As the PPC community erupted in outrage on Friday bemoaning Google’s unilateral decision to drop real Exact and Phrase Match keyword targeting, I started looking for ways we at SpyFu could help solve the problem.
In my research, I found an article by Brad Geddes publishing some actual campaign data showing the volume and performance of Normal Match types vs Normal + Variants. This is a gold mine, because with that, we can see how big of a problem this is (or isn’t). Pretty much all of my conclusions in this article are based on Brad’s data. If anyone else could produce the same spreadsheet Brad did, that would be an amazing help to figuring out how to solve this problem.
Here’s what I found:
Losing Phrase Match is 10x More Damaging than Losing Exact Match
This was not at all what I was expecting. Almost all I’ve heard and read about were articles talking about the death of Exact Match. Yes, that’s a big deal. But losing Phrase Match is nearly 10x more damaging to your bottom line.
Here’s an example: If you spend $100k per month on AdWords, and all of your keywords are Exact Match, then you can expect to spend about $102,400 after “Close Variants” is switched on at the end of September.
If your account is organized by Phrase Match, then expect to be spending $117,200!
Put another way, if your 100k is a fixed budget, but it’s driving on site conversion – let’s say it is driving 1,000 conversions currently. Guess what? If you’re using Phrase Match, that same $100k will get you about 95 fewer conversions – or 905 total.
My heart is racing as I write this. I feel myself getting angry. I don’t know about you, but 10% of my conversions = 10% of my revenue, and dealing with a 10% loss of revenue is not what I need now or ever. Especially not when it’s from a “partner” I pay $100k per month to give a flying Firetruck what I think.
Okay. Reminding myself to breathe. (LOL, flying firetruck… I have toddlers)
Finding a Solution
When I first heard about the new announcement, I thought losing Exact Match was the biggest deal. As such, my first idea for a technological solution was to automatically identify negative match keywords for “Close Variants”.
But, given the problem I see with Phrase Match, my suggestion is this:
Convert All Of Your Phrase Match Keywords into Exact Match
That, IMO, has to be your first step, and I think you need to do that before next month. Once you’ve done that, once you’ve saved yourself from the 10% loss in performance, then we’ll figure out that 1%.
Building out negative matches may be the best solution there, but here are a couple questions that need to be answered first.
- Google uses the example exact match [floor] will match “close variant” [flooring]. What if you are bidding $2 for [floor] and also $1 for [flooring]? Since “close variants” makes those technically the same Exact Match keyword, when someone searches for [flooring] would you be billed $1 or $2?
- 2. What if those were phrase match? “Floor” vs “flooring” and someone searched for “travertine flooring:” $1 or $2?
According to Google’s release:
“Keep in mind that the AdWords system prefers to trigger ads using keywords that are identical to search queries, so you can still use misspelled, abbreviated, and other close variations of your keywords. If you find that performance varies significantly between close variants, you can add the better performing ones as separate keywords and adjust their bids accordingly.”
So, the answer to both questions is $1. Google will still use the“exactest” match possible. So, we don’t have to mess with negative matches nearly as much.
But, it means that the only reasonable bidding strategy is essential reaalllly long tail.
I’ll be talking with lots industry experts over the next couple days. But, right now, my best recommendation is: convert all of your Phrase Match to Exact; just build out those “close variants” as Exact Matches and set the bids low.
Okay. That’s where I’m at with solutions brainstorming now. You can be sure I, and my whole team at SpyFu are all over this problem. We’ll figure out the best solution no matter what.
But, one more thing I had to know…
How Much Money is Google Making on all this Suffering?
In order to answer that question, I need to do some serious extrapolating. I started with Google’s own investor financial reports here: https://investor.google.com/financial/tables.html. Then, I took the estimates we got from Brad’s spreadsheet and plugged them in. I just took averages of Phrase and Exact match, since I don’t know what the real world split is (I kind of suspect there’s more Phrase match as a percent of spend going on, but I have no data to prove it).
Here’s a key assumption I made: I know that while there are like 250k AdWords customers, spend is in no way a bell curve. It’s a power curve all the way – something like Pareto’s (80/20) rule applies. So, an assumption I made was that since at least 90% of ad spend is controlled by the top 5% of accounts, and the top 5% of accounts are most likely to use “sophisticated” match types, I said that 90% of Google’s gross revenue from AdWords would be impacted.
Anyway, here’s how the data played out:
Google will make an extra $3.8 Billion next year. That’s fine. I think Google should make money – lots of it. The thing that sucks is that they’ll be delivering value less efficiently, and that will mean that customers on the edge – within 10% margin really – will be squeezed out.
In the process of taking away Exact and Phrase match, Google’s going to deliver traffic that’s worth about $2 Billion / year at today’s rates, but charge $3.85 Billion for it. So, customers get to eat the extra $1.8 Billion – and they can’t opt out.
That’s only $7200 in lost value spread across 250,000 accounts – a pittance, right? Well, then there’s that pesky Pareto fella. If you’re one of those top 5% — or top 12,500 spenders – prepare to pay $129k more or see $129k worth of fewer conversions.
