Monday, June 30, 2014

Facebook Can (and has) Manipulated You With Their Feed Algorithm

I came across this post from David Holmes over on Pando. It may have been in the news previously, but this was the first time that I'd seen anything about it. What happened, specifically, is that people over at Facebook undertook some basic experiments on their users. They modified the algorithm that delivers the feed to users, and for one set, they delivered more negative, less positive emotionally colored results, and for the other set, they delivered more positive, less negative emotionally colored results. As you might imagine, both groups essentially began echoing and amplifying the tone of their feed -- the negative group producing more negative content while the positive group produced more positive emotions.

The Pando post calls into question the idea that Facebook would simply experiment on it's user base like that, without any permission beyond the basic terms of service. And, while I appreciate the sentiment and feel that, at some level, there ought to be a deeper underlying set of online rights of self, I find the idea that businesses experimenting with algorithms without the consent of their user base to be wholly unsurprising. After all, much of this is the essence of serving online content -- A/B and multi-variant testing has been a common practice in online marketing for several years.

People want to believe that the content they are seeing is the same for everyone, that it's like a book. Once it's printed, that is what it is. But dynamic content, by it's fundamental nature, is manipulated.

For me, I see a much bigger danger implied by the Facebook study reference here. If you think of the ability to influence in this way, then the Facebook "scientific" paper is potentially a press release announcing their offering of an entirely new type of advertising. Imagine if you wanted to broadly shift public option. With the right amount of money and access to the right platform, you could pay for shifting the filter of the algorithm. This is not an Adwords type of program, accessible to anyone. Instead, it's a much more exclusive, strategic type of marketing. Probably expensive, with very limited availability -- like product placement.

Perhaps the scariest place for it, and it's most likely target, is with politics and political advertising. Consider how this type of behavior manipulation might be used for political influence. Depending upon who is operating the platform, how sophisticated they are with their manipulation, and what kinds of limits or controls that they place on this type of influence, the not-entirely-offbase conspiracy thinking can set your brain on fire. It makes you suspicious of any political content in the social media context; except, as Facebook has shown, people are manipulated without even knowing.

Clearly, it's an eye-opener and some serious food for thought. Is this where Facebook finally announces the real influence of the platform that they've been building, the platform that everyone has believed to be worth so much more than social media advertising has proven to be worth, to date?

Wednesday, June 25, 2014

Bad Brand Reputation Spillover: United Airlines and Chase Card

So I just received this email from Chase promoting an event. The subject header was, essentially, United Card from Chase. I realized, as I looked at my inbox, that I was having a visceral reaction to seeing the email -- essentially, my brain was saying, "oh, hell no" without even opening the email.

My reaction was not dissatisfaction with Chase, the credit card, or the promotional email, it was a resurgence of my anger and frustration with United Airlines. In fact, while I actually like the card and have been happy with Chase, I've found that I've been increasingly hesitant to use it simply because of the interrelationship with United -- the very premise that helped sell the card.

It made me wonder, how big are these anti-brand spillover ripples? How much do they affect your consumer behavior? Are there brands and services that you avoid spending money with simply because of their interrelationship with another brand that you're unhappy with?

Saturday, June 7, 2014

HBO's Silicon Valley: I expect better

Just so you know, I started this review of HBO's Silicon Valley after the pilot episode which, btw, I hated. Initially, I found that it was so disappointingly bad that it actually spurred me to write a review. But I keep wanting to give it a chance, to let it breath and open up -- this is, after all, a series from the same guy that did Office Space. Plus, there's something about watching movies or TV programs that are set in the place that you live, whether it's just watching for landmarks or looking for geographic continuity, it draws you in.

Silicon Valley delivered little over the course of it's seasonal run. Overall, there were maybe or four moments in the series that seemed amusing. And for those few moments which I would be hesitant to call gems, you had to endure the rest of it. Back at the first episode, here's what I started to write. I can't say that my opinion changed much.

So I worked my way through the first episode of HBO's Silicon Valley show yesterday. I say worked my way through because I've had root canals that were more pleasant -- at least I was medicated for the root canal.

More than amusing me or making me chuckle, Silicon Valley just made me angry. It was 50 minutes of time wasted not just because it wasn't good, but because it was a series of "funny" takes on "modern" Silicon Valley culture that have been done multiple times in other places. They trotted out the stylized tech company campus meant to be like Google or Facebook complete with the multi-person bike "used for meetings". Hollywood loves them some multi-person bike.

