Online Advertising Networks - MongoDB Applied Design Patterns - page 140

ECPM, CPM, CPC, CTR, ETC.

The world of online advertising is full of somewhat cryptic acronyms. Most of the de-
cisions made by the ad network in this chapter will be based on the eCPM, or effective cost
per mille. This is a synthetic measure meant to allow comparison between CPM (cost per
mille) ads, which are priced based on the number of impressions, and CPC (cost per click)
ads, which are priced per click.

The eCPM of a CPM ad is just the CPM. Calculating the eCPM of a CPC ad is fairly
straightforward, based on the CTR (click-through rate), which is defined as the number of
clicks per ad impression. The formula for eCPM for a CPC ad then is:

eCPM = CPC × CTR × 1000

In this chapter, we'll assume that the eCPM is already known for each ad, though you'll
obviously need to calculate it in a real ad network.

Operation: Choose an Ad to Serve

The query we'll use to choose which ad to serve selects a compatible ad and sorts by the ad-
vertiser's ecpm bid in order to maximize the ad network's profits:

from from itertools itertools import import groupby

from from random random import import choice

def def choose_ad ( site_id , zone_id ):

site = db . ad . zone . find_one ({
'site_id' : site_id , 'zone_id' : zone_id })
iif site iis None : return

return None
iif len ( site [ 'ads' ]) == 0 : return return None

ecpm_groups = groupby ( site [ 'ads' ], key = lambda lambda ad : ad [ 'ecpm' ])

ecpm , ad_group = ecpm_groups . next ()
return return choice ( list ( ad_group ))

First, we find a compatible site and zone for the ad request.

Ne xt, we group the ads based on their eCPM. This step requires that the ads already
be sorted by descending eCPM.

Fi nally, we randomly select an ad from the most expensive ad group.