A month ago, after four years and a quarter as CTO, I left The New York Times. It was an amazing four years, and I want to take a moment to capture some of what we accomplished, and some of what I learned in the process.
First, I will dearly miss the Times. The opportunity to sit around a table with leaders like Mark Thompson, Meredith Levien, and brilliant journalists like A.G, Dean Baquet, Joe Kahn and James Bennet, has been priceless.
It was quite a ride. We went from 1 million to 5 million digital subscribers, really turning the corner on our digital transformation, and ushering in a new era of growth for the company. That means a stable foundation for the mission, which at the Times, always comes first.
Why did I leave? More than anything, I had the good fortune to assemble a great team of leaders in Cindy Taibi, Brian Hamman, Erinmichelle Perri and Marc Lavallee, and at this point they have more to learn from unfiltered experience — doing the work — than they do from me. So the prospect of staying felt like lingering, looking over their shoulders, occasionally meddling. Weighing that prospect against the prospect of a new adventure, I chose the new adventure.
So that chapter has concluded, and I’ve begun another, running Engineering at Fastly, a company I have long admired. I’ll share more about my decision and my experience starting at a later date.
How do you evaluate Technology?
I’m very proud of what my team and I accomplished. But that question — how did we do? — is an interesting and nontrivial one to answer. How should you evaluate the performance of a Technology organization? My answer is that a tech org should be evaluated according to the business metrics by which the company itself is assessed. There are two reasons why.
First, while intermediate metrics like productivity metrics or performance metrics can be useful diagnostics, in the end the business results — in our case, subscription growth — are what matter. And focusing on the core KPIs of the business forces you to think hard about how to tie those intermediate metrics, and the work itself, to real impact.
Second, alignment is gold, and non-alignment is poison. A Tech team performing masterfully against misaligned goals is, just, a terrible situation. Sacrificing alignment in the service of self-defined functional glory is never the answer — fighting through the often painful process of alignment almost always is.
We did that. And, it was hard. I probably spent more time on defining, aligning and measuring goals than anything else at the Times. This included building a decentralized, cross-functional team structure, with a full OKR program that rolled up to enterprise goals set at the strategic level. That took a long time and a lot of work.
So the accomplishments I’ll talk about are all the ones that I believe helped move the business forward. By definition, tons of people across the company contributed to those results, and even the tech contribution was in every case the work of teams, not individuals. So all of the credit is widely shared. That said, here are some of the things we did that I am the most proud of.
Perhaps the single most impactful thing we did was to bring growth product thinking and practice to the Times. Growth product is the concept of leveraging product mechanics to drive growth, through a laser-like focus on the key conversion moments in the customer journey. It has become commonplace in the tech world, but it’s a big adjustment and a major lever for companies that have traditionally relied on direct-marketing driven conversion.
The key for us was finding, and then listening to, some super smart people. I want to thank Andrew Chen, Brian Balfour of Reforge, and Aaron Shildkrout for teaching us how to do it. Reforge in particular really helped us educate ourselves en masse, injecting great thinking and practice into the org. I think about 150 Timesians went through their Growth Series. (As an aside, I stake some claim to suggesting that they build an Experimentation & Testing course, which they did!)
The second key unlock for driving growth was investing in our marketing technology platforms. We did three big things. First, we built out a Customer Data Platform (CDP) using ActionIQ. Picking ActionIQ was a bet — they were and are a young company — but it paid off for us. Tasso Argyos, Action IQ’s CEO, proved to be a great partner to us. With them, we were able to marry the usability of a typical campaign management solution with infrastructure grade architecture and scalability. Meaning, they know how to make a useful product for marketers, and build tech you can actually integrate into a modern infrastructure. That was big.
Next, we rewrote the meter service that powers our paywall. This enabled us to run effective tests on all meter parameters, leading to two key outcomes: the significant lowering of our meter, and the introduction of a registration wall. Those two changes were big, risky steps that had a huge impact on conversion.
Last, we put a ton of work into making it easier for us to experiment with price changes. This was very challenging due to entanglements with the legacy print transactional system, and because it’s just complicated. This work is ongoing — but we got much better at varying, testing and changing pricing, resulting in better exploitation of the demand curve, more subscribers and more revenue.
Fix The Data
We couldn’t have done any of that without fixing our data infrastructure. This was one of the first things we tackled, replacing a mess of Hadoop (serious, go-home-for the-day-while-your-query-runs nonperformance!) and semi-homegrown stuff with a nearly pure GCP data stack. Personally, my main contribution was to hold us back from defaulting to AWS (which could have been ok) and to encourage us to look at BigQuery. Nick Ursa and Vishal Santoshi did the real work, mind-melding with the Google team with able guidance from Alan Beaufour.
