Automation Is Great. But With Great Power Comes Great Responsibility. This Video Explains The Risk of Using Bidding Algorithms During Peak Trading Periods.


Hi everyone, I wanna quickly talk about Google Smart campaigns. If you’ve not run them, it’s a fully automated ROAS campaign, so return on actual spend, where you say the product, you say the margin, the system goes off and generates those sales. I am convinced if you are running Smart campaigns during the Christmas period, you are not making as much revenue as you could do, and you’re probably paying more per sale than you need to.

There are two main reasons why this is happening. So actually, just before we go into that, we made a big decision yesterday. So the day before Black Friday, a couple of days before Cyber Monday, we have pulled, and we have binned all of our Smart campaigns. I just wanna talk through our reasoning why. So, number one, the algorithms are still not good enough, effectively. You might have noticed, the red flag that we noticed, well, we know for a set of products, the keywords that drive good ROAS, the ones that we know we should be there as much as possible. We’ve, on numerous occasions, searched for these terms, through different devices, browsers, et cetera et cetera, used utm_psw=0 parameter, and we’ve not been showing up.

So this was a little bit of a red flag, that the system wasn’t, hadn’t come to the correct conclusions. So that led us to believe there is actually a bit of an inherent flaw within the way that the Smart campaigns work. So if we take, a standard, I mean this is a relativity normal distribution for the keywords, verses the amount of revenue they would generate. So, it’s along the lines of the 80 20 rule, 80% of your revenue, coming from 20% of your keywords. We would recommend managing these in two or three different campaigns for each product. So historically we would have built, so let’s have one campaign for our rockstar keywords, and one campaign for our long term, you might think around brand terms, non-brand terms, competitors names, make and model type queries, and you would break this chain down, and manage each individual block of keywords in a slightly different way. But, with Smart campaigns you can’t do this, and what we think is happening, is our long tail is being supported by our rock star.

So our rock stars can’t fully rock, because our long tail is being supported by it. So, we’re not generating the revenue we would expect, and we are overexposed in the long tail, is our conclusion of what’s happening. So number two, the Smart campaigns are too slow to react. We all have noticed that the invention of Black Friday and Cyber Monday has fundamentally changed the way that people shop. So generally, they’ll shop closer and closer to Christmas, try and take advantage of sales, and the window between final Christmas delivery, and the kind of manic peak of online shopping, especially in the UK, is getting shorter, that window is getting tighter, so there’s less time to make changes. There’s less data so that you can evaluate during that period. So we need to enter that period in the strongest way that we possibly can. The Smart campaigns are not gonna be quick enough to make this conclusion. So we think Google’s documentation, Google recommends 50 conversions before evaluating any ROAS or Smart campaigns, sorry. So that’s 50 sales from each product that has been featured, not 50 web transactions or 50 checkouts, and if you’re a relatively high ticket product, you might not sell 50 a day, so say you’re targeting 10 sales a day, for argument’s sake.

That could be a five day period before the ROAS’s bidding strategy is accurate again. That’s far too long. So yeah, yesterday we bit the bullet, completely removed all the Smart campaigns within our accounts, and within the first hour of the disconnection, the products we disconnected have made more revenue than they had done the previous day. So, Smart campaigns are good. I have completely reevaluated the way that we use them, because there are some significant risks. Number one, they’re flexible, they don’t give you data, you can’t give them data in order to help them succeed. They are not flexible enough for short-term changes in competition, margin adjustments, so yeah, Black Friday, you’ve halved the price of your product, so your margin has halved. If you change your ROAS returns, so you could go from 500% to 1000%, so 20% cost sales to 10% cost sale, you’ve reset the algorithm. We’re back to that five day potential window. And the other thing that we’ve noticed, that we kinda hadn’t really considered when set them up, is, they suck impressions away from other campaigns.

So if you’ve got the same product featured in a Smart, and standard PLA, Smart seems to prioritize. So, in standard PLA you can set a priority high, low, medium, one, two, three. But, Smart seems to prioritize everything, and if it thinks it will work, it will nick it into a Smart campaign, so existing campaigns will suffer because of running a Smart campaign. So yeah, kinda do what you want with this information, but be very very careful entering this incredibly busy period using Smart campaigns. Cool! All the best guys, cheers!