For me, there were so many tired Silicon Valley tropes that imagine a mythological world that we supposedly live in:
  • The doctor pitching investment in the device he's developing; why, because I guess people in Silicon Valley people are supposed to be like people in Hollywood pitching their scripts.
  • Technology people describing their product with a long string of meaningless techno-babble keywords and phrases.
  • The universal 'demographic mix' of programmers
  • The great core technology breakthrough that will revolutionize the world.
Admittedly, we live in an amazing place and it would seem like there is the framework for a great story here -- perhaps that's why it's so disappointing to deal with the same old tired cliches.

Compare Silicon Valley to Amazon's now cancelled Betas series. There were a lot of things that I didn't like about Betas, but there were aspects of it that were way more authentic. Instead of being so Hollywood-formulaic, Betas felt more start-up, exploring territory that conventional movies and television don't go. The thing that actually grabbed us into that first episode was the bit with the Asian female character and the texting prank, something that never would have made it on network TV. And that's sort of what makes Silicon Valley so maddening. This is HBO. They aren't bound by the lowest common denominator decency rules and limitations of network TV. But clearly, they are still bound by its conventions.

Fundamentally, I think that Silicon Valley falls down exactly because it tries to find the humorous aspects of Silicon Valley.
Over the course of it's season, Silicon Valley was ripe with cameos. This seemed like yet-another Hollywood device to make it feel authentic, "if we just use real personalities, it will be more real". In that way, it almost seemed like it was more interested in making it seem real than making interesting characters and an interesting story.

Another one that was surprising, one of the guys over on the Pando Daily staff really seemed to like Silicon Valley. He actually wrote several rather positive reviews. It's actually one of the reasons why I put my initial review on hold -- I looked at his review and thought, "am I missing something?" It's a good reminder that, like restaurant reviews and various other opinions, to each their own. 

The One Moment
Looking back over the course of it's season, there was just one moment for me, one brief gag that spoke to a deeper truth of Silicon Valley. It was in episode 8, the season finale. It was like an actual cameo of authenticity. No spoilers, but let me just say that it starts as little more than a dick joke, a predictable bit of humor that flows through the season. But, in this one moment, there is a difference. The joke twists and the "engineer" takes over. Suddenly, you have an entire room working analytically to solve a theoretical problem without regard to the plausibility of the initial premise. This really is Silicon Valley in it's essence, this free-form willingness to throw resources and processor cycles on what-ifs, to engineer a world on a foundation of imagination.

If the rest of the series had been like this, Silicon Valley might be content I would recommend. Instead, the sum of its parts is about eight hours of wasted time.

Wednesday, June 4, 2014

Imagining an Enhanced Version of Gmail Mail Googles

Remember that Labs extension for Gmail that would help you avoid sending late night drunken emails? Essentially, you could define a period of time at night where Gmail would challenge you with a math problem -- if you couldn't solve it, the assumption was that you were too drunk and might want to reconsider sending the email.

Imagine if they could craft one that helped prevent people from sending those really stupid emails. You know, those emails with questions like:
  • What's the address of the hotel we're staying at?
    You know the name of the hotel, why are you emailing me to ask for the address? Have you heard of a thing called search?
  • What time is the meeting scheduled for?
    Did you look in your calendar? I sent the calendar invitation so that you would have a reference for the meeting.
How do you filter to block stupid?
I'm not exactly certain, but that's why it might take a team of Google engineers to overcome this problem. You probably need to scan the email for commonly asked questions and then score it based on an index of dumb questions. Maybe the person receiving the email could flag it as stupid in order to build a database of stupid rank.

Then advertising based on stupid rank -- StupidAds -- could appear. Need remedial Internet technology education? Send us an email.

I can envision an entire Internet ecosystem built around stupid rank, we just need to get the ball rolling. Anybody have the email address of somebody doing developing in the Gmail Labs?

Sunday, June 1, 2014

Customer Service by Algorithm: Airlines Optimize for Functionality, Profit

While you may have tired of reading posts about airlines and customer service, I came across this article, Airport Confidential, in the March 3 issue of Time Magazine. While the majority of the article is behind a registration wall, I'll give you a basic synopsis. In the article, they go behind the scenes with American Airlines and provide you with a peek into their flight management system and airline cancellations.