This was probably the single most important thing that we did. It was a lot of work, and migrating each data flow was painstaking work. It was also hard because we couldn’t see the benefits until most of the work was done. But it was worth it, and we ended up with a single source of truth for almost everything, all queryable by BigQuery, which we all know is a beast. That engine has driven an explosion of analysis led by Shane Murray and his team that really transformed our understanding of the business and our ability to test and learn.
We did a full cloud migration, shutting 4 leased data centers and moving our workloads primarily to GKE on the Google cloud. We chose Google partly because that’s where our data was, and partly because we felt it was the best first step in a journey towards serverless, which I think was mostly correct.
It was also HARD. GCP was still pretty raw when we started, and we suffered accordingly. It took a long time, with little to show along the way. But it was wildly worth it in the end: we saved a ton of money and gained a ton of flexibility. We also became much more secure, scalable, and reliable. The best part was that the migration became a huge forcing function to pay down tech debt of every imaginable kind.
I kind of hate to say that we rewrote everything, but we pretty much rewrote everything. We rebuilt our publishing pipeline on Kafka, our API layer on GraphQL (and now integrating the excellent Apollo platform), our web UI on React, our IOS app in Swift, and more.
We also handled organizational debt. Our Delivery Engineering team, a combination of SRE, DevOps and developer experience, was born out of the migration, created by Deep Kapadia to meet the challenges and harvest the opportunities of the cloud environment. We rethought nearly every aspect of how we built, shipped and operated software with an emphasis on resilience, performance and security. And we did it with far greater efficiency and a far more leveraged team.
The most unexpected thing that I learned was how important having a robust edge was to managing the actual mechanics of migration. We had deployed Fastly in haste right before the 2016 election, which worked out great. But the unexpected benefit of deploying Fastly came during the migration, when we were able to use the edge to control the testing and cutover process from start to finish, and to shield our customers from our inevitable mistakes. That made the whole process much more stable for our readers.
Personally, the most satisfying work for me was the work we did on our culture. I wrote about our approach here during the process, so I won’t go into too much depth right now. I’ll just add two things.
First, the most counterintuitive thing that we learned was that investing in a performance culture benefits the whole culture, which we worked hard to do, creating a professional development matrix that was quite demanding, implementing recurring 9 box exercises (that we actually acted on) and a 360 degree feedback process. We took promotion and hire decisions from manager/lead engineer (equivalent management/IC levels in our framework) out of the hands of any individual in favor of committees. We even relevelled people, including the wholesale abolition of the “architect” title. All of this was challenging for the team, but I truly believe left everyone feeling safer, because the expectations were clearer, the rules documented and transparent, and the results more equitable.
Second, this work was where I learned the most from the team. When I arrived, my thinking was behind the best thinking on the team. I quickly realized I had a lot to learn, and we really worked together to build our program. The most important contribution that I made was simply helping to lead the team through work that they knew better than me how to do. That was an amazing experience that really opened my eyes to the reality of servant leadership.
OK one last thing — I had a lot of fun with Sarah Bures and others reinventing our product development blog, Times Open.
We figured out how to grow the team. We went from hiring about 50 people in my first year to over 180 in my last. To do that, we had to rethink our process from the ground up, from the structure and size of the recruiting team, how we used cash for recruiting and hiring bonuses, how we interviewed, how we assigned candidates to teams, and much more.
R&D and New Products
I had the good fortune to work with Marc Lavallee on rebooting the storied Times R&D team. We worked hard to create a foundation and a rationale for R&D at the Times, in a lean environment in which R&D had to create room to dream and invent while still contributing to the business. We also faced unique challenges finding just the right ways to contribute in an organization where innovation is everywhere. That team is now doing great work.
I was lucky to have the opportunity in my last year to run our brand new product development, and to start two new consumer products, which I can’t describe yet because they are not public. But I can talk about the things that were the highlights of that process for me: getting a chance to work with James Bennet on identifying opportunities, and getting to hire two brilliant entrepreneurs-in-residence, Dorothy McGivney and Sarah Adler.
Too Many Other Things
This post is much too long and I have to stop so I can focus on my NEW job. But I am also super proud of the work Cindy Taibi did on enterprise productivity. We made a ton of progress bringing modern practices to a 160+ year old company, and navigating the crosswinds of the consumerization of IT. That will have to be another post.
Information Security was an adventure and certainly my steepest learning curve — and I learned a LOT. Huge thanks to Bill McKinley for his heroics, and Erinmichelle Perri for joining as our CISO and giving me the confidence that I was leaving the safety of the Times in able hands — far, far more than my own.
It was a privilege to work at the Times and be a part of their story. I couldn’t be more proud of the team, and I’m thrilled that they are each now free to grow to their next level of potential. There are way too many people, adventures, projects, failures, successes that I couldn’t fit into this post. I wish them the best and will be watching their progress, and I can only imagine what they will accomplish in the years to come.