute distributes the same way in every variant. The portal says so explicitly in that case.
If only six columns appear instead of seven, the data-driven model is missing. It is the only one that needs the whole set of paths, and it appears once there are enough of them, including paths without a conversion. The basis for it is recalculated once a day: if you cross the threshold today, the model appears the next day, not within the hour. While it is missing, nothing is distributed evenly as a substitute.
When the models do not differ, these four steps lead to the cause:
- 01Open the journey list and check how many journeys hold more than one touchpoint.
- 02If almost none do, signals are missing: check under “Data status” whether events were blocked without consent or not stored because of an invalid format.
- 03Check under “Setup” whether every channel reports impressions and clicks. A channel without measurement appears in no journey.
- 04Widen the click window if your purchase decision typically takes longer than the configured period.
Revisado por última vez el 14 oct 2026
En el portal del cliente
También se pregunta
- Why do I only see one model?
- Which model is correct?
- Compare models
- What does the spread tell me?
- Why does every model return the same value?
- All models show the same number
- Model comparison shows no differences
- The models show different numbers
- Seven different numbers for the same week
- Why do I only see six models?
- Compare two models
- What does the change in percentage points mean?
- What does “Unattributed” mean in the model comparison?
- Where do I see all models?
- What does “View touchpoints” show?
- What does the “Range” column mean?
- How much does a channel depend on the model?
Artículos relacionados
- How do the attribution models calculate?
Which models exist and how do they distribute the share?
There are seven models: six distribute a conversion by a fixed, disclosed rule, and the seventh reads the rule from your own paths. 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. Attributed and unattributed together add up to all conversions it calculates.
- What is post-view and what does it tell me?
Why should I count viewable impressions without a click?
Post-view means: a viewable ad that was not followed by a click is linked to a later order or lead and taken into account by your model. That gives a number to contacts click measurement cannot see, display in particular; it only works where your own tag runs, both sides have consented and the browser returns the view identifier. Viewable impressions receive a share from the Pro plan, and how much they count is decided by model and view window.
- Which metrics exist and how are they calculated?
How are ROAS, CPA and conversion rate derived?
Metrics come from measured quantities and, where attribution is involved, from the chosen model. Cost-based metrics require ad spend to have been imported or synced.
- How do I measure what the advertising actually achieved?
Can Metrunio measure the effect of advertising, not just assign it?
With an experiment: the switchback trial is available from Growth, two campaign groups from Scale and on the agency plan. An experiment compares two groups that should differ in exactly one thing and measures the difference along with its uncertainty. Attribution does not answer this: it says which touchpoint a conversion is credited to, not whether it would have happened without the advertising.
- How do I save a view of a report?
Can I save a report’s period, filters and model and open them again later?
Under “Saved views” next to the period selection you store the current report with its period, filters and model under a name and open it again with one click. The view belongs to you alone and does not change the tenant’s default model.