A Product Manager’s Farewell
Originally written December 17, 2010
Update (May 2020): I wanted to revisit this post since it’s been almost ten years since I wrote it and six years since I posted it on Medium. Some of the content holds up pretty well and other parts aged terribly. Most notably the first section on Content Policy does not match my current views and I wanted to include a note about that.
When working on Blogger, our stated product mission was to deliver “push button publishing to the people.” It’s hard to remember but in the early 2000s, the notion that the web could be used as a platform of self-expression was a radical idea. The web was where you went to get information, not to create culture. Even within Google, Blogger’s mission was not understood or embraced. So I ended up becoming even more extreme in my position that we simply needed to give everyone a voice online and that would usher in a new era of understanding. I brought this worldview to Twitter when I started working there in February 2007 and you can read an old blog post of mine that captures this idealistic zeal.
As with Blogger, there was always content on Twitter that we would rather did not exist. But we maintained the position that we would try to be as permissive as possible, removing content that was illegal or spam but trying to avoid making other types of content decisions.
From my vantage point in 2020, this was clearly a mistake for a few reasons. First, we cross applied the Blogger content policy but didn’t stop to seriously interrogate how different a platform Blogger is from Twitter. Because of the Twitter follower model new types of harassment and manipulation were possible that simply did not exist on Blogger. We also did not evolve the policy as the product evolved. Most notably, the @reply feature of Twitter did not exist from the beginning of the product and was a convention created by the user base. Once it began being supported as a native feature of Twitter even more avenues for exploitation were created (but the core philosophy did not evolve.) This pattern repeated itself as Twitter continually added features like trending topics or Explore and finally algorithmic sorting — each time we created new responsibilities as the platform operators to think about our role as curators. But the core philosophy of the content policy did not evolve as quickly. Instead we just forced the Support and Abuse teams to deal with the impact of those changes.
The other important error to note is that because the management team of Twitter was overwhelmingly white and and male, we did not see the abuse happening on the platform in the same way that a more diverse team would have. It is simply easier for white men to ignore the most vitriolic and awful uses of social media because they are so infrequently the target of those attacks.
Both of these errors have had profound negative consequences and I own my part in those mistakes. The primary driver for why I wanted to work in government in 2015 was to work on something that was inherently of service and centered in trying to improve the lives of others as opposed to on something that was, at best, value-neutral. (I wrote another overly idealistic post about that decision too.)
Having had that experience in the public sector, I’ve ended up thinking a lot differently about how platforms should think about their content policy obligations. What I’ve come to believe is that it is not tenable for platforms to maintain a stance of editorial neutrality. That is, it is not possible for platforms to say “we are simply the printing press, we don’t decide what is printed.” Twitter, Facebook and YouTube each make millions of decisions a day about what kind of content is allowed, what is promoted, what is flagged and what is removed. They are exercising editorial judgement every day, they just are uncomfortable about that reality.
The platforms are also in a pickle of their own making because they spent so many years trying to establish their “platform not publication” position that it has left them open to exploit by illiberal forces. I believe there’s no way out but to embrace their editorial position and to have a content policy that is consistent with their overall brand philosophy. The companies do not have good frameworks for thinking through these issues in legalistic or value-laden ways. And their attempts to think in those terms both doesn’t match the way the rest of the business operates and doesn’t move at a speed commensurate with the challenges they are facing. But they do know how to make decisions in terms of what makes sense for their brands; they do it all the time and it’s what makes decisions about spam and pornography relatively trivial for them.
All of this is obviously wrapped up in the moment we are in in 2020 when these issues have been brought to the fore. Ultimately, these problems may not be addressable in the context of the current paradigm because we’ve locked ourselves into a mutually antagonistic approach. And it’s possible that no real progress will be made until new platforms emerge and we try again. If any of these ideas, either the old ones below or the newer ones here, help in that goal, I’ll be happy even if upheld as a model of What Not To Do.
I stepped down as VP of Product from Twitter Inc in December 2010. I had worked on Twitter for almost 4 years and, on my last day, wrote a document with some lessons learned to share with the company.
I’m reposting that document now on Medium for a few reasons. First, it’s been 4+ years since I left which felt like both enough time where I was safe from divulging any secrets and also not too long to be completely irrelevant. Second, I somehow still get pinged about this doc from time to time. Finally, I wanted to revisit and annotate to see what still rang true.
I corrected a couple typos but otherwise it’s unedited from the original except for one redacted section as noted. It is super long so apologies for that.
