A social comment article about Gmail advertising in the WSJ a few days ago was interesting. It blamed Google for irrelevant adverts. Now, I’d take this problem and turn it on its’ head. What can advertisers and agencies learn from the article?
Lily
The first item to attract derision was the confusion of “Lily” and “Lily Allen”. That’s growers of lilies and a British singer. Let’s look at the adverts. Please go to our page that shows you AdWords adverts in another country, and enter “lily” as a search, and select the US as the target location. You should see a page with a bunch of flowery adverts.
Hit the back button and amend the search to “Lily Allen”. Few adverts and all specific to “Lily Allen”. So why does Gmail show an advert for Lilies but keyword search doesn’t?
I believe that the answer is, as usual with Google AdWords, subtle, interesting and revealing.
Negative Keywords
A negative keyword stops your advert from being shown if that keyword is present in the search. For content matching, the presence of a negative keyword reduces the likelihood of being shown in proportion to the number of times the negative word is shown on the page.
Chances are that the Lily growers are using “lily” as the broad match keyword. Google’s AI has learned that “Lily” and “Lily Allen” are two very different conceptual clusters and no longer tries to make money by showing “lily” adverts on “Lily Allen” searches. But that lesson isn’t known to the contextual matching system. You need to add more negative keywords surrounding the keyword, in order to tailor the content matches to the most appropriate content.
Budget
The higher you set the budget, the more widely Google will spread your content matched adverts. With Keyword Search, you control how far Broad Match will stretch by using the MaxCPC (the bid for the keyword). If you set a very high bid, then Google will try to find any way to get a click on that keyword. As you reduce the bid, you get closer and closer to just the words in the keyword, or at least you do so until Google’s Ai has enough evidence of high click through rates on conceptually related searches.
However, Content Match behaves differently. If you keep the bid the same, but increase the budget, you get shown to a wider and wider audience until the relevance is down to a loose conceptual match on single words in the keywords you have. As you decrease the budget, the effective CPM increases, and the range of sites on which the advert is shown will decrease to more and more focused content.
Back to Lily
So, if a lily grower uses a single campaign for keyword search and content match, and sets the budget high to meet the needs of keyword search, the high budget can be taken as a signal to content match to extend the spread of advertising beyond the normal bounds for a keyword search.
In other words, separate the campaigns. Budget, costs, keywords, negative keywords and adverts should be different for content match than for keyword search. You need to have one campaign for each, and you need them tagged so you can do web server log file analysis looking for other important patterns in the data… There’ll be another blog about that, eventually.
But is it important?
Well, if you show an advert about lilies to someone reading about Lily Allen, you might make a sale. But chances are that you’ve just triggered someone into “Needs Awareness”, the earliest phase in the buying process and the one for which you have the lowest and longest likelihood of conversion to a buyer.
Why not focus your money, initially at least, on people who’ve said they want to buy lilies? When you are rolling in wealth from that, then you can start taking a speculative punt on other keywords. The improved CTR from better negative keywords will drop your advert costs and improve your ROI, and that’s ignoring the effect of losing very early phase buyers for whom you’ve paid, but for whom you stand a much lower chance of converting. Google’s advertising will seem more focused to users, and when the whole system offers higher quality adverts addressing user needs, you’ll see higher usage of adverts by users… So this has a short term and a long term consequence.
Bill Spencer wrote,
Good to see you’ve got your mind working, Jeremy.
Link | November 6th, 2006 at 4:48 pm
Jeremy Chatfield wrote,
Hi Bill - it’s stopping the thinking that’s hard… right now I’m speculating around the new Google Custom Search Engines, and the possibility of publishers abusing those, without advertisers having any control. This would need site exclusion for Google properties? I don’t think that Google allows that. Experiment in progress :)
Link | November 6th, 2006 at 5:29 pm
Richard Ball wrote,
Back to the original WSJ article, I wonder what the full content of the email referencing Lily Allen included. I find it hard to believe that Google’s contextual advertising system cherry-picked a single keyword and deemed that relevant. There must have been some other keywords that might have related to plants or flowers or gardening, somehow. As usual, these MSM articles lack the real details needed to understand what’s happening.
At the same time, Google needs to be more forthcoming and explain to advertisers how they match keywords to published pages. Just saying, “Oh, create a good search ad and make sure you have content ads turned on and our system will work for you” doesn’t cut it. Anyway, thanks for the insight. I’ll try not to incite.
Link | November 7th, 2006 at 9:49 pm
Jeremy Chatfield wrote,
Hi Rich@Apogee! Yeah, it’d be nice to know. However, I’m also using some other research, not yet blogged… It suggests that as the budget increases the content match engine spreads adverts wider, so that it considers single words in a keyword to be important. I suspect that a default combined campaign with a high budget and no additional negative keywords, that’s the cause. But this does depend on Content Match respecting Negative Keywords, and I’ve seen a few claims by Google staff that negative keywords don’t affect content match, and other Google staff claims that they do - very annoying, to not have a definitive, credible word. Leaves it up to experiment…
Link | November 7th, 2006 at 10:05 pm
James wrote,
I elect Jeremy to head up the UN Google Watchdog group!
I’m under the impression that the Google Custom Search Engines use AdSense and we therefore have control as we do with content match. No?
Link | November 13th, 2006 at 7:31 pm
Jeremy Chatfield wrote,
Hi James - we’re experimenting with CSE, but it looks like AdSense rather than Keyword Search. The experiment is a little tricky. You have to set up a CSE, tie it to an AdSense account, put up an advert with distinguishing tagged URLs, find and click it, then detect it in the web server log files; all without triggering any of Google’s Click Fraud detection systems. It’d be so much easier if they told us what it did…
Link | November 14th, 2006 at 11:54 pm
Merjis blog » Blog Archive » AdSense and Behavioural Targeting wrote,
[…] then I’d be happier about it. AdSense doesn’t really belong in Behavioural Targeting. It doesn’t find me 18-30 year olds with an interest in extreme sports. It could find me people interested in extreme sports, extreme political opinions and those involved in an argument, or wanting to find out about the Olympics. That’s not Behavioural. That’s weakly linguistically related. […]
Link | April 26th, 2007 at 8:40 am
Click Fraud, Google AdWords and gclid | Merjis Search Marketing Blog wrote,
[…] Assume that I see a conversion rate of 1%. Assume that I have an AvCPc of $0.10 on keyword. I’ll spend about $10.00 for a conversion on keyword search. Asume that the content network averages the same conversion rate (this is not usually true - the true conversion rate is much lower for reasons discussed in another article). I can then afford to spend $10 on 100 clicks - the $0.10 AvCPC that we saw for keyword search. Google picks up 50% of that. They make $5.00 and the AdSense partners share $5.00. If there was just one site in the list, then that’s one AdSense partner that receives $5.00. I can easily see that this is good value and I’ll invest more to get more placements with them. […]
Link | July 16th, 2007 at 8:24 am