Expanded Broad Match Corrupted Around Aug 20 — Stop Using It!
by Jim Gilbert
Summary of the Issue:
Around August 20, 2007 many clients’ Google AdWords accounts saw their cost/conversion skyrocket — in one particular case we saw it was over 100% increase. It was caused by a corrupted “Expanded Broad Match” algorithm.
What Caused it?
“Expanded broad match” was expanded by Google way too much. AdWords began showing ads for “expanded broad match” terms that were just not relevant to the broad keyword generating them.
What to do:
1) Stop using “Broad Match”! If you do use it Google kicks in “Expanded Broad Match” and you can not opt out of the expanded part.
2) Insist to your Google contacts that you be allowed to opt out of expanded broad match!
(We have tried for a couple years, but have made no progress.)
UPDATE: 10/16 — see 10/16 note below
The Rant, Some Interesting Thoughts and the Detail — if you like the fun stuff
The Real Title of this Post should be “Expanded Broad Match - Google’s EPS (Earnings Per Share) Equalizer”
Rarely will you find me crawling all over Google’s AdWords PPC offering. Compared to the alternatives (competition in other words), they have done a technically good job, understand usability, continue to grow their footprint and are somewhat responsive to user (and agency) needs.
However, “crawling” is now in order. Since Google announced their “expanded broad match” it has been a sore spot with all those who understand how it works and what it does. For over 2 years we have begged Google to allow AdWords clients to “opt out” of the “expanded broad match” — broad match as it was originally YES, expanded broad match as it is today NO, NO, NO!
Around mid to late August and into September several very valuable AdWords clients cost/conversion numbers went VERY FAR SOUTH (like in bad & the ugly) — in one case over doubling! Careful investigation revealed that the increase in conversion costs were directly related to a VERY FEW broad match terms.
Even further investigation identified “some” of the new “expanded broad match” terms that had kicked in and destroying the conversion costs. Expanded terms that DO NOT RELATE to the broad term in anything resembling an acceptable manner.
Demands to the usual Google contacts requesting an option to “opt out” to “expanded broad match” were (again) unsuccessful. We provided the documentation and our reps agreed that these documented instances were “kind of far out there”. They even offered to help find more negative keywords to prevent it from happening. In one case we let them try to add effective keywords is and existing list of over 1,200 ones — trust me there was little they could do. See, there is no tool for identifying what Google expands broad keywords to. There used to be, but it conveniently disappeared.
We have great Google Reps, but they can only consult with their superiors on making these types of changes — the reps do not have the power to make something like this happen. Nor do they have “thousands of users” complaining — since their are probably not thousands of users who understand or can detect what is really happening.
Why does Google refuse to allow opt out for “expanded broad match”? The original explanation I received for implementation of “expanded broad match” was to enhance the AdWords user’s experience and provide them a better variety of related ads. Started out innocent enough, but as Google went public and had to answer to the ridiculous quarterly financial demands put on public companies by “Wallstreet” they probably realized that killing “expanded broad match” would have a severely negative impact on AdWords revenue. Furthermore, with continued “Wallstreet” pressure following Google’s first time “missed quarter” Google probably sees this:
My Opinion Only and I will remove it if Google will allow us to opt out of “expanded broad match”.
Not only can they NOT AFFORD to allow opting out of “expanded broad match” — by just loosening the knob they can instantly and dramatically pump revenue up in seconds.
Now it’s time to monitor the “expanded broad match knob”. We have put things in place to detect this “knob turning”. Like last time, wouldn’t it look funny if the knob was loosened only during the mid to late quarter periods?
This is only the beginning — those smart enough to monitor and detect wasted click spend due to unsupportable expansion of terms will at some point do more than just beg or throw fits.
P.S. Relax… I am not a cynic or pessimist — The facts of the analysis support these conclusions (except maybe my final opinions).
UPDATE: 10/16:
Appears the Expanded Broad Match Knob was screwed back in in a good bit. I screamed and hollered and Google backed off — doubtful. What is more likely is that they had pumped revenue all they could for the quarter ending Sep (3Q) and backed off during early 4Q (and Hoping they don’t have to do that ever again?).
