Google - Doing Less Evil

Published on January 28th, 2008 by Jeremy Chatfield

I frequently hold forth here that Google abuses the real source of its’ wealth - advertisers. The “Do No Evil” mantra is applied to users, but Google does horrible things to advertisers, remorselessly. After years of criticism from Richard Ball and me on the AdWords Help Forum, and joined more recently by other voices, Google is now advocating a different approach to the content network, one of Google’s main evils.

To indicate just how radical a change of approach this is, I’ll cite a few examples. Some of these represent areas where the new advice is, I hope, in advance of product changes.

• Google Accreditation (GAP/GAI) training material used to emphasise that the same AdGroup should be used for keyword search and content match.
• GAP exam questions sometimes assumed that keyword search and content match would be enabled for an AdGroup - no questions involved separating the two, or the reasons for doing so. When I next take the GAP exam, I’ll keep my eyes open for changed emphasis in the questions.
• Google’s Starter Edition enabled content match by default - so the least able and newest advertisers were automatically exposed to the greatest risk. I have not used Starter Edition or set up a new campaign in Starter Edition for ages - I’ll have to do so and see if this evil has been corrected.
• When creating a new campaign, Keyword Search and Content Match are enabled by default. Clearly this should be changed.
• Large accounts have access to Google Maximizers (or “Strategic Account Optimisers” or “Account Strategists”). As recently as a few months ago, these guys were stll advocating creating Campaigns that mixed Keyword Search and Content Network, treating impressions and clicks as being of similar value to those from keyword search.

Note that unlike text adverts, which require “Laser Straight Alignment” (thanks for that description, Firefly in the AdWords Help Forum), the content of the web page that Google selects for AdSense need not contain precise matches. That is, for keyword search, you’d need “rubber duck collector” as a keyword, the advert needs “rubber duck collector” and the landing page needs “rubber duck collector”. Google will gleefully (for the right conditions) match sites with just the word “rubber”, or “duck” (and possibly chicken, turkey, goose, etc) or “collector” (and probably “collection”, “collect”, etc) - this enormously expands the range of sites available, but weakens the already weak intent of this kind of user to purchase.

The new advice still misses out a couple of other key reasons to separate content match, and the ways in which advertisers can optimise the Content Network - ways in which they can optimise for their benefit and not that of Google.

• Bidding behaves differently in the Content Network from the effect in keyword search
• Budgeting affects something different in the Content Network

When Worlds Collide

When Google smashed together Keyword Search and the Content Network, they wanted to provide a simple interface, and they clearly regarded the two networks as complementary. So Google also merged how the bidding and spending controls were handled - but these controls have radically different uses in the two networks.

AFAICS, in Keyword Search, the bid affects the position, and the frequency of appearance. The higher the position, the more likely that third party networks will carry the advert. This positional change can result in a large difference in impression volume for a keyword, with relatively small changes in position. We’ve measured nearly ten times as many impressions in position 1, compared with position 5, for some keywords; for other keywords, especially low volume searches, the variation may be much lower, but often may provide a 2:1 impression volume change, as well as the increased CTR that you get from top billing. I can’t see this effect so markedly in the Content Network *above position 4*. I do expect this change in behaviour at position 4, because most publishers seem to use a single AdSense block, resulting in a maximum of four positions being shown - so impressions drop markedly below position four on the content network.

Bidding in the Content Network appears to directly affect only the position. If your position is low, lower than position 1 but at or above position 4, then increasing the bid appears to do two things - it increases the position and *reduces* impressions. Whereas if you are below position 4, increasing the bid to appear at position 4 or above will radically increase the impressions.

Why do you see a reduction in impressions in the higher positions? I believe it is because Google calculates how many clicks would satisfy the budget setting, and a higher bid will mean that fewer clicks are needed to exhaust the budget. The way that Google can control this is to either throttle the frequency of the appearance, or restrict the sites on which adverts are shown - either of these implies fewer impressions. AFAICS, when you make bid changes, Google throttles the advert - keeping the range of sites the same.

Budgets appear to be different in effect between Content Match and Keyword Search. In Keyword Search, Budgets control “throttling”. That is, if you have a budget of $100, but Google estimates that you’d get full advertising for $1,000, then your adverts will show only 10% of the time. Whether this is the first 10% of the day, or a 1 in 10 appearance through the day, is affected by the “accelerated spending” option in the campaign settings.

