I said I wouldn’t do it, but I’m going to be the adtech bro that writes a response to this article in the NYT that slams adtech titled “If It’s Advertised to You Online, You Probably Shouldn’t Buy It. Here’s Why.” by Julian Angwin that made it’s way around Twitter recently. I am sure I will get heat about this, because advertising creepy and bad (and it is), but there were so many factual errors and misrepresentations that I feel compelled.
Adtech deserves to be criticized by journalists. Hell it *needs* to be criticized by journalists. It’s an industry of mercenaries and external pressure is a disinfectant to the sludge that builds up from shady practices. If you want meaningful change though you have to come at it from a place of being factually correct, and do so in good faith otherwise folks will simply throw the opinion out the window. I don’t know if this was written as a serious critique with the intent to reform or was basically fan service to privacy fetishist crowd or what. I’ll ignore that Julia’s salary is paid by ads, the distribution of the article happened on Twitter and other ad supported platforms. Media companies *need* ads. Netlfix introducing an ad tier after how many years of just subscription should help end the debate that ads are optional for most media organizations. It’s the proven working business model and if we want more media, and we want well funded media, that is going to involve ads on some level so long as humans are involved.
For me, I’m largely neutral about ads as an industry. It’s a useful tool for lots of businesses and like it or not it’s basically the oil of the internet. It keeps the money flowing and the lights on whether we all like it or not. It’s also a tool that is misused and abused to the point of real world harm on a fairly regular basis. I am not claiming it’s perfect or even good, in fact parts are deeply rotten.
So let’s talk about what I am not saying with this article. I am not defending the status quo of the adtech industry as it stands. I am not defending spyware like data collection. I am not defending any law breakers. I am not taking a political stand. This is industry wonkery and nothing more than that. I am trying to bring a real politik energy to this debate from someone who works in the industry but isn’t a pro advertising zealot. I think advertising is a thing that is needed for businesses on some level. I think that there’s a lot of ways it can be used to empower smaller local businesses over very large ones to even the playing field. I think that most people, even the folks doing creepy borderline things in adtech mostly aren’t evil sitting atop their scrooge mcduck fortunes with their monocles and cigars guffawing at the poors. I think when things happen that have moral hazard it’s a series of small compromises that lead to harmful outcomes. There are obviously bad actors and I’m not talking about them today.
To truly reform and bring change you have to start from a place of understanding why things are the way they are with an open mind and nonjudgement. Only then can you start to make a series of similarly small rules that lead to outcomes that are mutually beneficial for everyone involved. Businesses can grow, media can get funded, technologists can innovate, and users can get useful ads while having their privacy mostly respected.
Julia opens the article talking about some brand of Razor’s I’ve never heard of that apparently advertises themselves as anti woke. I have no idea what that means for a brand in this context but it’s setup to obviously draw some sort of Trumpian connection to the advertising industry and upset the sensibilities of the average New York Time’s reader even though most folks I know in said industry are overwhelmingly liberal.
She claims that “But online, many ads are sold based on the many details that advertisers have gleaned about your behavior and interests from your online activity.”
Which is my first point of contention. No, individual advertisers almost never have access to that data. There is no way to meaningfully say in most contexts say Person did X. Or Person did Y online in terms of adtech data being passed. Certainly not to advertisers. Even getting into things like bid logs and win logs, the actual meat which does contain PII (personally identifying information) it shares data like device type and geo location, IP etc. Stuff that can be used to identify you, but is really like 10 pieces of a 1000 piece puzzle. It’s a sliver of half information that can be used to assemble a mosaic but has to be meticulously reconstructed to have any meaningful insights into. Bad actors do this, but it requires a ton of data, patience, and is ultimately throwing darts at a wall while blindfolded. Tech companies like Alphabet or Meta certainly have this info but theoretically it’s encrypted in databases and anonymized in a way that a Meta employee similarly cannot just access the “Person 34231” file and see all the weird things I google late at night.
The other thing, and I’m going to let you in on is an dirty industry secret. Are you ready?
Half the data is pure garbage or totally wrong anyways.
So even if you’re assembling the puzzle half the data is junk and you don’t even know which half further making the idea that advertisers have some folder about you somewhere with your darkest secrets even less likely.
She follows up with the claim that is mostly true but contains an annoying persistent lie which is “Tech firms track nearly every click from website to website, develop detailed profiles of your interests and desires and make that data available to advertisers. That’s why you get those creepy ads in your Instagram feed or on websites that seem to know what you were just talking about.”
The lie being your phones listen to you. They simply do not. The truth might sting but honestly you’re just not that special. Some mix of highly tuned probabilistic models and the Baader-Meinhof Phenomenon are responsible for this. I’m in deep, I buy the creepy data every day. My platform is using the creepy data right now. We’re working to get away from this, but the truth is that we just aren’t there yet. When we buy that data though you don’t buy a file on an individual, you buy an audience like “Ford Owners” that has a giant key of anonymized identifiers that links a bid in a firehose of bidstream data to someone in an audiences of millions. The “your phone listened to you talk about x brand” audience targeting option doesn’t exist. Adtech companies do track you in weird ways. This is not one of them.
The next claim I would consider absolutely false is “Already, we know that web tracking has decimated publishers.” which is also simply not true.
