What You Can Learn From People’s AI Marketing Failures

28 March 2021Ashley Maxwell8 min read

Touting the benefit of learning from mistakes is a little misguided since it suggests the world’s biggest losers would also be the world’s finest teachers. Learning from mistakes is never going to be as valuable as learning from success, but even so, here are a few AI marketing failures that are worth considering, even if only for their entertainment value.

Typical Affiliate Tool Mistakes

The AI marketing failures that have come about because of sophisticated affiliate tools is staggering. From the user who dedicated a budget of thousands to his product without categorising, and lost thousands because his product had the same name as a popular movie at the time.

To the affiliate marketer who figured that Geo-targeting was pointless because her adverts were in English, and figured only English people would be interested. Who was then dismayed to see that a small village in Africa had racked up 600 clicks on her adverts without a single sale.

What You Can Learn From Typical Affiliate Marketing Mistakes

Here at Vertanet, the first thing we drill into our marketing interns is “Small scale testing first.” Run tests, run as many as you like, but keep them small. If the marketer using the same keyword as a popular movie had set a daily budget of £3 rather than £3400, his losses wouldn’t have been so steep. If the marketer who neglected Geo-targeting had kept her project small, she would have noticed the odd click-through-rates in Africa and would have blocked the leak before the boat sank.

Target Guessed a Woman Was Pregnant

We created a blog post called, “What You Can Learn From Other People’s AI Marketing Successes,” and in the conclusion, we discuss what Target did. Their AI marketing tools are just as powerful as those used by other large chain stores, but they made the mistake of sending offers and congratulations to a woman for her new baby. Trouble is, they sent the messages before even she knew she was pregnant. Their AI was able to guess she was pregnant. Based on her purchase history within her Target online account, and so the marketing AI sent out congratulations and a bunch of new baby offers.

What You Can Learn From The Target Blunder

In truth, most modern marketing AI tools being used by large chain stores are able to guess things like who is pregnant, likely to get pregnant and so forth. The mistake Target made was to allow its AI marketing tools to make such a bold marketing move. It serves as a great lesson to all AI marketers that your learning computers do not understand discretion or subtly. They still need a human’s guiding hand, otherwise, your company ends up in the news because it started congratulating people for being pregnant before even they knew they were pregnant.

The AI Selling Blunder

This one goes back a bit, but there was a very large DVD and CD sales company that bought massive quantities of DVDs and CDs for just pennies at a time from ordinary consumers. It then sold the same DVDs and CDs on Amazon.

They were making a lot of money, and it didn’t really matter how much they charged because they were buying their products for pennies at a time. On Amazon, this company always seemed to have the lowest prices. This is because the company had an AI-powered selling program working for them that would lower the price of its products to just a few pennies below whatever the lowest seller was offering.

For example, if you were selling season one of Farscape in “Very Good” condition, and you were charging £5.30, then the AI program would change its price to £5.28. No matter how many times you lowered your price, the day after, the company would have beat you on price by a few pence.

However, some savvy consumers, such as myself, started to recognise the pattern. So, let’s say I wanted the box set of the Simpsons, I could pay the company’s price of £25. Or, I could list it on there myself for £5, and wait for them to lower their price to £4.95. However, what if other people saw my advert and bought my fake advert? The answer was simple, in my product description, I stated how the box was ripped, the DVDs were taped together, but I think they are in good condition anyway. Nobody bought my DVDs, the company AI lowered their price to £4.95, I bought their box set, and I removed my advert.

This went on for quite a while as I built my DVD collection. Then, I started telling people in my University. Before you knew it, the AI-powered company wasn’t as profitable as it once was. The AI system was taken offline, and they ended up selling their company to Music Magpie.

What You Can Learn From AI Selling Blunder

I wouldn’t like to take the credit for making the AI-powered company unprofitable, even if they were exploiting people with no money and taking their DVDs and CDs. I am sure I was not the only person who recognised the AI-selling pattern, and I am sure it wasn’t just my University that suddenly became wealthy in second-hand DVDs. Nevertheless, the company responsible could have fixed the problem very simply. Firstly, they could have examined their metrics and noticed the shift in behaviour. At which point, even a quick glance online would have shown them the problem. Secondly, they should have set a baseline price for each of their products, so the AI wasn’t able to sell at a loss. This may have meant their staff had to make some big adjustments in terms of purchase costs recording and price matching, but the fact they didn’t make the changes suggests they sat on ass for the entire time their business was becoming unprofitable.

How Steam Has Higher Than Average CTR on Suggested Gaming Offers

This is actually something I wanted to cover in the article, “What You Can Learn From Other People’s AI Marketing Successes,” but I had already taken up so much of the page writing about superstores that I didn’t have room to mention Steam.

