On Voluum, what is a "Visit" on an Offer?

[EricNguyen]

I noticed that in the voluum Offers report there is a "Visits" column. I'm curious what this number represents when I have landing pages with multiple offers. I was primarily interested in Voluum's machine learning functionality with regard to multiple offers per campaign but I believe it presents a reporting issue.

What I suspect is that a "visit" (as opposed to a "click") to an offer is actually a visit to a landing page for which the reported offer is downstream from that landing page. What I can't figure out is whether the visit is implied (offer A and offer B are within the flow/path and are therefore counted as a visit) or inferred (offer A is weighted at 40% and therefore counts as .4 visits, offer B .6 visits). When it comes to Voluum's "AI" it makes it even more complicated.

Sorry to nerd out on this. I'm wondering if the Offers report should actually omit the Visits column as it only seems to provide misleading information for flows/paths with more than one offer. My apologies if I've missed something obvious!

 

[Graybeard]

A visit would be some webpage that the user opened (in general).

AI WTF!
AI is in? What happened yesterday may not happen today
Or, are they expressing some (supposed) predictive analysis (and based on what).
Excuse my cynicism ...

 

[EricNguyen]

Thanks for replying!

They mentioned the bayesian algorithm by name in their promotional video for AI (though it's also referred to as Machine Learning which is a subset of AI). I understood that part (though I wonder if it should be a combination of Decision Tree and Naive Bayesian). I'm probably coming from a cynical standpoint as well but optimizing campaigns is my jam. I don't mean any disrespect because Voluum looks like a very nice product. Maybe the Voluum product manager can clarify with regard to what "Visits" represents in the Offer (and Affiliate Network) report.

If you have Landing Page "A" with three CTAs to three different Offers (we'll call them 1, 2, and 3), when the user lands on Landing Page "A" is that "visit" counted against each Offer 1, 2, and 3 for the purpose of reporting? It's kind of an interesting chicken/egg problem. The only way to report Visits over Conversions is to make sure you only have one offer per landing page; otherwise you're generating bias with landing page design. It's an interesting problem and one I don't think I fully understand.

Unless I'm mistaken, there is no such thing as an Offer "visit" unless a landing page only presents a single offer.

 

[EricNguyen]

I think you're dead on about rotating the presentation.

If I understand you correctly, are you saying to assign "visit" value based pro-rata on offer clicks? That's very interesting.

If you have 300 known landing page visits presenting offers A, B, C, one could infer visit credit pro-rata using your percentages:

Offer A: .167 visits per landing page view
Offer B: .507 visits per landing page view

etc.

Is that crazy? I don't think this is how Voluum is calculating it. Still, though, good thinking!

 

[Graybeard]

For that purpose I would simply devise a sales funnel and use my own domain redirection counters.

The bias is #1 or #2 (probably)
Change the order of presentation to prove that out?

Click A,B, or C is to a unique URI/URL and the 302 (redirections) are recorded.

So now I see; 300 visits to my landing page and the number of clicks each CTA gets on that landing page.
You can deduce the percentages including the page abandonments. A 50, B 152, C 20, +78 Page drops or abandonmentc

barry@paragon-DS-7:~$ bc <<< '300-(50+152+20)'
78

50/300
.16666666666666666666
152/300
.50666666666666666666
20/300
.06666666666666666666
78/300
.26000000000000000000

If you can get a pixel fired back or some other response from the remote source then you would have the remote page load too. The remote page could post back that client's access to the buy page and further to the purchase completion acknowledge page (in theory). In a perfect world ...

But if you had all of this data available you could create accurate funnels and predictions if you scale was statisticall accurate.

The key is data transparency