Volume 87: Algorithm? Algo-wrong.
1. Algorithm? Algo-wrong.
tl;dr: Algorithmic failure causes Zillow to lose big.
Zillow is an incredible success story. A true innovator in what was the sleepy category of residential real estate listings. Now, 15 years later, it’s hard to overestimate the impact it’s had on the US housing market since its formation in 2006 by Expedia alums.
By 2018, though, this was a business beginning to look a little stale, so it decided to get in on the iBuying game. iBuying, in case you don’t know, is a business model where institutional buyers use data science and algorithmic pricing to identify homes to buy, undertake modest improvements upon, and then flip at a profit. For Zillow, this looked like a no-brainer. It had access to the best data in the business, its own captive listing platform to sell on, and as much cheap capital as it needed to buy up all the homes it wanted. And buy them it did, expanding their approach in April of this year, before the whole thing spectacularly collapsed in embarrassment last week.
Amid one of the most significant bull runs in residential real-estate history, the Zillow algorithm got pricing so wrong that it now owns over 7,000 homes it acknowledges will have to be sold at a loss. This led them to shutter the iBuying business, reduce headcount by 25%, and watch as the stock price plunged.
I find it fascinating that while this is a spectacularly public example of algorithmic failure, it’s a risk that surrounds us daily. For years people have complained about software that sets inhumane shift schedules for hourly workers or HR software that reads resumes and rejects millions of qualified candidates, or adtech algorithms that operate as undecipherable black boxes, or racist law enforcement algorithms, or the stock market itself, where trading behavior is increasingly driven by machine learning algorithms that set the incentives for managers of publicly traded corporations, but nobody understands what these incentives are based on. (Aside from a Skynet quality to this reality, there’s also a pocket industry springing up in writing financial disclosures specifically to be read by trading algorithms rather than people.)
Now, don’t get me wrong, I’m not some Luddite. I fully understand that algorithms and machine learning enable all sorts of businesses and value creation opportunities that haven’t been possible before. It’s just our blind belief that these algorithms work and work in our best interests that is concerning.
I was reading a Twitter commentary about Facebook the other day. Someone commented that it wasn’t possible for a product to be sociopathic, which stopped me in my tracks because that’s exactly what the Facebook newsfeed is. That the Newsfeed algorithm amplifies hate is well known, but what is less well known is that it also algorithmically suppresses counterpoints because they add friction to the initial post, which reduces engagement. The net result? Mis and disinformation traveling unchallenged around the world. If this were done by human hand, we’d refer to it as sociopathic behavior, so why can’t we label it similarly when done by an algorithm?
Anyway, while the Zillow failure may be one of the first, it certainly won’t be the last time we hear about a corporation making spectacular losses due to algorithmic failure. As more corporate decisions are made by software, then by definition, the risk of that software making a bad or even catastrophic decision increases.
Increase far enough, and we might need to look back at the history books for a solution. Independently auditing the financial statements of publicly traded corporations emerged as a response to managers lying about what was in these statements, which created a material risk to investor capital. If black-box algorithms are found to threaten investor capital in a similarly systematic way, don’t be surprised if the independent auditing of such algorithms is eventually forced upon corporations too.
2. Greater scrutiny in the digital advertising ecosystem?
tl;dr: Rising costs, greater scrutiny?
Unlike many who like to opine on brand-related matters, I’m not an ad guy. My background being on the strategy consulting and design side of the ledger. However, as an outside observer, it’s fascinating to watch what’s happening in the advertising ecosystem, especially as it relates to digital media.
The 2008 financial crisis sparked a bull run in the shift to digital advertising that especially benefitted the likes of Google and Facebook. At the time, a deep economic recession meant corporations were looking for ways to maximize the efficiency of their media spend, and shifting it to digital was both trackable and cheap (relative to alternatives), and coincided with a shift in consumer attention to screens, mainly those in our pockets. As a result, more than one startup unicorn was built off the back of cheaply purchased remnant inventory during the early, heady days of programmatic buying. (Ironically, the winners today might be those who realize that the underpriced inventory increasingly looks to be in non-digital, so-called traditional, media)
Fast forward to 2020, and there was another spike in the curve as marketers looked at locked-down customers and budgets and shifted billions more dollars into digital media, finally reversing the eyeballs/media gap Mary Meeker has talked about for over a decade now.
