Please learn from our mistakes

No-bullshit lessons in business and careers. One mail every day. 15k+ readers love it. Join in?

Oops! Something went wrong while submitting the form.
TODAY’S STORY
18 Mar
,
2023

Why am I seeing this ad?!?!? 🤬

Just type "why are you showing me this ad" in Twitter search and you'll be in for a hilarious time.

Facebook is no different.

Nor is Youtube.

From showing ads to non-customers...

...to making false connections...

...to targeting ads to existing customers...

...performance advertising in its current state is F**KED.

Last weekend, I was talking to a friend. She complained that despite using filters on e-commerce apps to provide the algorithm some context, scrolling through the still pretty long list of products and brands was annoying.

I concurred.

According to Wordstream, the average Click-through Rate (CTR) for Facebook Ads across all 18 industries was 0.90% in 2022.

Here's the industry-wise breakdown of the avg. CTR.

Even if you take Retail — the best performer when it comes to CTR, with an average of 1.59% — this still means ~98.4% audience being targeted with these ads still doesn't click on them!

It's abysmal, to say the least.

But it makes sense, right?

The only context these ad platforms have around why you're making the purchase is all the data they collect on you via your browsing history, cookies, snooping over conversations, etc.

It's all quite loose, unstructured, and frankly, devoid of any valuable information.

For example, I may be looking to purchase “running shoes”. But I do not provide any context for why I need the running shoes. Am I training for a marathon, trying to improve my fitness, or looking for comfortable shoes to wear for my daily commute?

Performance marketing today is almost akin to shooting arrows in the dark.

But performance marketers know this already. Today, Facebook advertising, to a large extent, operates on the spray-and-pray paradigm. Performance marketers prefer casting a wide and loose net over tighter segmentation, because it is generally known to give better results.

And this is fairly obvious, considering that when you don't know much about the exact context in which your target consumers make a purchase (of which there might be many), the best strategy is to diversify your bets.

It's similar to investing in the stock market. You either go all Warren Buffett about it and treat it like a full-time job, investing most of your money in a few well-researched businesses, or you go to the other extreme and diversify your bets by investing in an index fund or a bunch of mutual funds. Investing with half-knowledge is dangerous, and that applies to performance marketing, too.

Think about the time when you’re ordering food online when you haven't had any meals throughout the day. You want to make the decision faster. You wade through the sea of options and decide quickly. The intent to buy is high because you're feeling hungry. But that is not the case every time.

You may just be browsing out of curiosity, or be a marketer who is researching an industry, without any intention of buying. But the algorithm misunderstands that research as an intention to buy.

That's context. And that context isn't captured by the system in any way.

The result is you are soon bombarded with notifications from the app nudging you to buy. Sometimes, it even happens that you're shown ads when you've already made the purchase.

This gets especially annoying when the product has a long purchase lifecycle, like a washing machine.

“Dear Amazon, I just bought a washing machine. Why are you recommending more washing machines? How many washing machines one ought to keep in their house?”

The marketers on the other end are equally clueless about why you aren't buying. And so they keep throwing targeted ads your way, hoping that one day you’ll bite.

To be more specific, the problem currently is that the search queries we enter on platforms like Google and Amazon do not capture the context around which we are searching.

As a result, advertisers struggle with poor customer targeting effectiveness, leading to wasted resources and frustration. Without having enough context, most ads end up being irrelevant to customers. Most of us swipe away or swipe up from ads while scrolling through Instagram, LinkedIn, or Twitter. Many of us even pay for premium versions of apps, just to not have to see ads.

(YouTube Premium, anyone?)

If your brand ads are plastered everywhere, it creates a negative perception too. In spite of personalization, if the same brand ad is served to me on different platforms while I use the internet, I am bound to get annoyed and think of it as a desperate brand.

As a consumer, I may also have insufficient funds when I search for products and therefore delay buying. This leads to precious money spent on a customer who would’ve converted without the brand trying so hard. Or I may not buy for countless other reasons that the system is currently unable to capture.

So, the question is: Can ads get less annoying?

Assuming most of this annoyance has to do with irrelevant ads and bad targeting, I have a strong hunch that this interaction between marketers and consumers is on the cusp of a drastic change.

Enter Large Language Models.

By now, unless you live under a rock, you must be aware of a paradigm-changing new technology that's taking the internet by storm:

GPT-4 and all its derivatives.

With the conversational capability services like ChatGPT provide, think about what happens when every search — instead of looking like a human trying to communicate with a machine in broken English — has a chat with the search assistant, like having a chat with an in-store sales representative.

Currently, our search queries on Google / Amazon look like

"best camera under 50,000"
"lehenga for wedding"
"best running shoes for men"
"how to earn money from home"
"tesla share price going up news"

But imagine a conversational interface that asks you more details in natural language. A conversation could go like:

“Hey, I'm looking to buy pants”

“Sure. Are you looking to buy formal trousers or casual pants?”

“I'm looking for formals to wear at a friend's wedding”

*the assistant shows a bunch of results*

*you, after going through them*

“Can you show me something that would go well with an off white Chinese collar shirt?”

“Sure. Here are some pants that would go well with a Chinese collar shirt.”


*you go through the filtered search results again*

“Can you show me pants for someone with thick thighs and a dark complexion”

“Okay, here you go.”

“Okay, now show me options that will deliver to my address in the next 3 days”

“Showing options that will deliver to your address in the next 3 days.”

Another example where context would be useful is someone looking for running shoes.

A conversational interface that uses LLMs could ask consumers for what purpose exactly they are looking for running shoes. The consumer might respond by saying they are training for a marathon, providing valuable context to advertisers.

With this information, advertisers could then tailor their ads to the consumer's needs, showing them ads for high-performance running shoes or ads for nutrition supplements or coaching programs to aid in marathon training.

Some of this context is captured even today using cookies, but you get the drift. More context sharing will always lead to better targeting. And by integrating GPT-like language models into search and e-commerce, consumers would be able to engage in a conversation with AI, and do just that.

An additional benefit would be that the shopping experience would feel much more personal and warm instead of feeling isolating and hostile.

Imagine being a person with an unconventional body type and then seeing models with perfect bodies on e-commerce websites.

It feels alienating. But it happens today, and more often than you'd like to think.

Integrating these language models into search would be like employing a bit of personalized customer consultation before you advertise, which increases the probability of purchase while supplying additional context for future targeting.

Honestly, the way ads are served currently is similar to what happens when you crash a wedding.

You might have gone there for the food, having no knowledge of who the bride and groom even are, but the hosts (advertisers and businesses) think you’re there because you care about the wedding.

It's no surprise then, that ads today are not only ineffective, but they are also annoying. Both share the same underlying reason:

Lack of context under which the thing is being searched for or a purchase is being made.

I think GPT-4 can fix this, (like it can fix everything if you were to pay attention to Twitter Threadbois and LinkedIn Influenzas) but unironically this time.

Feeling Lucky?
Subscribe to get new posts emailed to you, daily. No spam.
Oops! Something went wrong while submitting the form.
2k+ business professionals act on our advice every day. You should too.
Subscribe to get new posts emailed to you, daily. No spam.
Oops! Something went wrong while submitting the form.
2k+ business professionals act on our advice every day. You should too.