Search horror stories

Happy Halloween!

For an extra spooky episode, Head of Search Niki Grant, Digital Account Director Jodie Brookton and Digital Account Manager Tobie Jackson sit down to share their best search horror stories.

Available to listen to here.


Hello and welcome to an extra spooky episode of Unmuddled, a podcast from The Kite Factory where we take complex media matters and unmuddled them. I’m Nicky Grant, head of search at the kite factory. And today I’m joined by Jody Brookson, account director, and Toby Jackson, digital account manager. Hi everyone. Hello. Hi. Now, as Halloween is upon us, we’re taking the chance to share our best search horror stories. So think creepy campfire stories, but geekier effectively. I think it’s safe to say that most of us in media have examples of horror stories. And this is where something has gone spectacularly, often unexpectedly and sometimes completely inexplicably wrong. So we should clarify upfront that there are a few causes to such horrors. So from human error to feature bugs. But to save everyone’s blushes, we won’t be referencing any specific brands or people or companies. However, there’s a lot of trial and error involved in media, so we’re sharing some of our errors that we’ve seen to help you avoid the same traps. So we’re going to start off with one of the most common causes, which is human error. So loads of horror stories I think, occur in the industry as a result of human error. My best example of this, the kickers off is a very busy time for a client where someone in the team who didn’t usually work on that count offered to make some campaigns, to build some campaigns. They weren’t entirely familiar with the naming convention and they missed one letter out of the product code in the campaign name. Now unfortunately, all of that budget tracking was working off of a lookup in an Excel sheet for the campaign name because that look up couldn’t find that campaign. It wasn’t tracked. And what I ended up with was at the end of the month doing our finance reconciliations. And if you work in a media agency, you will know the horror in itself that that instills just to find that the overspend was 228 £0. It’s pretty big for missing out one letter. So this is the kind of stuff that we’re talking about today. Toby, have you got any examples of this? Yeah, I do unfortunately have a couple of examples myself that I’ll go through. So the first one is around bid strategies. So I’m pretty sure anyone who’s worked PPCs used bid strategies. Great things most of the time, but with a bit of human error in the mix, can go wrong. So at the moment we’re probably all preparing for Black Friday, the Christmas period, very, very busy period, especially for clients in ecommerce and industries like that. However, bid strategies at this time can be a best friend, but can also be the worst enemy. So an example I’ve seen is around bid strategies being used, black Friday or the festive period. So at times when your volume of sales and conversions is going to be rising to levels you wouldn’t see at any other period in the year. So by setting the strategies with constraints, so for example, target CPA or target robots, we can find that usually the targets that are set work fine. But when we enter these periods where everything’s changing, we can see that it throws the whole bid strategy out of whack. And actually what I’ve seen on the other end of things is setting it too low at times can mean that the ads don’t serve at all. So have some campaigns where it won’t be served at all during these peak periods due to the fact that humans have put two stringent limits on the campaigns. So this is because the data is not normal, I guess because it’s a weird time of year, because it’s Black Friday or something, that sample data just isn’t up to scratch. Exactly. And I think we’ve all seen as well around COVID and everything that’s been happening this year, with The Queen’s Death, for example, that things can change very quickly, perhaps on a daily or a weekly basis. So if we have these really strict rules applied to the bid strategies, it can mean that we face the same problems even outside of these peak periods too, which is tricky because you can’t avoid data going up and down. And we generally like peak periods, just not when we need to refer back to them. Exactly. So I’d say the best thing to do during these periods is to just go with a more flexible bid strategy to maximize conversions or maximize revenue will just mean that Google can flex it. So if your CPCs do increase, or if things do go up and down, it just gives you that flexibility to keep serving and keep things going and help you set your goals. Nice, I like that little bit of damage limitation there. Exactly. So moving on to another one I’ve got is around rules and scripts. So again can be really useful within the account. But something I’ve seen with rules and scripts go slightly wrong is when team members perhaps go on holiday or somebody’s left the account and hasn’t done quite an