The Earnings Story
If for no reason but to serve as a point of triangulation, I wanted to see how this fit into Google’s quarterly earnings picture.
I hate conspiracy theories. I think that more often than not, Google makes decisions irrespective of revenue implications. I think that they tend to want to serve the searcher well. I do often wonder if they are trapped in their use of “do what’s best for lots of users” development focus when building AdWords (where all users aren’t created equal). But generally, I don’t think that they as a company are overly focused on pleasing shareholders at the expense of customers. I find them admirable in that way.
But, I had to check. Here’s where the numbers took me:
Based on this, I thought, “A-ha! They want to tell a story of 20% annual growth. Currently, they’re heading to 17%. So, turn a little dial and poof! Annual report looks nice again.”
Then I had a thought. To get projected 2014 revenue, I just multiplied their (Q1+Q2) x 2. Later I thought their Q4 might be higher (holiday ad spending). Sure enough, when I look at their 2013 and 2012 Q4 numbers, they see about a $1B bump (about 12%). When I take that into account the same spreadsheet looks like this:
No silver bullet of earnings-focused dial turning. But, that might be a good thing.
Maybe we Can Reverse this Change
It very well may be that Jen Huang, Google AdWords’ Product Manager, is genuinely trying to make AdWords an easier product to use. Really, it does take people quite a lot of time to wrap their heads around match types – maybe they want to reduce the “clutter”. That could be what’s going on here. Most users don’t opt out of “close variant” matching. Hell, most users don’t use Exact or Phrase match at all, much less Modified Broad. Maybe they’re making that mistake of treating all the users the same, whereas 90% of Google’s revenue comes from 5% of their users. Wouldn’t that be the type of mistake Google could make? Building a product that pleases the 99% at the expense of the 1%; helping the little guy – just a little too much?
Seriously. Maybe it is a mistake. Maybe it could be reversed. Maybe if we ask really nicely?
Update #1 (8/18/14)
This morning Sam Owen posted more raw “Close Variant” data along with some great analysis on PPC Hero. I was hoping someone would do that so I could compare and validate it against my conclusions I drew based largely on Brad Geddes’ data.
Sam’s analysis ended up being very mixed, largely I think, because the sample sizes of the individual accounts were a bit too small. One account had only 8 conversions for close variants on Exact Match. In split testing, I call this the rule of small numbers.
The good news is that data from all 4 of Sam’s accounts can be combined to yield a more statistically significant sample. So, I did that here:
For Sam, Exact Match performance was actually improved for Close Variants. But, like Brad, he saw a significant 7.64% decrease in Phrase Match performance.
Also, he saw a 41% increase in spend on those Phrase Match keywords. That $100k spend I mentioned before? It goes to $140k. Or with a fixed $100k budget netting 1000 conversions? For Sam, 1000 drops to about 924.
Basically, Sam’s data validates the conclusions I made early. Phrase match is a much bigger deal than exact and can mean ~10% drop in campaign performance.
Combining Brad and Sam’s Data
The obvious next step is to combine Brad and Sam’s data to get a bigger sample – and one less biased by campaign organization structure, bidding techniques, etc. Here’s what I found:
Now we’re looking at a nice significant amount of data across five different accounts from two different PPC managers.
The good news is that the two data sources back each other up, and they further validate the conclusions made previously this article. Based on this sample, it looks like we don’t need to worry about Exact Match at all – it may very well improve performance overall.
I still say, convert those Phrase Matches to Exact Matches if you want control of your spending and your performance.
Beware of Phrase Match Close Variants. They are Google’s dingo that’s about to eat your baby.
Update #2 (8/20/14)
I’ve got some new data to report today. It all supports my current conclusions, but these are heavyweight data from heavyweight industry experts.
- Larry Kim of Wordstream wrote me to say that he ran the same analysis internally and found that Phrase Match close variants drops campaign performance by 11%. Compare that to my current best estimate of 8%. Larry’s data is presumably based on an aggregate of lots and lots of SMB accounts that they manage bids for. As a side note, I’m doing a joint webinar with Larry on September 9th, and we’ll be talking about all this match type craziness.
- Mike Rhodes of WebSavvy sent me about $1.5MM more spending data similar to what I got from Brad and Sam in the past couple days. Mike’s data is across several accounts, so it’s a nice aggregate sample. It’s also very likely to be based on Google.com.au searches, which is pretty cool to see if there are any differences.
In any case, his dataset is large enough that it is statistically significant on its own. Here’s what that raw data looked like:
Behold the sheer number of converted clicks in this sample. It’s like 20x the size of the other datasets combined. Note that the “sum of CPA”, and “sum of cr” are mislabeled: I confirmed the calculations are what they should be for “CPA” and “Conversion Rate.”
Here’s my analysis of those numbers:
More numbers are telling us very clearly: It is not losing Exact Match that you should fear, but FEAR THE CRAP out of Phrase Match.
Actually, what’s interesting is that both of these samples are now telling me that Exact Match (close variants) actually improves performance – either by between 0.75-4.43%. The worst estimate I’ve seen is it decreasing performance by 1%.
At this point, the advice I’d give is convert as much Phrase Match as possible to Exact Match, and leave Exact Match alone.