Snapchat Night Vision:
Ephemeral Does Not Mean Anonymous
An Open Source Product Strategy for Self-destructing Advertisers
It has begun. The $10 billion dollar startup with no revenue is finally making moves in advertising. Over the past few months, Snapchat has been keenly focused on on-boarding brands and distributing sponsored content in an unsurprising and long-awaited effort to “start making money.”
And there’s a lot of money to be made!
Snapchat is reported to have over 100M monthly active users. It is also particularly popular among female teens, an extremely coveted demographic among advertisers due to their callous disregard for their fathers’ credit card bills. While exact numbers have not been published, a back-of-the-envelope calculation can give us a sense of how much mullah this could translate to:
Looks good, but Snapchat will not become a (ten-)billion-dollar advertising platform by spamming its entire user base with movie trailers all day. There is only so much we can take, guys.
It doesn’t take Don Draper from Mad Men to know that when ads are more targeted, conversion is higher, users are happier, and there’s more business to go around.
Just look at the big kids in the space: Facebook and Google have incredibly advanced tools that allow advertisers to run campaigns on extremely specific segments of users.
For example, Facebook could serve a Ouija ad to a 24-year-old Medium blogger named Brett Goldstein who lives within half a mile of a movie theater, posted a status from Twitter saying he wants to see Ouija, and recently published a blog that mentions Ouija (edit: they just did).
This kind of creepy micro-segmentation is possible thanks to the boat loads of content that you, I, and billions of other users willfully share on each of these platforms every day.
Think about it: everything you say and do is being recorded and analyzed to help advertisers get you to click that big red buy button.
And that’s the problem for Snapchat: everything you say and do is deleted. You know. Because ephemerality.
So the question is this:
How can advertisers run campaigns for specific segments on Snapchat if the content necessary to segment users in the first place is deleted?
There’s an answer, and it may alarm you.
Ephemeral does not mean anonymous
Snapchat can actually still gather a lot of useful information about you while still deleting your embarrassing selfies.
Snapchat can do some pretty decent segmentation using your geolocation alone. They could even know that you have a bad case of Senioritis — they would have 3 years of location history at/near UC Berkeley and no records in an academic building in the last 4 months. The creator of Google AdSense has a startup called Factual that helps companies analyze location data in this way.
As previously mentioned, the most valuable information is actually the content you generate, not a stream of longitude and latitude measures. To capture the information stored in your snaps, Snapchat could run each snap you send through a gauntlet of metadata extractors in the minutes and hours before it is deleted.
Yes. You read that correctly: Snapchat can analyze the crap out of your pictures before they delete them.
In fact, Snapchat might already be doing this.
When you send or receive messages, we also temporarily collect, process and store the contents of those messages (such as photos, videos, captions and/or Chats) on our servers. — Snapchat Privacy Policy
This gets a lot creepier when we look at what kind of information they can actually pull out of our snaps.
Text and Speech Analysis
Snapchat can access text you type on your snaps, text you type in messages, and even text you write on snaps with the paint tool after doing a little handwriting recognition.
There is an entire industry around processing this kind of language data, and for good reason.
Language data can help answer questions around your interests, goals, and current activities. Even measuring the frequency that you use various words can be used to estimate age, occupation and home town — I use the word “hella” a lot (err… I mean hella), guess where I am from (answer).
Speech data from Snapchat videos can be converted to text and processed in the same way, but there are some interesting analysis that can be done with speech specifically. For example, there’s a startup called Beyond Verbal that has an API to extract frighteningly specific emotional undertones in your speech. Snapchat could know that you actually hate your friend by the subtle antagonizingly sweet tone of your voice in the Snapchat video you just sent her for her birthday.
Background Sound Analysis
While there’s not a lot to say about someone with cars or cats in the background of their Snapchat videos, music is an interesting data point — especially for a company as focused on the music industry as Snapchat. There are a number of companies that allow developers to programmatically perform Shazam/SoundHound-like music identification — Echoprint is a good example.
Considering the dire state the music industry is in today, it seems obvious that no one has truly mastered correlating music tastes with buying behavior or even personality, but I’d venture that Snapchat would want to hold on to your music listening history anyway. Tone down those Justin Bieber karaoke snaps, man.
Facial Analysis
Perhaps the easiest way to figure out who you are from a snap is by running a facial recognition algorithm on your selfies (30% of all photos shared by millennials are selfies). Using reverse image search or even private Facebook API access via some sort of evil partnership, Snapchat could connect up the rich information you’ve shared elsewhere to your profile on their platform.
Naturally, this might be one of the more controversial techniques discussed, but Snapchat was never known for having a strong moral compass.
It is possible to do some lighter facial analysis to extract things like gender, ethnicity, age, etc. through one of the many facial analysis vendors out there. There’s even an API that can decipher your facial expressions — even if you have a BRF.
Non-facial Image Analysis
Analyzing non-human content in snaps might be one of the most important elements of Snapchat’s snap processing endeavors, especially when it comes to retail.
We all know that when you’re trying on clothes, you send snaps to your “trusted fashion advisors” (best girl friends) to get their approval. Now Snapchat can get in on the secret.
Using retail image recognition API like TinEye (and your geolocation), Snapchat could determine what the exact items you are trying on are and alert the relevant brands to send you a sweet deal.
Its pretty interesting to think about where Snapchat can take their advertising platform in the future, but today, advertising on Snapchat is like shooting in the dark. But that’s ok — Snapchat is still in its infancy. And they’ll develop night vision soon enough.
Even Facebook was lazy about user segmentation early on: Facebook was only available in Ivy Leagues, Ivy Leagues have a lot of Jewish students, so advertisements targeting Jewish students made a lot of sense. Quite a chachem that Marky was.
As Snapchat inevitably grows and begins to do deeper (read: creepier) analysis on snaps, we as users need to reassess Snapchat as the “ephemeral social network” and think about what we really value in it.
Considering what happened in the Snappening, “ephemeral” clearly doesn’t always mean “ephemeral” like we thought. And looking at what Snapchat can learn about us while still deleting our snaps, “ephemeral” doesn’t look like it will mean “anonymous” in the future either.
But, you didn’t join Snapchat because it was ephemeral. You joined Snapchat because it was easy, because it was fun, and because it was real.
And that will never change.
If you enjoyed this post, click “recommend” below and follow me on Medium and Twitter.
I’m also pretty good at Snapchat. Follow me: MontyDelMonte