I jotted down a few valedictory notes below about things I’ve learned working here. This is not a strategic doc or a vision for the future; it’s much more tactical than that. But maybe you’ll find yourself in a similar situation in the future and will benefit from what I’ve seen.
Content Policy
When I left Blogger, we were embroiled in a battle over what type of content should be permitted on the service. Our policy had long been to be as permissible as possible, but this was being continually tested by new, offensive stuff. The problem is once you start getting into an argument about “what’s offensive” you’ve pretty much already lost.
Fortunately, I’m not much worried about that happening here. From the beginning, we’ve embraced the philosophy that more content is the best cure for offensive content (be it hateful or gross or harassing or whatever.) We have some structural advantages in the product (block, one-way articulation) that give users some control of their fates. And we’ve set up a team that is even less squeamish than I am about what we permit (this is what you want by the way; the least offended should set the standard because otherwise you’ll blink.)
Why is this virtuous? In addition to believing that more expression is better, the tactical reason is that you are more likely to do long-term harm to your brand through take downs than through letting stuff stay up. It’s the accidental or ill-advised take-downs that truly jeopardize your status (both legal and in the eyes of users) as a purveyor of information as opposed to mediator.
Once you start adjudicating content, it becomes hard to justify why not to do it in one more case. And then you start eroding at the core thesis of the product itself; that open and transparent communication makes the world a better place.
tl;dr version: Amac and Del know what they are doing here. Even if it seems wrong to you (and it has sometimes to me), you should follow their lead.
Fooled by Numbers
There’s no argument against being a metrics-driven product team so I’m going to make one.
It is imperative that we set numeric goals for the things we build. A key function of product management is knowing what success looks like. That’s easier when you have a number. Moreover, it’s important to be very honest and clear about how the service is doing in terms of growth and activity. These figures should be routinely reviewed by the entire company. In fact, they should be unavoidable.
However, there are two things to remember about Product Design by the Numbers. The first is that it’s really easy to trick yourself into drawing the wrong conclusions. This can happen by looking at the numbers over too short a timespan, by measuring the wrong thing, by willfully drawing the wrong conclusion. More often it happens because there are few cases where the numbers are actually “clean” in the sense that there aren’t outside factors you need to consider.
Second, there are cases (like A/B testing language for a signup button) where the numbers will probably support a clear conclusion. But in many cases, I would value subjective judgement and alignment to long-term vision over the metrics-based approach.
An example: I would posit that if we added threaded comments to tweets, it would increase engagement metrics. But I don’t think we should build that features because, in my view, the addition of an addition noun/verb complicates the product and sacrifices some of the core simplicity that is the hallmark of its success. You could argue that that “success through simplicity” should be explicitly measurable but I think that would only be true over a long time horizon and the numbers would be muddy and highly interpreted.
Instead of relying on a simplicity metric, I think it’s more important to have an expansive tangible vision of what is and is not the product. And then have a team that works realize as best as possible the full extent of that vision. Yes, this will involve qualitative, subjective judgement because not everything in the vision will be reducible to a metric. But I think when you’re dealing with a new and evolving form of media, like Twitter, there are many cases where you need to follow your instinct. We should resist the notion that this is sinful.
Fatalism
Now seems like a good time to confess that I am a fatalist, perhaps with respect to life in general but definitely with respect to products. To employ an overly-pretentious analogy, when asked how he approached sculpture Michelangelo famously said “I saw the angel in the marble and carved until I set him free.” (Except he said it in Italian and probably didn’t because these things are always made up.)
Similarly, I believe that products have an innate organic growth potential and the trick is getting the surrounding cruft out of the way so that it can realize its full potential. People talk about that Facebook had this genius ninja move to open up the product late to a wider audience, thus spurring its phenomenal late-stage growth. But really, the core product was always there; the choice was just how to expose it over time.
For Twitter, this means a couple things to me. First and most importantly, don’t try to be what we’re not. We were born with a certain DNA and we evolve, but we can’t just up and decide to grow wings and fly. Concretely, this means that we can stretch to do more “social” things but we can’t (and shouldn’t) try to reach into the core FB use case of “share photos with your friends.” Other examples: we shouldn’t add more verbs (checkins) to the system just because more social web actions are invented. And our strength is in public, one-to-many communications which suggests that private or group messaging is less a solid fit. (I think the right way to pursue some of these ideas is under separate brands than as Twitter.)