Steam is not an example of other people’s AI marketing failures. Steam is a success story with Click-Through-Rates that make others weep with envy. They suggest games to people through email marketing, through queue lists, and through the Steam GUI itself. Steam gathers data from your account. Here is what they learn about you:

• What games are already in your library
• Which games you have searched for recently
• Which of their gaming offers you have clicked recently
• What you have put into your wish list
• What you have clicked “Ignore” on within your queued list
• Which games you have reviewed
• How long you have played certain games
• Which demos you have tried
• Which gaming live feeds you have watched on Steam
• How much you have spent on games over the course of your account
• The average amount you spend on games
• Which times of the year you are most likely to buy
• How soon after release you buy games

If you have ever asked yourself how Steam is so good at compiling queued lists of games you may like, or how they somehow have an offer that is perfect for you every month, then the information gathering above is the reason why. Their AI marketing tools have a devastatingly high amount of very strong very reliable information to go on. Not to mention that when you look at games, you can ignore them or put them on your wish list, which is a big indication to Steam which games you genuinely do not want. Very few other businesses are able to get their potential customers to specify which products they definitely do not want.

What You Can Learn From The Steam Gaming Platform

Gather as much information as you can (legally) on your users and how they interact with your website, platform or network. Then, use that information to test your users to see how they purchase. Steam has all these areas where they can collect information because they created them, and you should do what you can to create information sources too. This especially includes things like getting user feedback. Letting people post reviews is becoming popular, and is a ripe area for collecting information on people. But, Steam’s “Ignore” button is a mark of brilliance, perhaps try something similar yourself. Steam’s queued lists of possible products is another piece of marketing genius that perhaps you can offer your account holders too.

The UPlay Website Converts Terribly

Ubisoft is not a bad, evil or manipulative company, but Ubisoft has made a few frustrating business and marketing decisions in the past. Yet, despite a few bland and robotic marketing decisions, Ubisoft has a pretty good online reputation, so it is difficult to understand why their Uplay store converts so terribly.

If you become intimate with their Uplay store, you see that they are failing because even though they have the same capacity for information gathering as the Steam store, they are just not using it. Uplay could gather just as much information on users as the Steam store does, but if Uplay “Is” gathering this sort of information, they are using it terribly!!!

Take the example of “Far Cry 5” and “The Division 2.” Both have angry characters, guns and shooting, but that is where the similarity ends. Far Cry 5 is an immersive, character driven, narrative story-heavy game. The Division 2 is a mindless shooter looter where fun is derived from levelling up. It is possible that fans of one game would be fans of the other, but it is also true that fans of puzzles may be fans of tennis. Yet, Uplay ran six-month campaigns where Far Cry 5 players were bombarded with The Division 2 adverts. Lovers of Far Cry 5 were yearning for an “Ignore” button.

Are Uplay learning their lesson? Probably not since the cross promotion of “Assassin’s Creed Valhalla” and “Immortals Fenyx Rising” has been an epic failure. Most hardcore gamers will tell you that those two games are very different, as are the types of people who enjoy them. But, you can’t tell that to the UPlay AI marketing computers.

What You Can Learn From Uplay’s Terrible Conversion Numbers

Though the mistakes Uplay are making are obvious, why they are getting it wrong is a little unclear. Obviously, they are not mining as much information as Steam, but even then, their examination of user play times and purchase histories should have helped them avoid the mistakes they are making. This leads to the only conclusion left, which is that Ubisoft is trying to force products on people who do not want them. Like the child who nags you to try his Plasticine sandwich because you don’t know you won’t like it until you try it. If one were to vastly underestimate Ubisoft, one may assume that their marketing department doesn’t understand their own games. “Assassin’s Creed Valhalla” and “Immortals Fenyx Rising” both have players with swords, and “Far Cry 5” and “The Division 2” both have people with guns. Could Ubisoft’s marketing department really be that out of touch? In short, use the information you gather, don’t force products on people, and understand your own products.

Final Thoughts – Get Involved

Be it the AI marketing mistakes mentioned in this article, or those perpetrated every day by companies that should have known better. The common factor with almost all of them is that a human hand could have interviewed at any time and stopped the disaster from happening. Whether it be the company the lets it AI-selling computers price its products so they cost less than the delivery fees, or the massive gaming corporation that doesn’t understand its own products. It seems that on every occasion, the damage done could have been mitigated, lessened, or avoided, if human decision making had re-entered the process.

AI marketing seems at face value to be the sort of place where human decision making should be left out, but even a child needs stabilisers on its bike before it becomes a champion rider. Plus, the point of AI isn’t to let it do all the testing and all the work, it is to make your job easier and/or more efficient. Taking yourself out of the process, and then wondering why things went wrong, seems a little short-sighted.