May 6, 2008
by Jeff James
Some experimentation with Google’s Conversion Optimizer has yielded positive results. 4 days into the experiment I’ve seen a significant increase in CTR and a 10-15% lower CPA. The conversion event is a lead and the conversion scenario across all campaign data is very much a homogeneous one. Many of the keywords are broad match geo targeted keywords (meaning keyword + location, the campaign is set to the entire United States) with essentially the same root and I’m using dynamic keyword insertion liberally in the title of the ads. Finally, the conversion event, if it happens will happen during the visitor session 9/10 times.
I had been using a tiered approach to CPA management with Omniture SearchCenter for the campaign. The rule formerly in use was roughly:
Assuming a CPA tolerance of $10 (for the sake of illustration)
- If CPA = $10, do nothing.
- If CPA {10 , 12} reduce max CPC by $.10
- If CPA {12 , 14} reduce max CPC by $.25
- If CPA {8, 10} increase max CPC by $.10
- If CPA {6, 8} increase max CPC by $.25
So in general, given that this evaluation is done at the keyword level the CPA should trend to the middle value over time and hover around it. It is naive in some respects because it has no contemplative ability beyond analyzing a ratio of 2 data points over the course of N days. It may not be making statistically significant decisions and the long tail of keywords which may only accrue 1 to 2 clicks per interval may never see any meaningful adjustments.
The Google Conversion Optimizer requires your campaign to have over 200 conversions over the past 30 days which in some cases is beyond the reach of an advertiser. For any high volume lead gen program this should be a visible option in the AdWords interface by now. Obviously if you don’t have conversion tracking enabled this isn’t an option.
As far as bidding parameters are concerned – the advertiser can specify CPA at the adgroup level. When setting the CPA values initially it will give you its prediction (CPA) and if you bid below it you’ll be warned that traffic may decline. I bid about 25% lower than their recommendation and I’ve seen no decline in traffic thus far.
As far as what the fundamental difference between rules based bid management and the conversion optimizer might be – well, it could be quite large. Beyond simply computing basic conversion math I suppose that the following statement by Google expresses adequately (how) excess value may accrue to the adopting advertiser:
“It uses the optimal CPC bid for her ad in each auction, thereby working to keep her average cost per conversion below her CPA bid.
Using real-time performance data allows the Conversion Optimizer to adjust Kim’s bids for better performance than she’d get with manual bidding alone. Also, because the Conversion Optimizer chooses a new CPC bid for each auction, Kim only spends money on the sites and search queries where her ads are likely to get conversions.”
Hopefully I’ll have continued success with the campaign optimizer. I suspect that within 24 to 36 months , the same technology will be available at a much lower conversion volume entry point depending on how predictive and sophisticated Google’s algorithms become as they continue analyzing trials using the current minimum level of conversion data. This is also a very elementary strategic consideration as the more Google can help automate advertiser success, the more they will be able to reduce revenue volatility and advertiser turnover. I wonder if they’ll ever release a landing page generator for advertisers…
Jeff
supporting data and FAQs – here
