I’m pretty good at math.
I can’t do calculus or anything really complicated, but I consider myself to be one of those people that are “good with numbers”. At the very least I know that $163.27 divided by 1 should be $163.27.
So I was understandably confused when Google reported that it was actually only $61.53….
Here’s how this came about.
I set up a new account for a client and almost immediately we added call tracking (the best call tracking – from the nice people at Call Tracking Metrics). We duly configured our goal in Google Analytics and imported this back into Google AdWords as a conversion.
Over the Christmas holidays things have been slow, but there has been some traffic and there was a phone call. Sure enough, this reported into Analytics and a couple of days later back into AdWords. In a way, I’m pleased that things were a little slow, because had we been busier I might not have noticed this funny little reporting quirk.
There are two interesting things that appear to be happening in the report Google sends. Firstly, the cost per conversion doesn’t include any cost incurred before conversions were imported. That’s fine – in fact it would be essentially wrong if it did.
But the other thing is a little misleading at best, and slightly more than deceptive at worst…. when we launched the campaign we applied a $100 coupon to the account. Google does not compute the value of this coupon in the calculation of the cost per conversion. So although the ad spend was $163.27 – the system subtracted the $100 coupon value and calculated the cost per conversion only in terms of the clients “real” spend.
There is a logic to this. Twisted, but there. Let’s suppose that the conversion was a much simpler conversion than a phone call, or a sale – a measure of time on site, for example – and let’s assume that approximately 25% of visitors would convert. In our case from the 36 visitors, we would have seen 9 conversions. The “true” cost of these conversion would have been $163.27 (at least, if you wanted to replicate this performance, this is what it would cost you) – so a cost per conversion of $18.11… but if we only count the cost without including the coupon then this figure would drop to $7.03.
Optimisation decisions probably shouldn’t be taken after so few visits, but, if they were, I am sure that the decisions would be different if the true (replication) cost were considered.
Am I being paranoid, or is this designed to mislead?