Budgeting Algorithm Problems for Panama - Updated

I’ve just been playing with Yahoo’s latest Search Marketing Platform, Panama (officially launched to the public today), and to my dismay, by the time I got to second screen, I managed to find a (major) bug in their budgeting system (hint: right click view image below and then zoom to see the screenshot properly).

There’s obviously a major mathetical error in their calculations & code, because at a certain point in the estimation ranges, the calculation breaks (and basically points to an infinity function that is simply not there). There is a 6-point 0 data range on either side of the break point. This is over and above the volume/clicks/cost ratios being totally out of whacked. I applied a logarithmic budgeting curve in order to save time whilst doing the projections - the numbers are pretty inconsistent. I’ve documented the price breaks in this spreadsheet as well.

Update 14 December: I’ve had a lot of enquiries today to clarify the problems presented in this post - so here they are. It would be handy to have the spreadsheet open.

Basically, the analysis was performed in order to get an idea of the slope at which Yahoo’s bid prices would increase to the maximum CPC. In plotting this slope, I discovered that the increase in CPC’s was linear to a point, and then became logarithmic. Also, in the beginning, at low recommended CPC based on low daily budgets, the amount of impressions actually drops until your CPC moves past the 50c mark - very weird.

The more concerning aspect around this opaque budgeting algorithm is the point at which the system recommends a 0 value and also eventually at the midpoint of the twelve “0″ values, there is an infinity error, which basically means that the maths behind the algorithm receives a number that it can’t handle (like divide by 0). Also, once your campaign budget passes through the “0″ values and error, the number mysteriously jumps about 3 fold.

Without looking at the calculations, I can’t say what the problem is, however this will seriously damage any credibility from advertisers as it appears that there are massive increases in cost toward the end of the spectrum, after a break in the algorithm. It seems like Yahoo needs to hire some of Google’s Ph.D’s in math or get better Beta testers, like yours truly :-).

If you enjoyed this post, make sure you subscribe to my RSS feed!