indepth enough handover and things can be left going on in the background. For example, I saw one on an account which was raised the budgets of a campaign by 10% if CPA is below a certain level. Now, when you keep applying a 10% day after day after day with it going unnoticed, you can find that the budgets go pretty high. That’s a lot of sense. Exactly. So yes, the budgets can go pretty high and end up finding that everything’s gone before it should. So definitely want to avoid that. And with scripts you can take that step further because you can’t even identify necessarily what a script does unless you can read the code. So again, you might find that unexplainable things are happening in your account that are due to script being left in there from somebody who used to work on it. So do your account hydrogen guys, is the lesson learn now almost as if there’s a ghost in the system. Jodi, what about you? Have you got any examples of human error? Yes, I do. I think it’s a reason why we all have quality assurance checklists when we launch our campaigns. But essentially what we were trying to do for a client account is grow the traffic so we kind of reach the ceiling in our current CPA, but we don’t generally bid on broad keywords because it does result in a high CPA. What you can do is apply audiences over the top of your search campaigns. So essentially it shows to only people that fit within your specific remit and are searching for the keywords that you want. So it’s really niche, right? So you can get really targeted and keep your costs down. To get those conversions within the Google Ads platform to apply this, there’s two options. You can do observation and you can do target. What we wanted to do was Target, but unfortunately there was a missed checkbox where it was set as observation. Oh, no. Yeah, so what we ended up doing was spending a lot all in one day, which unfortunately added up all that budget when that issue could have been solved by just having a quick quality assurance checklist that said yeah, it definitely says Target, not observation. So that’s something that we called in to look at. Yes, I’ve also seen something quite similar with an account I’ve worked on. So they were trying to use the same tactic that Joseph spoke about for remarketing, but again applied the remarketing audiences as observations that are targeting for the clients, saying hey guys, why is our remarketing campaign doing so badly? It doesn’t seem to be doing any better than our generic and prospecting campaigns. The reality was wasn’t actually remarking audiences. They were there, but they were just being observed amongst every other audience out there. So definitely a setting that you need to pay attention to even though it seems a bit inconspicuous within the interface. You know how carpenters have that measure twice, cut once, so you don’t cut something the wrong length that you is then irreversible. And I think that’s our same model of click once, but look at it 500 times if we say that it’s very rarely the problem that caused the mistake that causes the problem, though it’s often how long it’s left until someone realizes there’s a mistake that causes the problem. But I’d say your remarketing example is actually segued beautifully into what I wanted to discuss around the kind of technical horror stories that we have. Because whilst we experience human error and we try to avoid this by limiting the human input with automation and with technical features and that kind of thing, actually even those can go wrong. So you mentioned remarketing and how it suddenly started going really badly. I was working with a business a few years back who were trying to increase the usage of the app instead of people going to their website. The problem was we could only promote this particular product to their existing customers. So the way that we were doing that was using, as Josie mentioned, the target function for audience to say, we are only going to target people that fall onto this audience. That audience was collected via a remarketing tag on their website. Now, what effectively happened was the better we did at driving people towards the app, the fewer people went to the website, fewer people hit the remarketing tag. So the membership duration just started running out on everyone. And when you looked at the graph within the remarketing section, you think, what has happened to my life audience? Why has this suddenly disappeared? And it transpires. It was just because we actually had two completely conflicting strategies. One of them that said, we have to get people to the website to put them in that list to target them. And the other that said, goodness gracious, we’re going to keep everyone away from the website to try and get into the app. So even though there’s still a human element to it, we’d given the tag and instruction, we’d given the campaign an instruction, and unfortunately, it just kind of ran with it regardless, although we were very proud when we figured out that’s what it was, because