Other people much smarter than me such as Ari Paparo and Eric Seufert have pushed back on this in a more elegant way than I am about to briefly touch on, but I do want to push back on it. Technology is disruptive. I don’t care if you sold horses at the advent of the car or you were a whaler when electricity became a thing but it’s a natural consequence that is good. I’m not entertaining a smash the looms argument or socialism now nonsense. When media was both the content and a captive distribution channel whether via governments dolling out radio frequencies, needing the capital to buy a giant printer and hire paper boys, or build a television study it was a lot more profitable because there was less competitors. The internet has dramatically reduced the barrier to entry for creating media and because of that there’s more optionality. More optionality equals more competition and price parity. This can have many negative outcomes, but this is hardly the consequence of web tracking technology. It also has many positive outcomes as well such as elevating the voices of marginalized authors of various backgrounds and creating a robust competition in the media world. It’s been bad for the legacy gatekeepers of course who no longer have unquestioned narrative control accepted at face value. It’s a single facet or a larger trend and frankly the margins of media would have declined regardless even if it didn’t exist simply because of the staggering amount of new supply.
The actual paragraph that inspired me to actually write this entire novel was the following claim which is a fundamental misunderstanding of the technology powering the machine learning driven optimization of media. “Microtargeting has also enabled advertisers to discriminate in ways that are hard for regulators to catch. It is illegal, for example, for advertisers to use language in their ads suggesting that jobs, housing or credit opportunities are being offered to people of a certain race, gender or age or in other protected categories. But ad targeting means that advertisers can hide their preferences in the algorithm.”
It is 100% true advertisers can attempt to discriminate and it can and does happen. The idea though that an algorithm though can pick up racist vibes from a media buyer to automatically exclude someone of a protected class from seeing an ad for something like employment or housing opportunities is genuinely absurd. Machine learning optimization is inherently a black box on some level. You train it on a data set and that data set optimizes for an outcomes, usually based on some very banal goal like “target people more likely to fill out a form”. Those training sets can have biases of course, but again this is not something an individual advertiser controls for. In fact with Meta when you run ads in sensitive categories there literally is not demographic targeting even allowed. The machine learning driven model of media buying arguably reduces bias. In fact you could arguably be more discriminatory by the advertising model that Julia seems to hold in at least less hostile regard which buying directly from a publisher. If I’m only buying a print ad in some Yachting magazine or an equestrian magazine for a housing opportunity do you think I know the demographics I’m likely reaching? Yeah, I do too. The ye olde nostalgic media buying isn’t the panacea of anti-discrimination that it’s cracked up to be.
Julia goes on to lambast the brands that are apparently “allowed” to advertise claiming that “The products shown in targeted ads were, on average, roughly 10 percent more expensive than what users could find by searching online. And the products were more than twice as likely to be sold by lower-quality vendors, as measured by their Better Business Bureau ratings.“
My feelings on the BBB aside, I assume this means Julia also roams the aisle of Target on weekends shaking her head at brand name products versus generics on the shelves? Consumers wanting one thing versus the other thing is like the bedrock of marketing and has near nothing to do technology. In the days of yore I assume the New York Times rigorously tested every single product and service that bought an ad in its newspaper to make sure they only accepted the absolute best in each category correct? Of course platforms have a duty to keep obviously scam products from advertising, and tons of platforms like Meta have incorporated customer feedback into a quality score to even try and prevent this too, but the standard being set here is so high and outside the norm that I don’t think any advertiser ever had held themselves to this. I assume NYT has never run an ad for Kia because the performance is lower than that of a Ferrari correct?
My hyperbole aside this is what makes me sad about this entire line of thinking. Adtech has, despite its many many faults, empowered a massive crop of new small businesses, created by people that may have otherwise been marginalized because of who they are by the powers that be, and their downstream service providers and tech companies to be built. An absolutely massive amount of jobs we’re created, people jumped classes and income brackets on the back of what they built. Consumers got actual choices in the things they bought. This was a good thing. You no longer had to have a massive amount of cash to afford a national print campaign with the New York Times to advertise your business anymore. You didn’t have to hire some Madison Avenue Ad Agency to be able to take your shot. Access became dare I say more equitable, and I think this is one of the good things that has come out of this. Small businesses have the same seat at the table with Google as a fortune 500 company and can grow. Certainly hundreds of thousands of them, maybe even millions, did. No longer constrained by nepotism, or who they knew, or if they already came from wealth and could afford to start a company in the first place at all.
The final claim I take issue with is the following. “Jeremy’s Razors doesn’t need to know your family structure, your favorite sport or the name of your favorite singer. Jeremy’s could simply place its anti-woke ads near anti-woke content.”
Jeremy’s Razor’s (Unrelated to me I spell my name with an “O” and have never heard of this brand before this article) doesn’t have this information. It lives in large anonymized data sets entirely unavailable to Jeremy’s Razor’s known only to Meta’s machine learning. They don’t know who you are, they don’t decide what content their ads is placed next to. No human does.
Ultimately adtech is very broken. I could write entire books on ways that adtech needs to be, and could be fixed. This article fundamentally misses the mark and ultimately confuses the debate by providing, dare I say, misinformation about the actual nature of the mechanical ways these things work. Likely this was done to jack up the fear of the average consumer to push them towards Julia’s personal opinion. To be fair it is an opinion piece, but I digress. As Julia points out though it’s a $540 billion dollar industry, an industry that isn’t just going to pack it up. If we want meaningful positive change, and I do, that moves us away from the creepy targeting that does happen, an admirable goal, then we need to at least be realistic about the actual faults that exist and not just make up more dramatic sounding anecdotes about anti woke razors looking through your bedroom window at night writing down what posters of pop stars they see to apparently target their ads better and have a good faith conversation about realistic ways where we can smooth the edges on this industry and work to evolve things over time towards a place where advertising as a business model works for everyone, including consumers, in a more privacy focused way without sacrificing the good that opening up the ability for anyone to harness ads as a tool to grow their business. It’s possible, and we just have to be willing to listen and work together rather than create an unnecessarily hostile conversation based on truth not lies. I’m going to work towards that future, I hope you join me.