Satya Nadella's message to the Microsoft troops yesterday underlines the way consumerization has changed computing already: To Microsoft, everyone is now a "dual user" who uses technology for work and play. That's two chances to lose a customer if Microsoft products don't delight them.
To make sure that those products do delight, and do what people need, Nadella is turning to some of the tenets of Silicon Valley startups like LinkedIn, Facebook, Twitter, AirBnB, and Netflix: Data science and growth hacking.
Change agents and growth hacking
If you talk to people who work at Microsoft, you'll have heard them use some new language this year, with phrases like "change agent" and "growth hacking."
Getting comfortable with change and being involved in changing things is what Nadella pointed out that everyone at Microsoft is going to have to do; "Culture change means we will do things differently. Often people think that means everyone other than them. In reality, it means all of us taking a new approach and working together to make Microsoft better." One Microsoft, as you might say.
And growth hacking is a Silicon Valley startup term that's a lot more than just viral marketing, SEO, and A/B testing. It's about turning product development and marketing into a virtuous, data-driven cycle where you get more users by figuring out what users do and don't want; how they find your product and how they use it.
Josh Elman, now a VC at Greylock, tells a story about growth hacking in the early days of Twitter, when lots of people were signing up but few of them carried on using the service. Instead of emailing those users or trying to show ads to people who might be more likely to stick around, they focused on understanding what was going on.
"We dug in and tried to learn what the 'aha' moment was for a new user and then rebuilt our entire new user experience to engineer that more quickly."
The key was getting people to follow other Twitter users, so they were seeing tweets they would be interested in. "As we kept tweaking the features to focus on helping users achieve these things, our retention dramatically rose," says Elman.
His advice for growth hacking is very like Adam Pisoni's principles for turning a company into a responsive organization (something he's been doing at Microsoft as well as for Yammer customers). Find your heavy users who already love your product and find the features and the pattern of usage that made them into active users. Build things that attract new users -- whether that's your marketing or sharing from existing users -- and make sure there's a way for new users to get started that turns them into active users quickly. Then build more features that your old and new customers will love, and keep on going.
That means getting everyone involved in growth. Early on, Facebook had a growth team that included marketing, business development, product development, finance, and HR. It wasn't just trying to get more users; it was behind projects like the system for importing email contacts, making Facebook available in multiple languages by crowdsourcing translations of the interface, and even creating the Facebook Lite experimental interface (a slimmed-down version of the site).
One of the first times I heard "growth hacking" from someone at Microsoft was talking to Jeffery Snover about his "Just in time, just enough admin" toolkit for PowerShell at TechEd this year, when he compared fast releases and agile development to balancing on a bicycle. "You don't get stability by going slowly," he pointed out.
He went on to talk enthusiastically about the way Nadella was changing attitudes at Microsoft, in a way that appeals to many developers and engineers. "As an engineer, what I'm focused on is writing a ton of code and having customers use it and say thanks. Satya is fixated on usage, usage, usage -- what does it take to get people to use that code? He talks about growth hacking."
Getting that usage isn't just about developing the code of a product, and it isn't just about marketing it. Growth hacking says that those aren't the separate processes they've been in the past. It says that product features have to be based on data about what users do and what they want and measured against whether they attract and keep users, as does marketing; that building the product and getting people to use it are two sides of the same coin. Instead of throwing money at marketing with glossy adverts put together long after the product is finished, growth hacking says you need to understand your users so you can build features that help you get and keep more users.
According to Snover, Satya Nadella says to employees, "Take responsibility for adoption. Then have a hypothesis, say this thing will drive adoption and go test it. Instead of saying 'oh, the reason we don't have adoption is our marketing ,' take responsibility, do the marketing things you think will work and see what happens. If that works, go with it; if not, try the next thing. Don't throw the ball in someone else's court. You own it. Own it, own it, own it. Go drive adoption, make sure people use your product, make sure people love your product, make sure your product solves their problems."
From telemetry to data
One of the ways Microsoft will do that is data science.
Nadella's memo promises "We will be more effective in predicting and understanding what our customers need and more nimble in adjusting to information we get from the market." And as part of that "each engineering group will have Data and Applied Science resources that will focus on measurable outcomes for our products and predictive analysis of market trends, which will allow us to innovate more effectively."
That sounds exactly like what Dean Hachamovitch's team has been working on in secret.
The former head of Internet Explorer is now Microsoft's Chief Data Scientist. Microsoft won't give out any more details about his role, but another former IE team member, Rob Mauceri is now Director of Data Science, working under Hachamovitch. And his Twitter biography says they're "building a new data science team to improve MS products for phones, tablets, PCs, Xbox and more."
Hachamovitch is a natural choice to lead a team like that. Understanding how people use Internet Explorer has been a big part of the way recent versions of the browser were built. When the team started building IE 9, they knew things like over half the time people spent on their home PC was spent surfing the web and the way they decided which of the HTML5 APIs to support first was to look at what was being used by 7,000 popular Web sites.
"Telemetry is one of my favourite words of all time," Hachamovitch told us at the time. "We start from crash data; we look at crashes and analyze them, we see how much of it comes from add ins and how much from core IE. It goes through to usage data, which drives a lot of user interface improvements; how many people ever see this dialog, how many of them say yes or no and what can we infer about how confused they are. We have terabytes of this data we pivot every day."
Internet Explorer isn't the only team using telemetry -- nearly every other product team does -- but in the pre-One Microsoft days they probably used it differently. One thing the Data Science group can do is share what one product team has figured out with everyone else.
Then there are Microsoft's burgeoning machine learning capabilities. Power BI and the new Azure ML service give everyone access to machine learning techniques, but it's been honed by the way Microsoft is using machine learning, predictive analytics, and big data internally.
Now Microsoft is turning all that expertise on its own product development, not just by looking at what users are already doing, but predicting what they'll want to do next. A new data science team started up inside the Operating Systems group last November; now every product team at Microsoft will have one.