Whether written out in computer code or just rattling around inside people's heads, many PPC account managers like to use rules to help guide their bidding decisions. Acquisio's software, for example, lets account managers set rules like: "If actual CPA is less than one-half the target CPA, then raise bid 40%" or, "If actual CPA is more than 2 times the target CPA, then decrease bid 30%". But more often than you might think, you need to put the brakes on these rules, or they might break you. Trada.com lets its clients specify a target CPA that they want their ads to hit, for example $15 per conversion. And they can also specify a 'Keyword Spend Limit', like $30. That is, the advertiser is trying to pay $15 per conversion, but if the keyword spends $30 without a conversion, its ad will be suspended indefinitely. This might sound reasonable (who wants to let an ad run unfettered if it's not generating any conversions?), but it virtually guarantees trouble. Let me explain how. Let's say that an ad has a 5% conversion rate. If the ad's target CPA is $15, then you should be willing to pay up to $0.75 per click (target CPA x CR) for traffic. To aim for more than that brings in conversions at too high of a cost, on average. To aim for less gives up conversions at a price you are willing to pay. (We are considering each ad in isolation, with no budget limit.) So, we should bid whatever amount is necessary (for the sake of argument, say it's $0.90 per click) to make our CPC be $0.75. The 5% conversion rate means that the ad averages 1 conversion every 20 clicks. But, of course, sometimes it might take 25 clicks to get a conversion and some other times only 15 clicks. If you think about it, it should be obvious that 5% of the time the ad will get a conversion on the very next click after the previous conversion. So, 5% of the time that you check the ad's performance you will see that it got a conversion from only 1 click, for a CPA of $0.75. Woohoo! What a great ad! Of the 95% of the time the ad does not convert on the first click, 5% of the second clicks will convert. So, some of the ads will have a $1.50 CPA. Still a fantastic performance for an ad with a target CPA of $15, right? If you manage 100 ads, each of which has a 5% conversion rate, you are therefore virtually guaranteed to find some with $0.75 CPAs, some with $1.50 CPAs, some with $2.25 CPAs and, yes, some with $30, $40 or $50 CPAs even though all of the ads have identical CRs and all of the ads are at their optimal bid. Jerking bids around without accounting for this fact is likely damaging your overall PPC performance. How often will an ad get M clicks before a conversion occurs? Well, the fraction of keywords that will make it beyond M number of clicks without a conversion is simply: If 100 ads each have a 5% conversion rate, then about 36 of them will make it beyond 20 clicks without a conversion. Of the 36 ads that make it beyond 20 clicks without a conversion, 36% of them on average will make it 20 more clicks without a conversion. So, about 13% of the original 100 ads will see a CPA that is at least 2X the target CPA (and about 5% will make it to 3X the target CPA) based purely on luck. These 13% of cases are not 'bad ads' - remember, all of the ads in this example have the same CR and are at their optimal bid - but if you write a rule to shut off any ad that racks up 2X its target CPA (like Trada's system does), you will wind up shutting these ads off anyway, when they are just as likely to convert on the next click as any other ad. If you cut the bid some large amount (like Acquisio's system allows), you will still be punishing perfectly good ads, hurting yourself in the process. Similarly, raising the bids of ads with low CPAs will also cause problems. For 5% of the ads you will see a CPA of $0.75 even though each ad's chance of getting a conversion is still only 5% per click. In fact, on average, 40% of the ads with 5% CRs will convert with a CPA less than 1/2 of the target. If your CR is not 5%, the specific numbers above change, but the principle is the same: It's not enough to see that an ad got a CPA of "OMG $1.50!" or "what-the-heck $40!" If the bid is at the optimal level your best response, even to dramatic swings in observed CPA, is simply to do nothing.