Attribution
How do I measure what the advertising actually achieved?
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.
Can Metrunio measure the effect of advertising, not just assign it?
Note
The switchback trial comes with Growth, two campaign groups with Scale and the agency plan. On Start and Pro, “Experiments” shows in the navigation with a lock; the page there says what it does and leads to “Plan & usage”. On Growth, “Two campaign groups” shows in the selection with a lock “from Scale”. The “measured” level under “Check budget” comes only from two campaign groups. Experiments you created stay stored and reappear as soon as your plan includes them.
| Design | What is compared | For what | From plan |
|---|---|---|---|
| Switchback trial (on and off) | periods of the same scope | Does this channel contribute at all? | Growth |
| Two campaign groups | two groups in the same period | Which of the two setups works? | Scale |
- 01Open “Experiments” and create one with a hypothesis: what you expect, written down before the result.
- 02For the switchback trial, enter the periods when the advertising ran and when it was off. Switch at least twice.
- 03For the group comparison, assign each group its campaigns.
- 04Wait out the period and read the result. It states the verdict in one sentence, the numbers behind it, and the caveat.
What “no measurable difference” means
A full result, not an error and not a placeholder. It means: the observed difference cannot be told apart from noise. Metrunio then extrapolates nothing either. “Plus four per cent” from twelve conversions reads like a result and is not one.
Caution
For small shops a switchback trial often ends with “no measurable difference” or “not enough data yet”. It takes at least 100 opportunities per group and 10 conversions in total, and considerably more for a narrow interval. With around 100 orders a month, the difference made by a single campaign is usually smaller than the noise. The result then says “not shown”, not “does not work”. The planning from “Check budget” tells you beforehand how long the trial would need to run at your pace.
Why this is not a geo experiment
There is no separate design for it, because Metrunio does not know the region of a visit: it stores no IP address. If you target geographically, split your campaigns into two groups; then “two campaign groups” measures exactly that. It is not a geo experiment in the strict sense: that would need a control region with no ad exposure at all, and Metrunio would only see its purchases if measurement ran there anyway. Comparing two advertised groups tells you which did better, not what would have happened without advertising. The region sits in the campaign instead of in a guessed attribution.
Caution
No calculation can check whether the groups differed only in the thing being tested. A promotion, a public holiday, a press mention: all of it would land in the result. The caveat therefore sits on every result, so that it travels with it when you pass it on.
Note
The confidence interval follows from the principle that every figure says what it rests on: it states explicitly how certain a measured number is. Without an interval the difference would be a claim.
Last reviewed on 17 Oct 2026
In the customer portal
Also asked
- Measure incrementality
- What does the channel really contribute?
- Holdout test
- A/B test for advertising
- Causal advertising effect
- Which plan includes experiments?
- Why are experiments locked for me?
- Which plan includes the switchback trial?
- Is my Meta advertising worth it at all?
- Pause advertising and compare
- On-off test for advertising
Related articles
- 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.
- Where does measurement end?
What are the limits of Metrunio’s measurement?
Post-view is only measurable where your own tags may be served and the browser returns the view identifier: not every served ad is viewable, and not every viewable one can be linked to the purchase. Without consent, gaps arise deliberately, and a gap stays a gap, not a zero. Attribution describes assignment, not causal advertising effect.
- What do “measured” and “attributed” mean?
Why does Metrunio separate measured from attributed figures?
Measured means directly observed: an impression, a viewable impression, a click, a conversion. Attributed means a share distributed by a model; the product keeps both layers apart and never blends them. Alongside them are values you decide yourself (“set”), answers from your customers (“self-reported”) and figures an ad platform reports about itself (“reported”); there is no “estimated” layer.
- Which campaign actually delivered?
What do I move my budget on, if attribution does not measure effect?
The “Check budget” area sorts your campaigns into three levels: measured (an experiment found a difference), attributed (a model credits it a share) and not answerable (with a reason). Only the top level says anything about effect, and it always carries its interval.
- What does Metrunio cost?
Which plans are there, and what happens if I go over?
Five plans, tiered by events per month: Start (free), Pro (€79), Growth (€249), Scale (€599) and Agency (€299 with two client accounts, €49 per further one, 250,000 events per client account); every four delivered impressions count as one event. Above the allowance, every started 10,000 events cost €3, and a plan never costs more than the next plan up would cost at the same volume.