Adwords ‘Enhanced CPC’ Split Test Results

13 Comments Written on March 4th, 2011 by
Categories: Google Adwords

After reading Mark Ballard’s post on the Rimm Kaufman blog about the possibilities with Google’s new “Enhanced CPC” feature I was intrigued.  Google claims that Enhanced CPC can be a nice bridge between manual bidding and full-blown Conversion Optimizer (a campaign setting that allows Google to show your ads when and where Google thinks you’re most likely to get a conversion).

Enhanced CPC doesn’t take into account the CPA or cost-per-conversion you’re looking to achieve like Conversion Optimizer does, rather it  automatically adjusts your Max CPC bid, auction-by-auction (+/- 30% of your Max CPC bid), depending on whether Google’s algorithm for Enhanced CPC thinks you’re most likely to get a conversion.

Free Bid Management?

At the end of the day, Enhanced CPC is supposed to lower your max bid when you’re least likely to get a conversion and jack it up when you’re most likely to convert, lowering your overall costs.  If it turns out that this system works, it makes automatic bid management with manual over rides for zero fees a reality.

To determine the likelihood that a particular auction will convert for you, Google looks at a number of factors:

These include the individual words that make up the user’s query; the user’s geographic location, web browser, and operating system; and even time of day when the user is performing a search.

As Mark rightly pointed out in his blog post, advertisers can’t typically tinker with the settings for a number of these factors like OS and web browser, though they do have control over time-of-day ad scheduling and geo-targeting.

Setting Up the Split Test & Results

I decided to run my own test with Enhanced CPC.   In order for the test to be largely accurate, I had to ensure the campaign I was testing it on had no changes in the last 30 days or so to get an accurate baseline on all of the campaign metrics and conversion rates.  Additionally, I needed to ensure I didn’t change anything but the bidding method (changing from manual bidding to Enhanced CPC) to eliminate other factors accounting for changes in campaign performance or metrics.

That said, here are the results after 6000 clicks on a year-old Search only campaign (no Search Partners) taken over several consecutive weekdays that was untouched in the preceding 30 days:

Avg CPC- Start: .18c      End: .19c

Avg Pos.- Start: 1.7        End : 1.8

Impression Share- Start: 62.84%    End: 62.66%

Cost per conversion- Start: $17.60    End: $27.41

Conversion Rate- Start: 1.01%    End: 0.69%

Conversion Counts (1 per click): Start: 43  End: 31

The test period using Enhanced CPC was measured against the same days of the week and number of days as the previous ‘control’ week (aka the “Start”).

In a Word: “Meh”

End result?  Costs and went up slightly, conversions tanked, ad position dropped and impressions didn’t budge.

The big metric to measure of course is ROI, and in this case ROI dropped like a rock.

The odd thing however is that given the above results, you’d expect to see the Avg CPC  drop overall, which didn’t happen.

For me the net-net is that it didn’t work as advertised.  I thought perhaps it needed more time to ‘calibrate’ for maximum efficiency, but according to a number of PPC experts I spoke with it doesn’t really require the same lead time as Conversion Optimizer does to optimize itself based on campaign data.

What do you think?  Just another “bell and whistle” from Google to make Adwords appear easier for the masses?   Another attempt by Google to compete with 3rd party bid management providers?  Leave a comment below!

>> Subscribe to our blog posts via email to get more great posts like this one!

13 comments “Adwords ‘Enhanced CPC’ Split Test Results”

Hey Geordie,

Actual experiments/tests like this are extremely valuable.  It sure seems like another tool that google is trying to launch to compete with 3rd party bid management solutions.  You hit the nail on the head with your statement, “The big metric to measure of course is ROI, and in this case ROI dropped like a rock.”  ROI is the name of the game 🙂

how is this called split test? I thought a split test is to send half of the traffic to test and other half to the control groups It appears that you carried out a time series test.

Google’s root objective is to make more money. Don’t ever forget that. Everything they do is to help the company be more profitable because they are publicly traded and have a responsibility to shareholders. While they may market this feature as an altruistic tool to help you get more conversions, really it allows them to charge more for clicks that have a higher chance of converting. And who determines which clicks have a higher chance of converting? Google does. While Google may refine the feature to be helpful (you should always be testing) you need to remember that Google may not be evil, but they aren’t your friend either.

@bulent: Adwords doesn’t allow you to simultaneously run traffic to an A/B of with or without Enhanced CPC, so point in time testing is all you can do.  That said it’s still valid.

Great research indeed. We (at WordWatch Inc.) spent a lot of time as well, to increase conversions and decrease CPA, without any guidelines from our customers. But we are back at asking for a max CPA and doing a much better job since.

Thanks for sharing this, I worry the same about automatic bidding. I think they’ll bid what is best for Google and not what is best for my pocketbook.

The only way i know to split traffic is to use the Experiment tool in Campaign Settings.
Read more:

Thanks for this, really great to see someone testing the enhanced CPC feature and sharing the results.  I have just started a small test myself. It has not been running for long- spend has increased, but conversions seem to have increased too. We shall see.

One thing you can be sure of is that Google is going to maximize your daily budget, so make sure this is set at a level you will be happy with! Obvious, but many advertisers fear ‘losing out’ on auctions, so set their budgets much higher than they would really be happy with.

I ran a 1 week test and although many other factors could impact my ROI and the last few weeks have been kind of wierd, I still concluded that my conversions bottomed out completely as a result of using “enhanced”. Maybe it will improve in time but I lost some serious cash during my 1 week experiment with “enhanced”. I’ll try again in a few months I suppose..

Does politics (ie Obama’s speech) affect anyone else’s conversions, impressions, or clicks out there?? Thanks..

just trying hands at it – been 1 wk into – lets see.. I think u need at least a fortnight to decide, but true haven’t seen any improvements yet.. also read another post which claimed to have shown some improvements. maybe it is case dependent. also i think a strong conversion history is important for an algorithm like this to do well.

I have been testing the changes between CPC and eCPC with a similar sample size of data and I got similar results.

I was surprised by just how drastic eCPC bidding underperformed CPC bidding. In my case ROI was 35% higher with CPC vs eCPC. I noticed that google does not record the change of bid type from CPC to eCPC in change history so it is harder to know when exactly you made changes and therefore harder to track the ROI impact.

I don’t want to get to conspiratorial sounding here so I will look into using Adwords Campaign Experiments to measure the difference in ROI, but at the end of the day that tool still relies on Google to be honest about the results.

I have to say, Enhanced CPC Bidding – Google Adwords ~ PPC Blog is a really good blog site. I might like to show you my hi and thanks. Thank you!!!

Hi Geordie,

Thanks for sharing this. You might know this by now, but just wanted to point out that ECPC actually does function as a split test, you can ask for a report on the results from your Google rep. That will give you the most reliable data.

IMHO, you’d usually need more than just a couple dozen conversions/mo before any meaningful testing and positive results can be accomplished – otherwise there’s simply not enough data for the algo to make smart bidding decisions. For the most part I’ve found ECPC to be a beneficial complement, but I had a couple of bad experiences testing it with too shallow conversion volume.

Leave a Reply