Published: March 7, 2025
[Note: I first wrote this essay in 2023 with the intent of turning it into a video on YouTube or something... but that never happened. It's too important to keep to myself, so I'm publishing it here. I have expressed a lot of personally held opinions about a fundamental aspect of an industry I work in, and the opinions I express are not necessarily the views and opinions of my employer, which I have taken great care to not name on this blog.]
Remember when commercials during a TV show meant it was time for a bathroom break, and now you find yourself disgusted by advertising in general? That was probably before you spent a lot of time on the internet. Human society is fast approaching a moment of reckoning brought about by the rise of algorithmic social media platforms and the monetization of those same platforms by their operators. I have an unpopular opinion; hear me out (as if I didn't spoil it in the title of the post).
I'll say it again: At your core, you don't hate ads or ad-supported content. You may kick and scream and insist that yes, ads are the worst thing that has ever happened to civilization and we would just be better off without them ever. You may feel compelled to tell me in no uncertain terms just how wrong I am. That's your opinion and you're entitled to it, but the kicker is this:
Internet advertising has given immense power and utility to the advertiser by way of giving even more power to the platform. Think of it this way: the distribution platform has the capability and incentive to know as much as possible about who you are, what you like, and what your spending behavior might be based on those first two sets of data and your activity. The platform holding all this data about you doesn't necessarily have to give it to anyone else, either — it can stay in-house and still serve their business. As the distribution platforms grow larger and consolidate the market, the more data about you and your activity they can ultimately get.
DATA IS POWERFUL, AND UNDERSTANDING IT GIVES YOU SUPERPOWERS.
Data is a funny thing, and data science is probably more accurately referred to as art in some cases. If you take the time to understand seemingly unrelated data sets and correlate points within them, you have the ingredients for more data. This new data usually isn't nonsense — frequently, it's valuable. I'll give you a simple, yet very real, example: Imagine you're a sales manager at a small business and you want to track the business activity of your sales staff. You create a simple spreadsheet for them to enter very basic data about their work: the name of the client, whether the client is new, the date your salesperson pitched a job to that client, how much money the proposal was for, whether the deal closed, the date you got an answer from the client, and if the deal closed, how much money the client ended up paying you for it. This is all super basic stuff, and you don't even need a fancy CRM for it.
Sure, you can use this data as-is and create all sorts of textual and visual reports to give you and your staff a better look at their performance. You can even perform some basic statistical analysis to get an average of deals closed per month, for example. That's another data point, but it's really not new data, it's just related to what we already have.
If you've ever worked on a Sudoku puzzle, you know that it's frequently possible to place a number on a block by looking at where that number is (or is not) on adjacent blocks. A similar concept can also be applied to data, if you understand the data. Let's go back to the sales spreadsheet example, where you have a basic set of data about sales pitches. From this data, if you add a constraint or two, you can get multiple additional data points not given. One of the simplest ones is a close ratio — count the number of deals your salesperson closed and divide it by the total number of proposals. That's useful, right? If you take the sum value of proposals and filter them by the ones that didn't close, you suddenly have another less obvious data point: unrealized revenue. Getting more in the weeds, you can find the proposals where the final cost was different from the proposed cost and call that one your close delta, for example.
If you're not seeing the picture yet, consider the fact that these data points you just pulled out of seemingly nowhere can be used to create data of their own and start giving you some real insight about your staff's performance. Do you have someone with a lot of unrealized revenue but an otherwise high close ratio? It's quite possible that they're good at closing deals but may need some coaching on the pricing (or your business needs to evaluate its pricing). Do you see an average to mediocre close ratio and a close delta dipping negative? Perhaps this person has potential and could very well get a better ratio and neutral delta with some quality guidance over the course of a few one-on-one meetings. You get all this from data that doesn't seem related on its surface.
DATA SCIENCE HAS MASSIVE IMPLICATIONS ON INTERNET ADVERTISING.
Back to advertising. The internet ad distribution platforms have people who are paid very well to do what I just explained to you, using a lot more data, every day. If you were impressed by how granular the insights were by the extremely limited data I used in my example a moment ago, imagine what someone could extrapolate with a larger data set and even when comparing data from different people. You've seen this in action, and I guarantee you've been creeped out by it. If, for example, you meet a small group of friends at a coffee shop and you talk about a specific pair of running shoes you saw at Target the day before, there's a chance that you'll go home later that night and start seeing ads for those shoes on Facebook. You never looked at the shoes in any app on your phone or on any web site, you only talked about it... and now you're seeing ads for it. This has to mean your phone is listening to you in the background all the time and using that to sell you advertising. There's no doubt about it and you're 100% convinced of this, right?
Your phone is (most likely) not listening to you. I'm not a paid shill for surveillance capitalism or anything, because if I was, I wouldn't make all this effort telling you about why it's bad. This isn't because your phone is not capable of doing that — and trust me, the microphone on a modern smartphone is of fantastic quality — but because it doesn't have to. Capturing, uploading, and processing speech reliably at a massive scale is horribly expensive and inefficient. If you're suddenly seeing ads about a product you only talked about recently, and you're certain you never searched or looked at anything online about that product, it is a near certainty that someone around you did. The ad platforms understand the data they're collecting. If a group of devices were in close proximity with reduced activity during a specific time frame, it's likely that the owners of those devices were engaged in an activity together. Based on the activity of one of those devices either just before or just after that time frame, the platform can draw connections and generate actionable data about all of them. That is a lot of power, and it's creepy as hell.
