Why am I seeing this ad? AI, ML & human error in advertising

Ad platforms create equal opportunities for businesses but not equal outcomes.

They’re mostly marketed as self-service and easy to use, however, there are new features added regularly and open-ended ways to set, structure and target. Meaning, countless ways to spend—creating winners and losers in advertising.

This is where machines and digital advertisers are needed, to provide a profitable outcome.

Enter AI, ML and experts as freelancers, via agencies or housed in some of the world’s biggest companies, equipped with ample data, tech and educational resources to match people with companies via ads on search, social, and elsewhere on the web.

But, are the machines still in infancy or too heavily relied upon and do the experts always get it right?

Well, how often are you seeing ads that are irrelevant to what you wanted or where you were or who you are?

An irrelevant ad is an ad paid for by the company advertising but can return zero value as it’s of no use to the person receiving the ad.

As a digital advertiser via my company Adboy.com, I’m always curious as to why I was served an ad and if the company paying makes or loses money from it.

Something I’ve noticed is that in easily avoidable errors, ads can be served to existing customers, people with irrelevant needs and people that can’t be or are far less likely to become customers.

With this article, I’m going to give you the lenses of a fastidious digital advertiser. You’ll spot errors like these for yourself and know how they could occur, what the negative impact could be and how they can be avoided.

Advertising to existing customers

GettyImages 925925584

Image via Getty Images / Olga Kashurina

While scrolling through your news feed, you may have seen a sign-up ad to something you are already signed up to. I can recall two recent ones, a free month trial for the music streaming service I already pay for and free ad credits for new accounts on an ad platform I’ve been using every day for years.

Both companies have an incredibly large user base so this isn’t a small loss. And, given that both companies are adtech, I suspect it common that companies spend to acquire customers, already acquired.

If you see a sign-up ad from a company you are already a customer of it’s likely that they are trying to reach people similar to you, or interested in their offering or because you are already familiar with the company. Targeting expansion is also an option and left up to machine learning to serve broader than what the advertiser has set.

With both of the ads I’ve mentioned, the advertiser may count acquiring me as a new customer if I logged in and used the apps after the ad was viewed or clicked, which I would have.

An advertising campaign like this is most likely set to get the most amount of users within the advertising budget. As a user already, I’m probably the cheapest and when the machine or expert learns this it will use my data points to spend on showing more ads to people like me, existing customers.

So, the expert in charge looks great in their reports as they show cheap new signups for the company ad spend, but in fact, it’s money down the drain and opportunity lost.

This is easily avoidable because whatever can be targeted can be excluded. When signing up, details like your name and email can be collected and when using an app, content that only a customer can access is viewed. This data can be used to create audiences in ad platforms like Google and Facebook ads, and added to a campaign as an audience exclusion, which would stop existing customers from seeing ads.

Advertising to people that can’t be acquired

I’ll draw on an example from a talent recruitment ad that was recently served to me (male, not a parent, Melbourne based) by a marketing agency. It read, “Are you a mum and parenting influencer based in Sydney?”

By clicking on the “Why am I seeing this ad?” button, I can see they want to reach anyone who visited their website previously, this is known as retargeting. This specific ad should only have been served to anyone who visited their site AND is a Sydney mum.

Just because someone can trigger an intent that the advertiser would like to target, it does not always mean they need to see a certain ad or any at all, as this is only one layer of the cake.

There are triggers for why an advertiser would want someone to see and ad like retargeting, keywords, behavior, context and interests.

And then there is the who their customers actually are, like gender, location, parental status, income and age. These are all things that can be used in combinations with the signal for intent to target their ads.

When an advertiser combines targeting of why someone should see an ad with who someone is, they better their chances of serving ads only to the customers that they can acquire.

With a few clicks, you should be able to see a similar mistake for yourself. Try visiting a bunch of online retail websites that sell both male and female-specific products then keep an eye out for the ads you’re served shortly after. You may see ads from some of these sites, directed to your gender opposite.

Advertising to people that are not the perfect match

Image via Getty Images / dane_mark

You may have been searching online for a service you needed and thought you found it, only to enquire and find out you and the business were not a perfect match. In many cases, the business’ advertising can avoid this.

