EUGENE – Google’s pseudonymous parent company Alphabet Inc. announced late last week that it would discontinue the use of controversial in-email advertisements. For over a decade, the company scanned the content of users’ emails for keywords and then displayed relevant ads right in Gmail. Despite ongoing protest from consumer and digital rights activists, as well as a handful of wiretapping lawsuits, email scanning only recently came to a halt because ‘[paid] business customers were confused by the distinction and its privacy implications…’
For years the trade-off for free Email was that Google would scanned message content for key phrases to better target you with ads across its other products (like Search, and even in Gmail). Now, because corporate clients voiced confusion over the differences between paid and free Terms of Service, Google’s in-mail ad units are a thing of the past.
– sent from my Chromebook
2004: Humble Beginnings
A June 2004 write-up (with a hilariously tin-foil hat name), “The Internet Ad You Are About to See Has Already Read Your E-Mail” explains it this way:
When Google announced in April that it would test a free Web-based e-mail service that offered users vastly more storage space than its rivals, it introduced one twist — the ads that users see when they read their mail would be related to the subject mentioned in the message.
So if your friend sends you a message about his vacation in Florida, you will see text ads for beach resorts to the right of the message. When your mother writes about her new digital camera, photography ads appear. – New York Times
Certain ad topics were off-limits – dating, sex, drugs, guns, etc. despite being some of the “most lucrative categories for other Web-based [marketing].
With questionable success, Google also tried its hand at early sentiment analysis – gauging whether a user is excited or frustrated based on word choice and context clues. “So if your mother complains that her digital camera is a dud, the recipient is not likely to see a camera ad on that message.”
How It Work(s/ed)
Back in 2004, the Gmail product was exclusively for single user accounts (as opposed to full companies with their own domains just managed by Google), and was only available via invite from current users.
Susan Wojcicki, then Google’s director of product management (and current CEO for the YouTubes) explained the technology behind the email ads
First, computers scan a message to determine what concepts are being discussed. They look at the words that are used, their frequency and where they are in the message, among other factors. The technology distills this data into key words that describe the content of message.
The second step involves comparing those keywords to Google’s database of advertisers, who have bid to have their ads placed next to messages containing certain key words. Google’s system considers both the relevancy of a given ad to a given e-mail message as well as how much the advertisers are willing to pay.
Worth noting as well – Google has also said that they scan the content of Gmail messages for signs of illegal activity – child pornography, malicious software and attempted hacking schemes. The success rate of these types of searches, however, is hard to come by.
One day I’ll get that darn blue check mark and maybe these giants will start taking my requests for inquiry seriously. Crawl that, Googlebot 😉
Missing the (Data) Point
As we’ve explored in other posts, data in isolation (length of rope, butler, observatory, body) on their own can be relatively innocuous. But when taken together in aggregate, data can be strung together to paint a broader picture (The butler killed Mr. Body, with a rope in the observatory).
When Google handles your email, they have access to two major types of data. Meta Data, which we’ve explored before as well as Body Content:
Meta Data attached to your email (who emailed whom, when, and about what subject) on its own may be fairly harmless, but coupled with other existing data points like subject matter and the purchasing behavior from in-app ads, over an extended period of time?
Body Content is literally the words you send in your email. From Dears and To Whom it May Concerns all the way down through the last link in your overly complicated signature. This is the stuff that privacy advocates are most concerned with – it’s like giving a 3rd party the freedom to read your letters en route, and then sell the info they find to build new billboards along your commute.
The implications for this type of access starts to add up really fast – ESPECIALLY if key info like context is absent.
Here’s a goofy but telling example:
How many times have you left a meeting and shot off a quick follow up email that two days later makes no sense, because you’re now lacking context?
Great – let’s hand that responsibility off to a system that isn’t sure what I meant when I sent: “Sure, can’t wait to kill a Sam Adams after work on Friday.”
Nothing’s Free On the Internet, Moore’s Law, and other Comm 101 Greatest Hits
Also worth noting – users received a whopping full gigabyte of cloud space for FREE to store their Gmail messages. That may not sound like much, but adjusted for inflation and Moore’s Law, a 2004 GB is roughly equivalent to all of the rack space in Menlo Park, California in 2017. (Not really but you get my point.).
Now that users have a full gig of space, they’ll ‘never’ need to delete an email again. Effectively giving Google a trove of meta and body copy to continually mine for more data with which they can better target you. Think about it: a user with 500 data points is way more likely to be honed in that a user with just 10, and as a result is way more valuable (see: Google’s own ROAS definition).