THE INTERNET'S WAY OF TARGETING YOUR AUDIENCE
Now that we've established just how deep you can go with data that's fairly innocuous on its own, consider the fact that ad platforms just give their customers (the advertisers) the ability to query this data and execute an ad buy targeting its result set. It's like that episode of the US version of The Office where a phone technician comes in, shows Michael how to use the intercom function, and just ... leaves. To use an example bordering on the absurd, the number of white women aged 30 to 35 years who have a Bichon Frisé they go skydiving with every Tuesday can be counted on one hand... probably using the whole hand as the counting object, not the fingers, but I digress. If a company has developed an absolute must-have, life-changing harness for thirtysomething white women who go skydiving with their Bichon Frisé every Tuesday, they could use the platform's tools to buy an ad that will only be shown to those people. (Okay, that one person. Her name is probably Claire.)
My funny example is fairly innocent, but it's easy to see where knowledge of enough seemingly unrelated data points that aren't super intrusive on their own can paint a picture that implies access to very intrusive data. While the stakes aren't very high when it comes to skydiving with dogs, that's just the example I came up with. You can choose your own adventure coming up with other not-so-harmless scenarios.
YOUR FEELINGS ARE REAL AND VALID, BUT THE CIRCUMSTANCES ARE NOT
All this setup leads to the crux of my argument: advertising in general, and your opinion of it, has been poisoned by how it's done on the internet. There's an adage you've probably heard by now stating that using a free product supported by ads makes you the product. I believe that argument only works in the context of the granular targeting enabled by modern internet ad distribution platforms. People in general don't have a problem with the concept of sponsors bankrolling the creation and distribution of content they're interested in, and people in general want the people creating this stuff to get paid for their work.
Let's consider Claire and her dog. The fact the harness company can go to an internet ad platform and plug in the parameters to slice and dice the entire audience down to Claire does one thing: it substantially lowers the effort required to convert the target audience and as a result, the ad content suffers. I'm sure you've seen every oddly specific ad for a T-shirt being hawked on Facebook.
PASS THE CHICKEN WINGS AND BEAN DIP, LET'S WATCH COMMERCIALS
It hasn't always been this way. One of the last vestiges of real creative and memorable advertising content still exists, and it's during a championship gridiron football match the NFL would really prefer I pay them money to mention by name. These ads are so well made, some people literally tune into this football game to watch them. This is the other extreme from Claire and not a great example, but there's a compromise here: the quality and tolerability of advertising content is inversely proportional to the hit rate of audience targeting. With Claire browsing the internet, the advertiser can get close to a 100% hit rate. They could probably just show her a picture of the harness and she'll buy it. As the number of misses in targeting goes up, more creative effort is required to convert the audience and the ad content just gets better.
My day job is in broadcasting: real, traditional broadcast media. It may seem a bit disingenuous for me to talk about this given advertising revenue literally pays my salary, but if there's anyone well positioned in the world between the internet and the media industry, it's probably me. Traditional broadcast media, and even print media to a lesser extent, has survived for decades with a business model of selling ad time or ad space based on aggregate, non-identifiable profiling of their existing audience as a whole. Both sides do extensive market research and surveying to craft their product and determine generalized information about who the audience tends to be. Where there's an acceptable level of overlap, advertisers buy. No one needs to know what apps are installed on your phone, what web sites you visited last week, or what exactly you bought when you went grocery shopping yesterday. By necessity, there is some “miss” in the targeting, which puts the responsibility back on the advertiser to plant and grow the seed about whatever product they're talking about. I firmly believe the traditional media advertising business model is the more ethical choice.
TURNING THIS SHIP AROUND
What needs to change? First of all, advertisers and consumers alike are going to need to learn to tolerate some “miss” in the ad landscape — probably more on the advertiser side than anything. Consumers get the extreme version when turning off ad targeting on internet platforms. When it says “your ads may be less relevant”, it's not kidding. You get some really off-the-wall stuff. If advertisers can be okay with a lower hit rate and actually put in the creative work to convert the audience, the ads that aren't so relevant are less grating. Internet ad platforms also need to go on a data diet so savvy advertisers aren't given the keys to the kingdom, or better yet, spin off market research on behalf of publishers to get out of the business of collecting and maintaining the data on their own. The ideal situation would be advertisers dealing directly with the publishers. This could be either the evolution of today's ad distribution platforms where ad sales are outsourced or the ad platform is compartmentalized and sold as a service to the publisher, similar to a collaboration suite like Microsoft 365 or Google Workspace.
Internet content publishers also need to not be afraid of allowing consumers to opt out of the advertising ecosystem and provide revenue directly. On the other side of this, consumers need to understand the option and its implications. I don't get why people are so averse to subscribing to things on the internet. Meta is being forced to provide a subscription tier by European regulators, but in my opinion, this scenario just makes sense to roll out worldwide. I would pay for Facebook if it meant no more ads and a less aggressive algorithm. The algorithmic feed is incentivized to fuel outrage, which drives engagement. This engagement activity just provides more data about you which can be used to target ads, and the cycle repeats.
I'm just grasping at straws here, but this is a worthy conversation to have as people are seeing the consequences of internet advertising and trying to get away from it. When clients are asking your radio station exactly how many impressions a commercial will get, you know the entire pool has been poisoned by the internet. As a one-way broadcast medium, there is no way to accurately measure how many people heard a spot. We can rely on market research and ratings survey data to make an educated guess, but couldn't say for sure if 462,137 people are listening to our station at 1:21 PM on a Tuesday...
...but you can reasonably assume Claire and her Bichon Frisé aren't. They're in an airplane getting ready for their jump with this brand-new harness.