2

If I attempt to calculate a position based model off of the Multi Channel Funnel (MCF) report or API, this does not match what is reported in the position based model in the model comparison tool.

The reason I'm doing this is because I'd like to use the Position Based model in some other reporting, and would like to get that data in an automated fashion via the API. As the Model Comparison Tool (MCT) doesn't have an API, but the MCF does.

The MCF apparently has the required elements (via the mcf:firstInteractionValue, mcf:assistedValue, and mcf:lastInteractionValue metrics) (based off of this article). And Google calculates the Position Based model with 40% credit to first and last interaction, and 20% to middle interactions, I figured I could do the math on the MCF values and achieve the same number of the MCT. Which does not seem to be the case.

For the same source/campaign over the same period (a one week period a few weeks ago, to hopefully avoid data timing issues), doing this math off of the MCF results in a conversion value of $13,724.06. MCT reports $14,014.07 (MCT reporting $290.01 higher).

Example: Table of MCF Calculation

However, the MCF report First Click and Last Click value matches the MCT Fist Interaction and Last Interaction values. So the issue appears to be with assisted conversion value. I tested this with some other channels, and the difference was generally that MCT was higher, usually between 1% and 2%.

So, question:

  1. Anyone else trying to do this?
  2. Is this some sort of issue with sampling of the assisted values?

Any help would be appreciated, or at least an explanation of what is going on with Multi Channel Funnel assisted values vs. what I was expecting from Model Comparison Tool.

EDIT:

I created a custom model that attributes 100% of the conversion value to the "middle" (vs. 20% when used in the provided Position Based Model) to see what the MCT had for this value, as it appears to be the main issue. Looking at same time period as above example.

With this custom model, the "middle"/assisted is $15,514.40, whereas the MCF Assisted is reporting as $13,399.20.

So, Model Comparison Tool "Middle" is not the same as Multi Channel Funnel Assisted Conversions.

I'd like to know why this is the case, as I feel like they should be the same.

  • Hmm... Three thoughts that aren't quite an answer: 1.) Your API response will tell you whether that data is sampled (developers.google.com/analytics/devguides/reporting/mcf/v3/…); 2.) Did you subtract out First Interaction Value from Assisted Value? Assisted includes first; 3.) Do you have multiple kinds of conversions with values set? The API will return all. However, #2 will decrease the MCF value and #3 could only increase the MCT value, so if MCF is already smaller than MCT...they are not going to solve that. Let us know about sampling level! – Reve Jan 15 at 17:27
  • @reve: 1. Not sampled. – ryantkelly Jan 15 at 18:00
  • ...and I hit "add comment too quickly" @Reve: 2. No, I did not. And if I did, the 20% of (Assisted - First Interaction) would be crazy low, throwing the variance off even more. 3. I have multiple conversions, but only one with value (including the others in my MCF data did not change the conversion values), so that seems good. All great points, though I'm afraid looking like dead ends at the moment. – ryantkelly Jan 15 at 18:05
  • Well, one more though - is your lookback window the same? The API, as far as I can tell, will always use a 30-day window, and increasing that in the MCT would give the correct direction of difference. I'm assuming this isn't the issue but it's the one new idea I've had. :-) – Reve Jan 15 at 18:27
  • Yep, good thought/confirmation - but my web views are both with 30 day lookback. – ryantkelly Jan 15 at 18:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.