Don’t Get High on Your Own Supply

Don’t Get High on Your Own Supply

Over-emphasis on data can lead you down dangerous paths

Since Google arrived on the scene making the kind of money from data that was never seen before, data became the center of technology. Facebook’s entire business model has been built around data and newer ways of monetizing it. Data can do no wrong. Data is the ultimate power.

This worship of data quite naturally led to glamorization of the role of data science and everybody is building predictive models in Silicon Valley. Consequently, Data scientist jobs also pay more than software engineers according to Indeed Stories of how having data led to incredible impact are dime a dozen.

What about the flip-side of having too much data? Is there one?

Voter sentiment and social media

2016 elections changed America irrecoverably. Facebook with its wealth of data about people’s sentiment could predict that Trump could be swaying the crowds. Mark Zuckerberg took a bet on Trump based on this data, allowed the Russian bots to overtake his platform and decided to look the other way while Russians toyed with our democracy.

Trump’s victory triggered a cycle of right wing content to be created and consumed more on Facebook. Increasing numbers of Trump supporters were on Facebook. People were posting hate content: engagement metrics soared. But this created an echo chamber on Facebook. By 2020, the only voices that could’ve been heard on Facebook were the people posting daily controversy from the white house and if you did any sentiment analysis on that data, you were bound to think that Trump would win again. At this point this is selection bias: Facebook only has the audience that shares loud and hateful content, and if you believe Facebook has the world’s data, its easy to believe thats where the world is at, and the idea to look beyond this data would not even occur to you.

Eventually when Trump lost the election in 2020, Mark Zuckerberg has found himself at the wrong side of the White House having cost the Democrats the 2016 election. As a corporation as big and international as Facebook is being on the wrong side of DC can be very damaging and in addition to Tiktok stealing Ad revenue from Facebook, losing political pull might also be a big factor in struggles of Facebook.

The HGTV addiction

When Zillow was new, it was an amazing experience that changed how people expected to look for homes. For an app, once you hit a certain user base, the key focus that drives valuations is engagement metrics. Zillow needed to have people spend more time on their app. To get more people to spend time, you need inventory. To get more inventory more people need to be willing to sell. For more people to be willing to sell their homes, they should see significant upside. Welcome Zestimate!

Home shoppers spent more than 5 billion minutes on Zillow via our apps alone. This averages roughly 16 minutes per person in the U.S., or the total equivalent of 9,500 years. Home shoppers viewed homes more than 2.5 billion times, up 150 percent year over year. That averages to nearly 80 homes per second.

At one point in 2018, my one bedroom one bath condo in downtown San Francisco had a zestimate of over a million, in 2022 my condo sat on the market for months without being able to fetch $700K. That’s how overestimated Zestimate has been. But most users won’t realize that the Zestimate is a manipulation tactic to have more inventory to fuel user engagement.

That was all hunky dory, until Zillow thought one day: Wait a minute, if we can control the prices of the real estate by controlling buyer and seller expectations and inventory, why don’t we buy and flip homes and get in on this margin? The investment team that was buying houses probably didn’t cross paths with the Zestimate team that was trying to drive engagement metrics. Or maybe by this time Zillow has forgotten the reason they created Zestimate in the first place.

Using data to manipulate customers, ultimately had them fall victim to their own tactic overestimating the amount of money they could make flipping homes. The black swan event of the pandemic suddenly created a materials shortage and increased contractor expense. Zillow lost more than $881 million trying to flip homes and laid off more than a quarter of their workforce - in the best real estate market in history.

Before the data arrives

Before the Pandemic, there was never data to prove either way of large companies could operate entirely remotely. So they kept hiring and kept building expensive offices because that is what everyone did and that is what all the data suggested needs to be done. But after 2+ years of record growth in many businesses with entire workforce working remotely: you have data to prove remote working works better in a lot of cases. Prior to the pandemIc there was probably just occasional surveys where people expressed wanting to work from home once in a while, but now the survey results are making it hard to justify the real estate expense of having office facilities.

Putin seems to be consuming his own Propaganda

A great example of self created misinformation bubble is Putin during the invasion of Ukraine.

The Late Show with Stephen Colbert - Putin Is A Victim Of His Own Propaganda March 31, 2022

Russia recognized that Facebook and other social media platforms were engineered to pick up on polarizing news. They strategized that spreading hate news could challenge the democracy in America. Russian propaganda has been creating fake news tirelessly trying to bid Americans against each other, but within Russia they have been telling Russians how good Putin is showing the divide in American society as a failure of democracy. I guess it was a good strategy and one that almost worked, until Trump lost the election and democracy won, proving that Americans, though divided, were not as much in the clasp of Russian propaganda as Putin might have hoped.

Being data driven is not a bad thing.

It’s important to recognize that decisions can be made with common sense and in absence of data. Looking beyond data is as important as looking into data. Black swan events occur that change the world and all the data becomes incorrect overnight. Or new data invalidates old data.

Another problem with being overly data dependent is that you are limited to the data that is there and that’s why you can only do what has already been done. Data puts you in a box.

Visionaries are able to imagine scenarios where current beliefs are challenged. Steve jobs was able to imagine removing the keyboard from phones, which had never happened before and 100% of the phones came with a keypad. Would you have asked him if he had data to prove that this would work?

References