Earlier this week, a few of us on the PPC Chat hashtag on Twitter were discussing dynamic keyword insertion. The different opinions were interesting.
Some people were not fans of using DKI at all:
Others, like me, use it a lot. The conversation then twisted and turned to a discussion of whether it’s ok to use DKI for competitor ads or not.
It was so interesting that I decided to expand beyond Twitter’s 140 characters here.
DKI must be used correctly.
There’s a huge misconception out there that DKI inserts the user query into the ad. It doesn’t – it inserts the keyword you’re bidding on. So one key is to make sure you’re bidding on keywords that you’re ok having in your ad copy.
Misspellings, for example, can be great keywords but terrible for DKI. If I’m going to use DKI, I put misspelled keywords in their own ad group and don’t use DKI there.
You’ll also want tightly-themed ad groups. Otherwise it’s nearly impossible to write ad copy that makes sense for 50 different keywords.
Tread with caution when using DKI for competitor names.
Part of the Twitter conversation centered around using DKI for competitors. I have done this successfully on more than one account, without repercussions. In my experience, the engines usually end up using your default text anyway, not the competitor terms.
But others had different experiences:
These are all valid considerations. Discuss the strategy with your client or boss before trying DKI with competitor keywords. When I worked in-house, my boss loved the fact that we were doing this. I’ve had clients who love it too. But I’ve had other clients who said no way – they didn’t want ill will with their competition.
Test, test, test.
As with most things PPC, DKI is worth testing. We inherited a client who was using DKI across the board. We immediately decided to test ads without it. What a mistake. Click-through rate plummeted like a cement block – I mean, CTRs were 1/10 what they were with DKI. That test didn’t last long.
Test DKI in different parts of the ad, too. I see it most commonly used in headlines, but you can use it anywhere. Try it in the middle of the ad, or even in the display URL. The results may surprise you.
Of course, there are other pros and cons to using DKI, as this Wordstream article points out.
What’s your take on DKI? Love it, hate it, don’t care? Share in the comments?
My vote is to lose it! My thought is, if you are taking the time to set up your campaigns and ad groups correctly, your ad text should reflect the keywords within the ad group. Through that process, you would not need DKI. D\
That’s what we thought with the client whose CTR dropped 10x without DKI. 🙂 Mostly, though, I agree.
I still love DKI though and definitely vote to Use It. 🙂
This probably is a naive question. Does Google provide a report about how often the ad was shown with DKI?
You mean with DKI vs. default text? Not that I know of. That would be useful for sure.
I have always been with Justin and Andrew on this .I felt the cons outweighed the pros. However, I am starting to sway in certain situations. As you said, test. Not doing it a lot just yet, only in certain cases where I think it would be beneficial. However, even in those cases, I am A/B testing against a static ad. The CTR is usually good on DKI ads, but that doesn’t always mean it’s the best converting ad.
As for DKI in competitors adgroups. I’ve never done that, but may test. Expanding on my tweet, I would think that it could generate people searching the business name, not because they are in the market for similar services or products, but because they need the contact info to talk about something else – complaints, reschedule, etc. But if the CPC is low and the conversion rate is high, it may work. But again, as Melissa said, that needs to be discussed first and then tested.
I think I’m the only one on the “use” it train, but I use it very, very sparingly. I find myself with really, really long product names – think 35 or so characters – so they’re impossible to use in ad copy with any sort of real relevant context. You can sneak in wayyyy longer stuff using DKI or dynamic headlines.
My record is a 40 character headline 🙂
You’re not the only one. I use it. 🙂 Great example of using DKI in an ecommerce setting!
I meant the only OTHER one. I’m curious who these “others” are you speak of in the blog post though!
Well, I see enough DKI in the wild to know that others use it. 🙂 Some good, some bad…
DKI has always thrown a level of complexity into the testing model that made it difficult to use with any certainty.
As others noted, I always felt that if I have properly themed my ad groups and ads then I have accomplished the intention behind DKI, right? Many of our adgroups are very tight with 10-15 keywords. Most are represented within the ad text or title. I’m covered right?
Well that is not what I am hearing from you. DKI has given you that one extra push of relativity between keywords and ads. Whereas I have 3-4 ads and 10 keywords, I could be missing out on the relationship with the remaining 6 keyword variations. And that is the lift you are seeing that is worth pursuing. Am I understanding you correctly Mel?
Solid topic once again which denotes the importance for all SEMs to take the time to revisit and re-evaluate features and functions.
Exactly, Jerry. I’m saying for many of my clients it gives that extra lift, and over a large set of keywords it definitely moves the needle. Many accounts won’t need it, but many can benefit and IMHO it’s always worth testing.
I am kind of neutral on DKI, but I have to bring up this aspect – how do “close variants” play into DKI? I know that the actual query is not inserted, but I would assume that “close” variants trigger ads? Based on what I’ve seen insofar as what can constitute a “close” variant, I have hesitation about using DKI more. I should note that this is also an issue even for tightly themed ad groups not using DKI.
That being said, it clearly works well under some circumstances, so like everything else in PPC, it would seem, “it depends” is the correct answer once again!
Good food for thought though. Would be interested if others shared some case studies so we might be able to figure out if there is any rhyme or reason as to when it works well.
I haven’t seen a decline in performance with close variants. The only oddity is that it puts your keyword, rather than the user query, into the ad – but it’s always done that, so broad/phrase match keywords might not exactly match the user query anyway. Either the close variant is relevant, or it isn’t – IMHO DKI doesn’t affect anything in that case.
As you say, it depends! Thanks for your comment as always!
What a lovely subject. Reminds me of black and white hat SEO 🙂
As in everything. I would say it depends.
But what hasn’t been taken up at all is the point of organization. Especially for smaller clients where time is limited, DKI is beneficial. I have a client whose account includes tons of keywords but not too much budget being spent due to limited queries. Instead of breaking the keywords into tons of ad groups which makes testing a pain, I just split them up into a smaller group of ad groups and using DKI add the keywords into the ad. The message varies based on keywords due to the organization but they match the query often. That safes me tons of time of bloating the organization which makes optimization too hard. CTR is great and in this case the CR is a dream.
Another aspect which isn’t mentioned is the fact of headline length The headline is limited to 25 characters but Google often makes an exception with DKI. In German, some words are ridiculously long and I was only able to get the keyword into the headline using DKI. I think it was 28 characters.
I can’t wait to the point where I can steer dynamic ads with “positive” keywords because, also here, Google makes use of longer headlines.
Of course using DKI with keywords that don’t match the landing page is a bad idea and will end up in low CR.
I think DKI cam be wonderful if used correctly! That applies to many things, though. If done correctly and in the right amounts anything can be great. It just depends!