9 Things I Dislike about The Google Adwords Automatic Matching Idea

Pay per click advertising works best when you maintain control over your match types, negative keywords, ad rotation, and landing pages. The minute you relinquish control to broad match and other “lazy” modes of traffic, the minute your cost-per-lead/sale can go through the roof. This got worse in the Summer of 2005 when Google started making broad match into “expanded broad match” – greatly increasing the situation an ad will display by loosening restrictions — and now it sounds like things are going to go broader still. I’ve read that on May 20th Google is expanding the beta (or possibly launching) the Automatic Match system (nicknamed “Screwgle” by Paul Boutin at Vallywag) and it sounds like broad match gone amuck meant to improve Google’s profitability.

From Google:

“Automatic matching shows your ads on relevant search queries not already captured by your keywords. It works by analyzing the content of the landing pages, ads, and keywords in your ad group. It then shows your ads on search queries relevant to this information. The system will continually monitor your performance on these queries and adjust its matches accordingly. Automatic matching aims to show your ads only on queries that yield a high clickthrough rate (CTR) and a cost-per-click (CPC) comparable to or lower than your ad group’s current average CPC. This way, your ads receive additional targeted traffic at a similar cost to your current traffic. Automatic matching won’t allow your spend to exceed your budget, and it also won’t affect the traffic you’re currently receiving. In addition, automatic matching will have no impact if your campaigns already capture the majority of relevant traffic.”

I’ve not used this system ( I wasn’t in the beta ) but even just reading this email I can already list at least 9 things I dislike about it.

1. It does the equivalent of keyword research on the fly, trying out keywords to see if “they stick” based on Google’s data. This removes the human from keyword research. My experience is that 80% of keyword research is ELIMINATION of poor quality choices before you spend money on using them in your campaign. In other words, there is a lot of intuition applied based on advertiser domain knowledge and automatic match eliminates that phase. While some may think this approach is a good way to learn keywords to use, I think there are better ways, especially when they involve harvesting keyword data and then doing some smart detective work on the results.

2. The almost impossible task of creating negative keywords that will negate the constant algorithmic display of Google-chosen keywords. Think “automated embedded match without the associated automation on the negative embedded match side…. How often will you check it? How often will amateur or new advertisers? I’m guessing never. So you will very likely end up advertising for phrases where you do not offer the product being searched for – a classic waste scenario.

3. The enabling of the system by default. Many advertisers won’t know what hit them. While Google has assured everyone that this is an optional system, it would be likely that it will be turned on.

4. It further dismantles the smart idea of keyword-specific URLs and landing page parameters, not to mention dynamic keyword insertion. Caveat – I have no idea how Google will handle existing exact match and phrase match keywords – I’m assuming it will use the old rule of most-restriction-first-shown.

5. It spends your budget. All of it, on your behalf. With my clients, I prefer to ease “up” to their daily budget. That means starting with precision matches and loosening them gradually to find a sweet spot between relevance and volume. This way you don’t waste money or possibly reduce quality score while you tweak your campaigns or testing (by throttling our ads.)

6. It encourages lazy adwords management. When your ad campaigns are running, having them unattended can create an unhealthy black box between you and your pay per click. The variety of auctions, bid types and match types already has advertisers perplexed – and adding another dimension will make it worse. I really worry about “rush” and “fire and forget” Adwords solutions. They are insanely wasteful. This is why I hate turnkey PPC setup programs.

7. The only way you can see what queries are generating ads is to turn on a Search Query Performance report and watch for irrelevant queries – then create negative keywords for them as they come up, whack-a-mole-style. A never ending cycle that few advertisers will remember to do.

8. Advertisers will need to look at the special Automatic Keyword Performance data in reports to see how well these automatic matches are working and add special tags to your destination URLs for logfile analysis. This means yet another tag on the URLs.

9. Google does a pretty bad job of limiting broad match. With this already causing grief, are we really going to make things better by loosening it further? Yes, I know – some of you will say that broad match has its place in pay per click marketing. As Brent Hodgson expertly points out, perhaps that it’s a way to cast a net and capture heretofore unknown keyphrases. If you have endless funds, yes. But it’s a bit like testing newly designed bulletproof vests on real humans – it’s nice when you get it right – but when you don’t, it’s bloody expensive. I think other keyword development tools + log analysis + intuition + internal search logs make a hell of a lot more sense.

The right way to do Adwords

Take the time to grow a keyphrase list, divisions into thematic adgroups, and built up a collection of negative campaign and adgroup keyphrases that use phrase and exact match. Easing into your budget using split testing of ads and optimization of landing pages, you can eliminate broad match (conceding: except on your experimental campaigns.) This process alone can improve the conversion performance of your campaigns by a factor of three or four.

I should underscore that I’ve not used the system, but I’m very familiar with adwords and extended broad match issues. Is there a chance that the algorithmic remnant-sale system will work? I guess, but only if your Adwords campaigns were AWFUL to begin with and you’re willing to throw money to the wind and see where it lands.

Update:  Some real world data about this feature has been posted by Jim Gilbert and Mike Churchill