The Limits of Agile Having managed an agile development team for the last few years, I’ve gotten used to the fact that estimates of time and effort can be optimistic in the extreme. I understand that. We hire talented developers and they tend to naturally view all projects through the lens of their own ability. Assuming all goes well and assuming there are no delays from external sources and assuming development partners deliver on spec and on time, our estimates would be perfect. In practice, that’s a lot of assumptions. Apparently, this is a common problem for agile teams. I still believe that agile development is superior to any other process. But many researchers have begun to dedicate time and energy to fixing the blown estimates and deadlines that seem to plague many agile projects. (See below for an example.) Why does this matter? As agile processes move out of the pure development group and begin to guide other departments, it is worth reflecting on the strengths and weaknesses of breaking down projects into user stories. The biggest weakness is that we all tend to overestimate our own capacity for work. Having once, while feeling exceptionally well-rested and engaged by a project, completed X amount of work in a day, we now consider X to be our standard output. It isn’t. We all have bad days or bad night’s sleep or bad projects that don’t engage us. Further, we estimate based on our own capabilities without considering the externalities that may make our estimates invalid. It’s fine to say that our estimate “would have been right if the client’s internal tech team had delivered on time.” But in practice, delays were probably foreseeable and that should have figured into our analysis. All of us want to under-promise and over-deliver But agile estimating seems to encourage us to do the opposite. Next Steps: When evaluating a user story during an agile estimating session, ask yourself what could go wrong and how that might affect the process. Read More AI versus the Trolls Jigsaw, one of the many fragments of Google that now operate as a quasi-independent company, has announced a partnership with The New York Times website to improve moderation on the site through Conversation AI. Conversation AI is an artificial intelligence engine that attempts to recognize trollish behavior online and remove offensive and insulting language or spam from the digital conversation. Since The New York Times already has human moderators who do a relatively good job of guiding the tone and tenor of the comments, Discussion AI will help with flagging certain comments for moderation based upon language. Critics are quick to point out that context is the key to determining which comments are offensive and AI is bad at historical context. So a determined troll could find their way around AI-powered moderation by being a little more subtle in their language and references. Why does this matter? 2016 was a bad year for the internet. Trolls, fake news, and partisan nastiness debased the level of online conversation to a point where people are starting to disengage. Indeed, some platforms (like Twitter) may already be too far gone to be worth saving. Google or Facebook, companies that depend on a robust and engaged audience of internet users, have a financial interest in improving the level of conversation. Unfortunately, there’s only so much human moderators can do about the sheer volume of trolls. The anonymity of the internet breeds trolls as a natural bi-product. There are privacy and free speech concerns in handing moderation over to artificial intelligence. But most of these forums are private sites controlled by corporations, so they can control speech to whatever extent they wish. Next Steps: Most social media professionals are well aware of the danger of attracting the attention of trolls. It pays to err on the side of safety, even if it makes your social presence more vanilla. Read More Winter for the Telcos AT&T, Verizon, Sprint and T-Mobile had their day in the sun. When all cellphones could do was to make calls and send texts, they were free to rake in the profits knowing that it was prohibitively expensive for anyone to build a competitive network. Then smartphones came along. Suddenly, the Telcos came to the attention of Google, Facebook and Amazon. Crappy connections and high prices were limiting Google’s profitability. Google does not like that. As Google moves forward with Project Fi and other tech behemoths begin to eye the inefficiencies of the Telcos’ mobile network, it’s starting to look like the golden age of the Telcos could end. Verizon seems to be trying to get ahead of the trend by purchasing some content providers. Ironically, the end of net neutrality under a Trump administration could hasten Google’s entrance into the telco space. The old telcos have been trying to get rid of net neutrality forever so they could make money on an internet “fast lane.” But Google makes money on all lanes and they lose money on reduced internet traffic. The telcos should be careful what they wish for. Why does this matter? For consumers, it probably doesn’t matter. Year after year, the telcos compete with cable providers for who can be the most hated corporation in America. But this leaves open the danger of unintended consequences. It doesn’t really improve America’s substandard cellular infrastructure to drive all the telcos out of business. It would be much better to introduce enough of an alternative to keep them honest and encourage them to improve service and reduce prices. One hopes that Google is only using Project Fi to pressure the telcos. But there is a long history of animosity between Google and AT&T in particular. And providing cellular service is a zero-sum game. Next Steps: I got the Project Fi sim card for my new Pixel phone. But you should probably continue using your telco. Because I’m selfish about these things. Read More Digital payment and the economy Money is changing. This change extends far beyond cryptocurrencies or digital payment or peer-to-peer marketplaces although it certainly encompasses all of these changes. Cash is slowly dying. First because cash is terribly inefficient. A consumer may have sufficient wealth to make a purchase but if her means of payment (cash or a credit card) are not nearby, she won’t make the purchase. This has a negative effect on the economy. Payment also has a negative effect. Long lines at retail locations lead to cart abandonment at much higher rates than what we see through ecommerce. Increasingly, we see a continuum of different methods of purchase and payment, from going to the store and paying with cash, to ordering ahead through an app, to performing the entire transaction online and waiting for delivery at home. There is also a decline in centralization. The economy into which most of us were born depended on large actors (states and big companies). Money flowed towards these actors as taxes and purchases and then flowed out again as wages or social services. Increasingly, we are looking at a peer-to-peer economy in which transactions are small, instantaneous and direct. The economy, in the memorable phrase of Patrick Collison of Stripe, is becoming “higher resolution.” Why does this matter? Honestly, it’s hard to say how or why this will come to matter. Complex systems exhibit surprising emergent behavior. However, we are already seeing a transformation in the monetization of the internet. Initially, digital was funded using advertising-supported content. Now, as the internet expands into more and more of the economy (think: Uber or Amazon or Airbnb), we see that digital funds itself through sales of goods and services. This is good news since the digital ad marketplace is sick with fraud. Next Steps: I recommend listening to the podcast below from a16z to begin to understand how these changes will affect us all. Read More