it was a little bit of a riddle. Can anyone make me feel better by giving another example of these? Yes, I’ve got one here. I wouldn’t say it’s necessarily as down to the machine as yours, but keyword insertion, dynamic keyword insertion within ad copy. Something I haven’t used as much in recent times. I’m not too sure about it. I don’t know what you guys thoughts are on it as a general tool to be used in PPC, but an example I have was with a retail client, and I think this is one that probably rings true for the retail industry in general and anybody who’s thinking or has used dynamic keyword insertion. Now, the way it works for anyone who isn’t aware is that when the user searches on Google, on the search engine, they’ll find that the ad itself inserts the keyword they’ve used. Now, this can be really helpful because it makes the ad copy super relevant. It makes it look like it’s added, tailored exactly to what they’re looking to search for. However, if they’re searching for something slightly obscure or slightly weird, maybe something you don’t stock within retail can be a bit of an issue. So if somebody’s searching pink fluffy socks and you just sell socks and they’re definitely not pink or fluffy, you’re essentially false advertising. You might get clickthroughs. I think they’re going to find that perfect pink fluffy socks for Christmas or Black Friday, and in reality, you haven’t got any. Wouldn’t you be disappointed as well, if you were very excited about getting your pink fluffy socks, I mean, I know where to get mine from. They’re already sorted. So, yeah, I’ve got to worry about this with any new entrance on our Qwen search and beware. There was an instance of this with Ebay years ago, I think it was going back 2016, something like that. And they had nosebleeds Divorce was one of them, that they had winning lottery numbers as well as well. And I think it was effectively a broad match campaign because they were well, we sell everything and obviously it’s user generated their listings. They don’t even really know what they sell until people go to list it. So they were like, I don’t know, maybe someone will try and sell a divorce on ebay, who knows? So it is a perilous approach, I think. How about you, Jodi? Have you seen any machines go rogue? Yes, a little bit rogue. So, a few years ago, I was asked to consult on a client who had seen their CPA go through the roof in their brand campaigns. And for those people that run paid search, you’ll know that the brand campaigns are always the most efficient. So this is a massive concern for them because the majority of their conversions came through brand and now they were paying. It was almost triple the CPA that they had before. So we did a little bit of digging, had a look at what changes have been made in the last couple of months and it transpired that a new bid strategy had been added for the brand campaigns. This was an impression share bid strategy, which to me makes complete sense because if you want to own all the searches for your brand, you would want to aim as close to 100% impression share as possible. What had happened though, is the algorithms for the bid strategy had caused the cost per click to rise from £1 to around £7. That’s a big horror story, right? Halloween sound effect. Exactly. Adding all the ghouls. So, yeah, I was happy because we found the solution. It was to add a bid cap. But what this meant is that the algorithm had essentially just seen the opportunity to spend more and more and more despite there being very low competition for that particular keyword, because it was a brand term, it’s trademarked. It was a brand that has got a specific name. So you wouldn’t actually accidentally buy it if you were typing in something more broad. The moral of the story is that if we don’t look at the performance of the bid strategy on a regular basis, once we set it or analyze where our current CPA should have been and then set the bid cap, it can go wrong and the system will want to get more of your money. I think that’s the challenge with these things, isn’t it? It gathers momentum. So once it’s started, until you identify what it is, that algorithm, that strategy is going to continue to kind of inbreed the problems that it’s had from the beginning. Exactly. It kind of brings me onto my point I was thinking of before, but yeah, just to build on that, we still sound really strange recently. So again, related to brand, so it was a client we were working with brand terms, not pure brand, but sort of brand with something else, which again, usually you’d find fairly low CPCs. It’s unlikely that you’re going to have too many competitors in this marketplace voting against them because of certain rules that apply. Anyway, we found in a certain campaign, lots of money was going towards it. The CPCs again were super high than we’d ever seen for anything around later before. And again, this makes me think that the machines weren’t necessarily working as well as they should be, split it out into its own campaign. So