On Monday a customer arrives in your shop through an Instagram ad and looks at teapots. On Thursday she searches Google for your name and clicks the search ad. On Saturday she types your address straight in and orders for €120. Whose order is it?
That depends on a rule you choose. Such a rule is called an attribution model: it distributes an order’s revenue across the measured touchpoints before it. Metrunio has seven.
The calculation
| Model | Instagram (Mon) | Google (Thu) | Direct (Sat) |
|---|---|---|---|
| First touch | €120 | €0 | €0 |
| Last touch | €0 | €0 | €120 |
| Last non-direct click | €0 | €120 | €0 |
| Linear | €40 | €40 | €40 |
| Position-based | €48 | €24 | €48 |
| Click before view | €0 | €120 | €0 |
Position-based gives the first and the last touchpoint 40 per cent each and splits the remaining 20 per cent across the touchpoints in between. “Last non-direct click” skips the direct visit because a visit through an ad comes before it. “Click before view” calculates the same here, because there is no ad that was seen without a click.
The seventh, the data-driven model, sets no fixed rule. It works out from your own paths, including those without an order, how much attributed value would drop out of the model if a channel were missing. It needs a basis for that: 100 paths, 20 conversions and two channels within 90 days. Until then it does not appear, and nothing is spread evenly in its place.
When an ad was only seen
A second case shows why the names of the models can mislead. Someone sees a display ad in the morning, does not click it, comes to the shop directly in the evening and buys for €80. With post-view, which comes with the Pro plan and needs your own tag on the ad, “Click before view” gives the €80 to the ad that was seen: there is no click, so the viewable impression counts.
“Last non-direct click” never considers viewable impressions. It gives the €80 to the direct visit, because that is the only touchpoint of the kind the model distributes to. Under the model with “non-direct” in its name, “Direct” gets more here.
Which model is right?
None can be derived from the data as the right one. Every fixed model is an assumption about which touchpoint counts for how much. More useful than searching for the right one is a different question: does the choice of model change your decision?
Metrunio’s model comparison puts two models side by side on the same orders and gives the change in percentage points per channel. The table with all models shows a range per channel, from the smallest to the largest share. If it is narrow, the channel’s figure hardly depends on the model. If it is wide, your decision rests on an assumption, and you write that down when you pass the figure on.
The comparison comes with the Pro plan. On the free Start plan Metrunio calculates with “Last non-direct click” and says which channel would shift most under a different model.
Fix the model before you compare. Picking afterwards the model that puts your favourite campaign in front gives you a justification, not an answer. And none of these figures says whether the order would have failed to happen without the Instagram ad. Only an experiment with a control group answers that.
À retenir
In the example, Instagram gets €0, €40, €48 or €120 of the same order depending on the model. The same three touchpoints were measured in every case.