Second, it means that we have to be very intellectually honest with ourselves about how much and to what extent we can deflect the organic growth of the product. Don’t get me wrong, there’s a lot we can and must focus on in this area. But our goals must be tempered by the reality of large numbers and our experience in what we’ve seen.
Finally, some would conclude that if everything is fated and all we can do is plod along, why bother? The answer is that even if things are pre-destined, our choice is in how we act along the way. We ought pursue our destiny with conviction and dedication because that is how we achieve excellence.
Planning for Surprises
Now, just because we live in a pre-determined universe doesn’t mean we won’t be surprised. After all, we only get to experience the present (which is one reason I recommend Vegas, see below.)
One of the operating principles I pushed for most was “Let’s not pretend we’ve got it all figured out.” We’ve been continually surprised by how people use Twitter and I hope that is true for some time to come.
From a product-design standpoint, to me this argues for building for more holistic, general experiences rather than prescriptive ones (I think we probably violated this rule with Retweet but that’s a whole thing.) More concretely, we should introduce basic capabilities that allow for interpretation by users. So with Events, for example, I think it’s important we create an experience that allows for a wide range of organic events, rather than just large-scale, partnered events. It’s by tapping into the scale-free nature of Twitter that the features we build will achieve the most success.
A corollary to this is that when we have a vision for a feature, we should organize around that effort holistically rather than build it piecemeal. I think we lose the most operational efficiency when we choose to have one part of the whole built by Team A and then turned over to Team B. For this reason, I’ve been somewhat skeptical of prototype exercises that don’t have a clear home or route to productionization. I fear that too often, for full-fledged feature-based work, we’ll end up with things that we simply can’t get out the door. Sometimes this is done in the name of testing out what works, but we should be clear about the types of features that can actually be tested this way. As I argued above, for core capabilities in the product I think we need to align to an overall vision for Twitter rather than discover that in pieces. To be clear there are a set of problems for Twitter that must be tackled iteratively. The way in which we build the Giant Brain Machine* will be through algorithmic iteration. We will continually evolve the way we target ads and content to users and this type of innovation must be lead by engineering, done in a highly metrics-driven way and performed iteratively over a long period of time.
Some notes on Culture and Communication
There is a lot that’s intangible about culture and the first rule ought be “Don’t get too precious.” The culture changes with every person you add and if you simply try to preserve what you have, you’ll stagnate.
But there are some concrete qualities I would focus on. At the top of the list is open and transparent communication. I think Q&A at Tea Time and company-wide meeting notes are important. They give the opportunity for inspection and tough questions. Yes, it means that sometimes information is presented with less interpretation. But I’d argue this is worth the price to foster a more shared context of what’s going on and greater trust between employees.
Obviously, there will be times when things are discussed behind closed doors. For example, if management is contemplating a re-org. However, it’s in this case where open and transparent communication can help the most. The instinct with a re-org is to try to keep things confidential and some information will necessarily not be shared. But once the decision has been made that a change in structure is necessary, it’s important to act quickly to start talking about it. First of all, once any set of people start to know about a re-org it becomes difficult to act as though it’s not happening. Second, it takes a long time to effectively communicate a re-org. You have to spend the time actually explaining the context for the change and getting feedback. And there are always more people to talk to than it seems. But if you do it right, folks are less surprised and more able to hit the ground running; the work you put in beforehand ends up saving you time afterward.
With respect to Tea Time in particular, some of my favorite moments have been during employee presentations. In the Bryant St. office, Evan Weaver explained a situation we were having in Infrastructure by explicating a poem. It was one of the most amazing things I experienced in my whole time at Twitter. I think that bottom-up quality is important with not just Tea Time but the culture as a whole.
That should just about cover it. You guys will have to take it from here. Don’t hesitate to reach out if you think I can help.
Sincerely,
Jason Goldman
*Giant Brain Machine. We’ve used this term as a short-hand for the long-term vision of Twitter since about 2007. It means an information service that knows what you want before you want it. The manifestation in the product would be a timeline of content that is exactly what you’re looking for and is immediately delightful (fun and useful.)
I think we can deliver this on a user’s first visit (i.e. logged out, no account) to Twitter. The dream is that they hit the Twitter icon on their phone, we infer stuff about them based on what we can get on the phone and show them a great set of tweets. We then lead them into an experience to refine and tweak that timeline. If on the web, we prompt to save with an email and password when they navigate away. We don’t prompt for username creation until they try to create content.
If I had to pick one, and only one thing, that Twitter needs to deliver on in order to grow 10x, it’s this frictionless consumption experience.