And, while the news media has become obsessed with the price of bacon in recent months, a less observed side-effect of billions of extra dollars pouring into digital media has been that the ads are getting more a lot more expensive too. While it’s still trackable, it’s much harder to say digital ads are cheap these days because even with the Apple-induced reduction in the ability to sell surveillance advertising at a premium, the cost of digital ads is most definitely going up.
And when costs go up, typically so does scrutiny. And if that’s the case, where’s scrutiny most likely to fall? Here are three areas that feel well overdue for a transformative impact:
Fraud
Experts believe the cost of digital ad-fraud will rise to $100bn globally by the end of 2023. Estimates are that as much as 50% of your total media spend is being viewed and clicked on by bots rather than real people. Ironically, the more narrowly you attempt to target specific audiences, the more likely you will be the victim of this fraud because the bots make more money pretending they’re the customers you really want. Now, while it’s been observed that the price of fraud has already been factored in, that doesn’t consider the rising cost of digital ads. With that in mind, a full-scale assault on ad fraud will be more, rather than less likely in the future. Just don’t look for it from within the current system because crappy incentives mean too many advertising people are making too much money by turning a blind eye.
For more on ad fraud, check out Dr. Augustine Fou.Transparency
Digital advertising today is an engineer’s fantasy playground. The sheer complexity and scale of what they’ve built over the past ten years is astonishing. Google alone serves 11 billion ads daily. Unfortunately, however, the side effect of all of this complexity is that the opaqueness of the digital supply chain has become a haven for intermediaries, black-box algorithms, and a myriad of opportunities to rip you off. Transparency, this is not. In a major piece of analysis last year, the forensic accountants over at PwC found that roughly 50% of your total media spend is probably consumed by ad-tech intermediaries (meaning 50% of your spend never made it to a consumer, or bot, at all) and that 15% of that spend was completely untrackable - meaning the money simply disappeared into thin air. In addition to a lack of transparency, sub-standard engineering solutions to human challenges, mean important journalistic topics get defunded, while media dollars are unwittingly funneled into some of the worst hate sites on the Internet.
For more on how the transparency-free world of digital advertising leads to the unwitting funding of hate, Check My Ads.Attribution
For the longest time, Google was showered with wealth it didn’t deserve because of the neat jazz hands of “last-click attribution.” Last click just means overweighting the last click the consumer makes before going to a website to purchase, which, surprise, surprise, is often from an AdWords ad on the Google search page. We’ve since figured out the obvious: last-click attribution is a crappy metric because there are often more important things that happen first, but we’re still far from figuring this stuff out correctly. We still hear horror stories of brands shifting marketing spend heavily to digital ads focused on demand gen/activation only to find sales slide and discounting increase (a marketing death-spiral of a combination). Some smart economists are doing good work here, but if digital ad costs continue to increase, look for a hard shift toward econometric models focused on contribution to cashflow and a shift away from the rose-colored attribution dashboards the likes of Google and Facebook peddle.
For more on the economics of advertising, check out Grace Kite over at Magic Numbers.
It’ll be particularly interesting to see how this all shakes out, not just for the advertising technology industrial complex but also for the advertising holding companies. Last week, Omnicom announced that aged CEO John Wren is lining up slightly less aged doppelganger, Daryl Simm as his successor. (I know they say advertising discriminates against the over 40’s, but it sure looks like an old white man’s game from the top). Anyway, it’s zero coincidence that the anointed one comes from a media background. After all, this is where the likes of Omnicom has been making all its money in recent years, and if nothing else, he’ll know where the bodies are buried.
But, assuming greater scrutiny does fall on digital media, it will crimp this cash cow the advertising holding co’s have become dependent on. And if that happens, look out for a raining domino effect as some or all of them start jettisoning underperforming assets - likely the creative agency networks that operate on single-digit margins, which will almost certainly be tempting targets for private equity companies out bargain shopping when they’re being dumped at firesale prices.
3. And then there were three. GE’s, that is.
tl;dr: Three GE’s marks the end of an era.
My, how the mighty have fallen.
GE has long been one of America’s most storied corporations, founded in the 1800s by Thomas Edison and famously supercharged by Jack Welch to the point that it become the world’s largest corporation by market capitalization upon his retirement in 2001. A pedestal it then fell spectacularly from in 2008 due to its exposure to the sub-prime lending markets. After much soul-searching, it slimmed down into a much smaller company by dumping most of its finance businesses and is now headed toward life as three even smaller companies with the announcement this week of a proposed split into a healthcare company, a power and energy company, and an aviation company. Phew. (Great news for branding agencies, by the way. Two new brands to create and a third to revitalize. Every new business team in the country is currently going ga-ga trying to figure out who they know over at GE.)
