Construct a reasonable assumption from a corollary.

[RobertFisher]

Most are aware when doing work on Local citation sources, if a phone number is to be used as a tracking number and it is not the main number used for the business in Local, it should be an image without alt text identifying the number. With a major citation source like Yellow Pages, we choose to use the regular business number which, of course, creates an attribution problem. If the same number is used in organic and Local, where did the traffic come from: YP? organic listings? Local? G+?

With one client, we use a sophisticated proprietary call tracking mechanism that is the property of the franchisor. We also utilize GA and WMT. My question is: Is there a way to draw a reasonable corollary to other (non-call) conversions from overall traffic?

If we know that for referring sites using an image number that those sites generate as many calls as they do contact forms filled out can we draw a reasonable corollary that other sites would be similar? On my example I have included, Angies list shows 2 visits and one form submitted. The next referrer shows 7 visits and 2 forms submitted. If each of these referrers are using image numbers and we know we got one call from Angies and 2 calls from the other, can we draw a reasonable assumption of: YP had 5 forms submitted, they would have had close to 5 calls additionally?

So, if forms to calls for A is equal, and forms to calls for B is equal, can we assume that for a third or fourth referrer, calls to forms would be near equal with rare exceptions?

If not, is there a way you are aware of to statistically draw an inference from this situation?

Obviously, the big unknown is do people from one referrer have a significantly different motivation, etc. than those from another? One thing we know with this client is there is no affiliate traffic referring so there is little likelihood one or two referrers are distorting the numbers.

I realize this takes some thinking, but I would appreciate a couple of you who did better than me at statistics and modeling to throw your thoughts in here.

Thanks a ton,

Robert

rKeWl72

[SEO5Team]

Made a note to send you my next statistic related question directly , well done.

[RobertFisher]

SEO 5,

Thanks for chiming in. I am going to blame you for helping me with this, as I had almost forgotten posting it. After reading your response - and it is near impossible to do any tracking on that one source - you did make me realize what I was missing in the assumption.

Though I did not try to shake the cobwebs from my Latin, I think that the question is, can ceteris paribus be applied in this case? In other words - is it reasonable to assume that if you have 10 referrers and searchers from 9 act a given way when presented the same page/info, that searchers referred from the tenth will act that way as well. I think you can IFF (pulling out the old algebra terms on you there; IFF is for "if and only if" the referrers are from the same type of vertical - directories for example.

Let's assume the referrers are only directories like YP, Yelp, citysearch, etc. [I think when you use referrers that are of a different type, _s__ay pinterest, groupon, xyz marketing, other_s, the correlation is much less valid.]

Let's also assume one of five has no call data and where we can track the calls ( assume 4 referrers) for every 1.7 to 1.9 calls we see one contact form filled out.

Can we now construct an equation and draw a reasonable corollary: Where the referrer without the ability to track calls is Echo, and where calls (Z) is the missing data, then,

you can reasonably assume that the number of calls from Echo is between:

1.7(Echo contact forms) = Z and 1.9(Echo contact forms) = Z

So, I have to say to you - Good Answer and Thank You SEO 5!

Best to you and yours,

Robert

[SEO5Team]

Hi Robert,

Since noone responded , thought we'd chime in. Your assumption would make sense in this case based on all things being equal (Ceteris paribus) haven't used that since University:) . If source A and B are giving you accurate data when it comes to phone calls, we can assume that source C is also the same when it comes to the calls. The only accurate way to test this would be to initiate some sort of call tracking ( even if it is only for 30 days) on that referring source and confirm it for sure (then revert back to the usual #)

Hope this helps.