Let’s not forget that when get offered free stuff on the internet, it most likely means that data about your use is actually the product. Simply put, [Modern Media Company] makes way more money charging advertisers than they ever would charging users.
2004 Flashback Attack:
The basic idea: Google will offer you a gigabyte’s worth of email storage capacity — by one count, up to 500 times that offered by its competitors. But it also plans to scan the contents of your email messages in order to display advertisements relevant to your online conversations.
“Because many people wouldn’t have to delete email, [Google] could potentially search for communications a year or two, or 10 in the past, with ease,” –Gmail: What’s the Deal?, Electronic Frontier Foundation, 2004
Imagine what Google’s ad partners could do with 10+ years of search, email and chat history…
Scaling to Enterprise Clients
As Gmail’s popularity grew among early adopters, it became obvious that the web-mail service would be useful not just for individual ‘free’ users, but for corporate accounts as well.
Companies wanted to pay Google to manage their employee email in an easy to use, browser-based environment. It alleviated the nightmare of getting multiple systems to play nicely with one another, not to mention integration with Google Docs, Google Calendar and even GChat. (Sorry Googles Wave and Buzz, I haven’t forgotten you even if everyone else has).
To expand Email at scale to enterprise customers, Google had to make some compromises – chief among them: no longer scanning the email content of paid users in order to protect the privacy of corporate clients.
Let’s imagine for a moment a fictitious company, World Wide Wickets Co., that has been struggling with in-house email servers for 5+ years.
All of their data from everyday accounting to HR is sent through internal systems. To switch to Gmail for Businesses, World Wide Wickets would need to be guaranteed that their content not be scanned – especially if that content was sensitive or confidential for one reason or another.
With Great Market Share Comes Great Scrutiny
While scanning was eventually discontinued for paying customers, Google continued the practice for its free-level users for OVER A DECADE. This went on despite mounting protest from both privacy and consumer advocates.
In fact, Gmail’s history is peppered with lawsuits and ongoing criticism.
Despite this, the product still accounts for a whopping 20% of all email client use as of June 2017 (just behind ‘iPhone’ – which really isn’t an email client so much as it is a default app).
Electronic Communications Privacy Act
In 2010, Texas resident Keith Dunbar filed suit against Google, claiming that by scanning his email content, Google was in violation of the Electronic Communications Privacy Act of 1986.
Generally speaking, the ECPA “prohibits the unauthorized interception of wire, oral or electronic communications,” except in situations where a company has obtained prior consent. Whether accepting a Terms of Service agreement, like when a user signed up for Gmail, counted as such had yet to be determined in a court of law.
By 2014, the case expanded in scope and plaintiffs hoped for a class action certification:
The [ECPA] has a provision that allows companies that serve as facilitators of communications, like e-mail, to monitor the contents of them, but Congress stipulated that the provision was intended for quality-control purposes, like combating spam.
Plaintiffs in the case are arguing that making money off targeted ads doesn’t really constitute quality control, and that serving ads has nothing to do with sending e-mails, which is what a Gmail user has signed up with Google to do…
Google’s defense centered around implied consent:
In a section of the motion that was widely noted, Google also argued that non-Gmail users had no expectation of privacy when corresponding with Gmail users.
“Just as a sender of a letter to a business colleague cannot be surprised that the recipient’s assistant opens the letter, people who use Web-based e-mail today cannot be surprised if their communications are processed by the recipient’s” e-mail provider, the lawyers wrote.
Ultimately, the courts sided with Google, and denied the class action motion:
The court found that questions relating to whether class members had consented to the practice were too highly individualized to satisfy the predominance requirement based on the myriad disclosures available to class members.
In other words, Google argued successfully that users individually ‘read’ the terms of service agreement. Proving widespread wrongdoing was nearly impossible because it was a case-by-case basis.
The Terms of Service, which the court described as “vague at best and misleading at worst” wound up being the nail in the coffin:
Ironically, the fact that the court considered Google’s Terms to be vague or misleading and the fact that the Terms were not presented uniformly to end users appeared to actually help Google… it led to more individualized inquiries as to whether the users had given their express consent.
2017 and Beyond
Still here? Haven’t lost you yet? Great! Let me know what you think – and don’t worry – there’s not Terms of Service for reaching out to me here.