outside of our standard brand campaign and actually the CPC’s dropped about a quarter of what they were before. Same bid strategy, same targeting, same keywords. But as you said, breaking the trend of that bid strategy, continuing to do what it did seem to do the trick. So yeah, one that might be worth a try again, if you’re facing a similar problem is just rebuilding, spitting out and seeing if that does anything. Yeah, and it’s the same with budget as well, isn’t it? If there’s too much in one place or if there’s too much too many assets, too many keywords, too much targeting under one budget, you get a Hungry Hippos effect. I don’t know if you want to explain the Hungry Hippos effect for anyone that wasn’t alive in the 80s or early. I was born in 93, I’m within the 90s as well. But other people seem not to have the joy that is Hungry Hippos, although you are. It’s a very good game. It’s a very good game, a board game, if you could call it that. They play when I was younger. It definitely has given me some problems with my wrists in later life. But essentially you’re tapping on a little lever and you have a hippo that’s mouth extends to try and grab as many plastic balls from the center of the board as possible and it can get very fiery and irate. So a couple of arguments over it, but good fun nonetheless. So if you’re looking for a Christmas game this year, if they still sell it would recommend but going that back to face search. So I think with anything that’s shared within Google, there can be this issue between Google as a means of making money and for themselves and advertising, they will often push towards the hungriest hippo in that sort of marketplace. So if it’s a campaign and you have multiple ad groups, you can find that an aggregate with the most potential to spend. So the broadest keywords that is the campaign that will end up spending the most, even if it’s not the most effective campaign for your goals. The same thing can happen with shared budgets across multiple campaigns or portfolio bid strategies. Whenever something shared, you’ve got to be careful and look into how the money is being spent, because you can’t necessarily trust that Google is going to push it towards the place to achieve your goals. It may just go to the place that has the hungriest hippo who’s going to spend the most money. Yeah, and I think it’s probably worth adding. I know that there are a lot of a lot of people out there whose immediate response would be, well, it’s Google. They’re there to make money. Obviously, they’re just going to spend as much as they can. But arguably, if you’ve informed a system that you would like to get traffic and you’d like to get conversions, and that system identifies an opportunity to get higher volume, it’s what it’s going to do. So it’s down to those parameters that we need to put in place to make sure that we’re not allowing it to run riot. Exactly. And just keep an eye on everything. I think as it goes, there may be times when you need to flex it and step in and make changes. Other times it might just be a case of cost or higher, but we’re still doing well and just keeping the client informed. So I think it’s just a balancing act of trying to keep an eye on the trends, making changes where appropriate, or just reporting on it and letting it run. If you trust Google doing a good job. Absolutely. Jody, have you got any final tips for anyone looking at avoiding horror stories in the future? Yes, I think create as many QA spreadsheets as possible is my advice. Things get lost in translation. You can train your team as amazing PPC specialists, but there is still that element where there’s new things added to Google Ads and Bing Ads all the time, where you can’t necessarily, like, make sure that everything is set down. So once you see those new things come out, think, okay, what should I add to my QA checklist because of this new feature and you can live a happy non horror story life? I think that sounds like a fantastic idea. I suppose my element to add would probably be to audit every so often these kinds of automation to the bid strategies you’re running, because I think I always refer back to this again, old references, talking old money. But the quote from Only Falls and Horses would trigger with his broom that says, I’ve had the same broom for 20 years. It’s had seven different heads and 14 different handles. And it’s kind of what search campaigns are, that they Frankenstein’s monster over time with all these different elements. Sometimes you just need to take stock of those elements that you’ve got because they’ve augmented. So much over time, it probably resembles nothing like what you actually think is in place. So doing a bit of an audit and just seeing what you’ve got available every so often might help refresh your memory and make sure that you’re updating anything that does need to be updating. Thank you so much for listening today. This has been an episode of Unmuddled from the Kite Factory, and we hope you have a very happy Halloween.