Nicknamed "The Cancellator", the MRP-like software manages all of the carrier's flights across the fleet. In calculating and managing their operations, it takes into account a broad range of rules and considerations like flight crew availability, plane availability, and a host of other considerations. It makes decisions and calculations in real time, attempting to optimize for things like connecting flights, sometimes delaying take-off of one flight in order to ensure an optimal level of connections.

The article details how considerations like limitations on the hours that pilots or flight crews can work provide alarms or penalties that also must be considered and calculated into the schedule. While everyone understands weather, the article points out how complex and intertwined the entire system ecosystem is, where planes and flight crews all need to be in specific places and specific times in order for the entire system to operate. It mentions an example of one time when weather delays made things so bad for JetBlue, that they had to suspend their entire system just to move planes and crews back to the proper starting points for, essentially, a system reboot. This is Just-In-Time airline passenger manufacturing.

The Implications of Algorithm-based Airline Operations
Reading the piece, the big picture message that that they sell you on is, "flight delays and cancellations? It's not our fault. It's the system." At least, that is the thrust of the piece. But there are a number of other things that I took away from reading it.

First, system operation factors like delays and cancellations are not really differentiators in terms of the airline product. More or less, what's implied by the piece is that all of the airlines now use these same kinds of systems. Sure, you may be grumpy because XYZ airline missed your connecting flight, but that missed connection was equally likely on every other airline. Efficiency, as managed by the algorithm, should essentially deliver functional parity, depending upon how tightly you control the input parameters and the tolerances.

One example of the how the algorithmic performance might be tweaked would be to adjust the available resources. If you have flights that are being delayed by late flight crews, you could eliminate that by having 2x the number of flight crews, with one set left idle in the various cities. You might still encounter situations where weather stranded crews in locations, but with double the number of required resources, the chance of unavailable resources would be reduced. But it would also be an increased expense for the airline.

While some of the algorithmic factors may essentially be "acts of god", you've got to know that others are driven by business operating strategies. The article alludes to this in it's list of factors that are more likely to get your flight canceled. If United Airlines is running four flights a day from San Francisco to Chicago and two of those flights are only at 60% capacity, they can cancel one of the less full flights and remap passengers to the various empty seats, schedules be damned. From an operations standpoint, you're going to want to minimize those empty seats, and since the entire flight ecosystem is so complex, the business can easily attribute the cancellation cause to a host of other reasons -- flight crew delays, weather in another part of the country, or airport flight controls.

The Algorithmic Impact on the Customer Service Bottom Line
Most businesses these days are managed by sophisticated software and algorithms, so it's hardly surprising that a business that's as complex as an airline wouldn't use software to manage flights. The Time Magazine piece paints the airline operational algorithm in an "algorithm is god" perspective, ala Google's search algorithm. For the airline and their front line customer service, this is a helpful message -- "hey, it wasn't us, it was the algorithm. All of that cancellation stuff is beyond our control."

But there's a big difference between the airlines and Google or the software that runs an Amazon fulfillment center. You measure Google and Amazon by the results of the process, while the airline result is as much about the process of getting from point A to point B as it is the result.

For me, this article was another important lesson in understanding why the base level of service provided by an airline is probably more important than any 'premium' incentive. Take my recent flight to Chicago on United Airlines as an example. Because San Francisco is a United hub and Chicago is also a United hub, there are several flights back and forth each day. With the frequent schedule, there are multiple schedule options and lots of seats, so you can usually find a lower priced ticket on a plane with more open seats. But since it's a regular route with more flights, it's no problem for United, via their algorithm, to cancel the flight and pack the surrounding flights. Contrast that with an airline light JetBlue or Virgin America. With fewer flights, their ecosystem is probably more dependent upon maintaining schedules than optimizing for capacity.

And because the major carriers can do more to optimize for capacity, there are fewer empty seats, more 'Premier' passengers on a given flight, and correspondingly less chance of upgrade. It's not just your imagination. It's why, nowadays, you're more likely to have an empty middle seat on a JetBlue flight than a United flight.

This is why the base level of service, things like how much space you have in economy class and whether the airline charges for checked bags, is so important. The simple reality is that, unless you pay a premium surcharge for seat class upgrade, you can expect to ride in a packed economy seat on a full flight. Wouldn't you rather have that extra inch or two of leg room and, ideally, not have to battle for overhead space with uncle Fred and aunt Ethel?