October 11th, 2007 at 2:43 pm
[...] Originally Posted by laurawp do you mean select the best match type for the keyword (depending on it’s “genericness”) or would it be better to go with broad match but ensure I’ve got a really thorough negative keywords list? You could use 1 keyword on broad match and 1999 negatives in an adgroup BUT you might still be getting traffic you never even considered due to google’s slack expanded broadmatch algo. personally I think the expanded broad match algo is bordering on click fraud .. You put the keyword in your account in good faith , you and anyone else you care to ask would reasonably think that your keyword is going to be triggered when it or a highly relevant search term is typed in.. Yet in reality your advert is triggered on terms that vary on a scale of relativity from exactly that term to only the most tenuous link.. vibrators to ‘vibrating tooth brush’ .. true .. yes .. they both vibrate but toothbrush is totally irrelevant, may as well be a flight keyword and it shows you paddle steamer ads..both are modes of transport after all.. from vague we go to over specific and irrelevant .. you may be broad matching on ‘printer cartridge’ yet expanded broadmatch may show it for the term ‘12A1970′ which is great BUT only if you sell Lexmark 12A1970 printer cartridges… this is where negatives come in.. you may have to add even model numbers brands you may have negatived already.. imagine the list of negatives there for a product like cartridges. or irrelevant model numbers.. also watch out for expansion ie.. you bid on “Samsung ML-1430″ and it also shows for Samung ML-1450 .. which you may not even sell, or have in stock etc.. the list goes on..expanded match is just plain baaaad Do a search on expanded broad match on google and you’ll find posts like Expanded Broad Match Corrupted Around Aug 20 — Stop Using It! · SEMClubHouse you should be able to choose between broad match and expanded broad match so you can opt out of it like you can for the content or search network, their disclaimer is facile and simply glosses over a potential spending black hole. Just using broad match is pretty lazy, not to mention potentially dangerous from a ROI point of view, you should be doing highly targeted adgroups, adcopy, using phrase and exact matching and negatives.. and where broad match works.. then yup use that too but it should be the last resort not standard issue. __________________ Shane Dreamweaver Templates [...]
October 12th, 2007 at 1:12 pm
We’ve been dealing w/ the exact same issue, started about late Aug and really peaked into the “sphere of awful” about 10 days ago. Cost:conversion was through the roof. We notched back the main offenders using phrase and exact, left a few proven phrases in broad, and are watching closely.
This has such a dramatic effect on ROI it’s just silly for Google to do. Need to throttle that down. Then again why not keep it aggressively broad if you’re Google? The big fish with huge spends probably don’t even notice or care, while the smaller players who are careful about conversion costs get hammered.
October 12th, 2007 at 10:35 pm
Nice article. The more pressure that can be put on them about this, the better. Expanded broad match is a crock. Previous to implementing an extensive and ever expanding negative keyword strategy, maybe 10% of the terms in the search query report for broad match were relevant. Another 50% were totally irrelevant horrible matches (everything from “used jeans” to “wholesale automatic weapons”) and the rest, well, let’s say they ranged from cartoon sex clips to things that get people put into sex offender databases. Let me tell you, it’s great to learn a client’s brand was associated with that.
October 15th, 2007 at 12:18 pm
Wow - Thanks. I have been fighting that for over 6 weeks. At least now, I know where to look to make the correction.
October 15th, 2007 at 5:05 pm
Yeah we picked this up too;
http://forums.digital-m.co.za/showthread.php?t=2278
But we think we have evidence that “broad match” was expanded to include a user’s search history, and not just semantic variations of a search phrase. It makes sense to try this if you think about it - at least for users anyway. If someone does a search for “apple” then Google have a lot better chance of matching an ad if historically you’ve made a lot of searches for computer terms, music, or fruit; but for us it’s not worked out well. I agree, Google can’t lose here.
November 5th, 2007 at 4:53 am
[...] Their recommendation is also to just stop using broad match. The put the blame on the new feature on Google trying to meet Wall Street quarterly targets: [...]
November 5th, 2007 at 5:37 am
Thanks for the writeup Jim.
After we got burned in August, we followed your exact advice: don’t use AdWords broad match at all to avoid expanded broad match altogether. Check out the awful sales numbers I got out of expanded broad match and how much better things look without broad match at all.
March 5th, 2008 at 2:46 am
[...] Para más información ver aquí [...]