However, on the Content Network, it appears that the Budget affects the breadth of sites to which you’ll be exposed. That is, Google uses the Budget as an indication of desperation for clicks. If you set a Content Network budget of $10, then Google will only show the advert to sites where it is pretty sure it can make the $10 - a very close content relationship with your keywords. If you set the budget to $1,000 (for the same Campaign), it appears that you’ll be shown on sites with looser and looser relationships to the keywords you are using, until Google exhausts the most likely inventory.

An economic/business analysis of the system shows that this makes sense. If Google is trying to optimise revenues from advertisers, then those with low budgets are best satisfied by showing a low volume of adverts on sites that are more likely to click. However, higher budgets allow Google to address less relevant inventory and possibly extract revenue from sites that would otherwise only show public interest adverts. Consequently, increasing the budget will result in a lower CTR on the content network.

If you look at that business analysis, you’ll see that the budget is handled with respect to the advantage to Google, not the advertiser. That is, if Google has an excess of publishers with rubbish content, then advertisers with high budgets will get shown on the rubbish sites *so that Google can make more money*. As has been extensively documented elsewhere, and now admitted by Google, you need to control keywords for Content Match in a different way than that for Keyword Search, and you need to consider bids and budgets in a different way.

Until Google changes practices (disabling Content Network in the Starter Edition, turning off content network in keyword search oriented campaigns by default, etc) this advice will have little impact on Google revenues. Those who know what they are doing have tended to disable the Content Network as a matter of standard practice. Those who don’t - well, I suspect a goodly part of Google’s revenue growth can be attributed to ignorant advertisers having their adverts shown on any willing publishers’ sites.

If I were a Wall Street analyst, I’d be wondering whether Google’s slight reduction in Evil, may also signal a slowdown in revenue growth. IMO, the risk for Google is that small advertisers appear to be getting more wary of the giant search company. I’m encountering resistance to Google AdWords based on SME’s passing around their experiences - which, if they have gone down the DIY route, is usually pretty dismal, and especially so when they’ve seen 80% or more of their budgets spent on Content Network clicks. The question in my mind is whether Google’s revenues can continue to grow if they don’t become less evil to advertisers. While Google continues to enjoy a near monopoly of user search, revenues should continue to be pretty good. But if Yahoo! finally gets some kind of plausible act together, or MSN produces a search engine that works reasonably, advertiser defection is likely to be rapid - Google’s real relationship with advertisers hinges on its ability to bring in volumes of users. If that fails, then the revenue stream comes under threat. This change of policy may reflect new understanding inside Google that competition really could damage them, unless they do something visible to help with the perception of click fraud that continues to bubble away.

The real question for most small advertisers is whether there really is any point to Content Network advertising. Given that users are in a different frame of mind, the CTR will be lower. Usually the conversion rate is also much lower. Because it appears that the Campaign budget is used to control exposure to relevant sites, then many campaigns will be needed - to make sure that different AdGroups get the right targets. The costs of managing the multiplicity of campaigns and the lower volumes of buyers from these routes can make the management costs for the activity far too high to justify significant targeting effort.

Recommendations

These really split into two areas - those for advertisers and those for Google. And, to some extent, to me…

Small advertisers should disable the Content Network, if they haven’t already done so. Chances are that this will reduce spending but have no adverse effect. A small fraction of advertisers, mostly specialists in handling this type of advert, would see a negative consequence - but it is easy to re-enable the Content Network should you notice any damage to sales.

Google needs to follow through the changing advice by improving what they actually do, rather than just changing the advice. Changing the advice is a good start, though.

I need to dig a little more into whether the GAP exam has been revised to test this new set of recommendations, and to see if the AdWords Starter Edition has been revised to reflect the new Best Practice recommendations from Google.

Marks to Google for beginning to provide some sensible advice about the Content Network, instead of the misleading fog they previously offered. I see some reason to be more optimistic about Google’s relationship with advertisers.

Updates

2008-01-31 I’m always worried when I see Google playing nice to advertisers, when I don’t understand why. Google has been eating into Yahoo! and MSN market share for search - so expect keyword search broad match to dilute relevance in favour of income. But read this live blog of the analysts report for Yahoo’s fourth quarter 2007. Looks like Yahoo is growing in one area - display ads. Competing with, and beating, Google. Whew. I feel better. Looks like Google is feeling the heat from a competitor and is responding. The article has not been revised to reflect this.

2008-02-02 Inserted new paragraph after the examples of Google’s previous Content Network advice. This expands on the difference between keyword search and content match - it is closer to extended broad match at its most aggressive.