Anyway, GE is personal to me. Immediately following the Welch years, at the beginning of Jeff Immelt’s time as CEO, we won a global brand revitalization job for it. It was the first major pitch I’d ever done and was the reason I ultimately ended up moving to the US.
Looking back in hindsight, all of GE’s subsequent travails were there to see when we worked with them. And, while history has not looked kindly on Jeff Immelt’s tenure, I’d suggest that Jack Welch left the company on a precipice that meant he’d had to have been a super-CEO to thread the needle toward success, were it even possible.
Under “Neutron Jack,” GE famously had two mantras that had pretty dire consequences. The first was to let go of the bottom ten percent of performers every year, with the idea that this would leave you with only the best people and eliminate the sleepy bureaucracy he’d inherited. And to a certain extent, this was true. Some of the most impressively cogent, lear, and driven business leaders I’ve met worked for GE at that time. However, it by no means created across-the-board success. One of the negative impacts of the “cut 10%” mentality was that it went on for too many years. After the fat had been cut, it wasn’t just the underperformers that were eliminated, but the muscle of the business as entire business functions were systematically excised, like marketing. Because GE had so few marketers and essentially nonexistent budgets, they had all sorts of issues with basic things like buyers in major categories, like healthcare, not knowing that GE was even in that business.
The second mantra under Welch was the statement that you should be “no 1 or 2 in your market, or get out.” Now, while this sounded good on investor calls and was intended to push the business toward markets where it could win at scale, what it achieved was the opposite. Instead of using its considerable scale and scope as a business advantage, GE atomized itself into smaller and smaller parts to meet the “1 or 2 definition.” You see, the fastest way to be no1 in any market is to define that market very, very narrowly. This is why you’d meet people with job titles like “Joe Bloggs, President, GE 5mw Wind Turbines, Northern California.” Why? Because they were no1 in sales of 5-megawatt wind turbines in Northern California. Sigh.
At this time, I remember becoming somewhat skeptical about branding books. Here I was trying to help solve this crazy atomization problem, exacerbated by GE having systematically dismantled its marketing function, while at the same time reading a glowing write-up of GE as the paragon of a master-branded architecture (I think by Kapferer, or maybe Aaker, I can’t remember). I distinctly remember that what I saw and what they were writing about were two entirely different things.
However, these were tiny issues relative to the 800lb gorilla that wasn’t just in the room but lurked over everything, called GE Capital. Many years earlier, GE had established a captive finance arm that meant it could offer financing on the sale of its own products. Under Welch, this finance division was supercharged into a broad-scale lender through hundreds of acquisitions to the point that it came to represent 50% of total revenues. As a result, there wasn’t an aspect of the riskiest areas of lending that GE wasn’t involved with globally. You name it if it involved lending out money for high returns, GE was doing it. Why? Because it had become addicted to the superprofits generated by credit-rating arbitrage.
Very simply, if you were a sub-prime lender making risky loans, then your credit rating reflected this, and your cost of borrowing was high to make up for the risk you represented. However, if GE were to buy you, its AAA credit rating meant the cost of borrowing dropped precipitously, allowing GE to pocket the difference as pure profit. The problem, of course, was that the more of these acquisitions they made, the more risk was being absorbed into the business and the less accurate that AAA rating really was, until…Boom. The 2008 financial crisis rained the whole thing down around their heads.
Now, at the time, we had no clue this was going to be what ultimately crushed the company, especially not as a wet behind the ears 20-something-year-old branding consultant. But I remember us building a whole ring-fenced architecture around some of the less salubrious consumer lending businesses. Like in Europe, where a GE-owned business had a name that was literally synonymous with the term “loan-shark.”
As for the work, well, I’m pretty much convinced that what we did remains the world’s most complex brand architecture consolidation. We dealt with over 400 non-integrated acquisitions and thousands of things branded GE, trying to bring them all into some semblance of order at scale. I haven’t worked with them directly for almost 20 years, so I’m not sure how much of that work survived to this day, except that I often still see the baby blue and the typeface we created for them. For type geeks, GE Inspira (no, I didn’t name it) is a custom font loosely based on VAG and designed to reflect the curves inherent in the GE monogram. I’m a big fan of the typeface because it’s lasted almost twenty years and still feels distinctive, modern, and instantly recognizable as GE. A rare thing in these times of custom designing Helvetica over and over and over again.