Not all leads hold the same or any value. Sometimes, an ad doesn’t go direct to an online sale, but instead starts with an online conversion like a lead and then potentially converts to a sale in the offline world, like at a physical business or customer location or over the phone.

Advertising like this is usually optimized to get the most amount of online conversions for the advertisers budget. With all conversions being treated as equal and assigned the same dollar value.

However, just because the potential customer converted online, it does not mean they can, in fact, become a paying customer. A business then has to convert that lead into a sale and the rate that they convert leads to sales can differ by variables that a business can identify from looking at their own data.

Let me give you an example, a plumber runs lead generation ads to people 10 miles around his base (including the CBD) searching for terms including “cheap plumbers” and “best plumbers.”

From a cost per lead basis, all looks good but the plumber has to turn away a lot of work because in many cases he does not want to accept the job or the customer does not want to hire him.

So, he takes a look at his data to see if he is getting the right type of leads, ones that he can, in fact, turn to sales.

He discovers a few key things, the most sales come from areas he can get to within 20 minutes of receiving a call. He gets the most amount of leads from the CBD, however, he rarely accepts a job if it’s to an apartment without on-street parking or if it has a low clearance car park, common in the CBD. He also does some shopping around and learns he is 20% more expensive than most of his competitors in the same areas.

With this information, the plumber adjusts his targeting criteria. He spends more on the areas he can drive to within 20 minutes, excludes anyone in the CBD from seeing his ads, excludes any search terms including cheap, decreases bidding on low-income earners and increases on high-income earners.

The plumber finds that the optimizations increased the rate that he converts leads to sales. He spends less time on mismatched inquiries and gets more profit from his advertising.

Unless advertisers upload their offline conversions, the machine or expert doing the optimizations will treat all online conversions as equal value. Whereas, if the customer and business variables are known, the data can help target and optimize campaigns for increased profit.

Advertising to people with irrelevant needs 

Lastly, search ads are used to target people as they need something but they can also capture people with irrelevant needs at the same time.

The actual words used to search and find an ad can differ from the words a company has intended to target people with.

For instance, a tax accountant may be advertising for new leads and target with the keyword “Accountants” but receive traffic for “Accountant jobs” “Accounting bookkeepers” and “CPA Accountants”.

In this scenario, the tax accountant pays to advertise to and receives inquiries from people looking for employment and services outside of their expertise.

If the tax accountant is relying entirely on machine learning with campaign types like smart and dynamic to run their ads and this traffic appears to be performing well, it will continue to run and increase spending on this type of traffic. If the ads are being managed, it would have never appeared in search had they done one simple thing: specified that the words “jobs” “bookkeepers” and “CPA” were to be ignored in the event any user includes them in a search term.

image4

Image via Adboy.com / Adam Zelcer

You can find irrelevant search ads yourself in many industries by adding employment, education, niche and location words to your search, like in these examples below.

image3

image6

Image via Adboy.com / Adam Zelcer

image5

Image via Adboy.com / Adam Zelcer

Search campaigns have three keyword tabs. Search Keywords that you add to target, Negative Keywords that you add to avoid and Search Terms where you can see the actual words people used to find the ads.

image1

Image via Adboy.com / Adam Zelcer

Advertisers can prevent receiving irrelevant search traffic when setting up their campaigns by adding any irrelevant keywords they can think of to the negative keywords tab. On an ongoing basis, the search term report can identify any irrelevant traffic that can be used to add negative keywords to the list.

I’ll leave you with my favorite example of an error by experts—one that you’ve just learned to avoid yourselves. This one is actually of negative value to the advertisers and positive value to the potential customers. Check it out by Googling phrases like “worst AdWords agencies” or “worst Google ads agency.”

In the search results, you should see a number of names from the digital ad industry greeting you with ads at the top of the page. These are SEM professionals who are advertising their SEM talents via SEM in a bid to get SEM clients—and, in the process, their failure to make proper use of the platform is revealed.

Advertisers diligent about the four areas I’ve covered in this article would be hard to catch wasting a dollar on irrelevant ads—and, their advertising would be more effective.