Rand Fishkin's Data-driven Disney World
Rand Fishkin does things a little differently than your average techie. It’s immediately noticeable.
Moz, the SEO and marketing company he founded, grew out of a blog, not a dorm room. He identifies as a feminist, and tweets it for the world to hear. He believes marketing should come before product. His mustache is elegant, pointy, Dali-like, and on the other side of his pompadour is a closely shorn undercut that probably scores many points in the Seattle scenes (after my interview with Rand, I show his picture to a friend in the cafe lobby of KEXP, and she immediately responds, “I see that guy everywhere and have always wondered about him.”)
But Rand’s most fascinating divergence manifests while he’s describing his product philosophy.
However quick to disclaim that Moz’s engineers and product managers spend a lot more time on tools than he does personally, he admits that when he gets involved on a project, he can become pretty meticulous in oversight.
“When a product feels as if it comes from a singular vision, there’s a kind of magical experience,” Rand says. “It’s not that you can’t trust engineers and product people and designers to do stuff on their own. It’s just that you need someone with that overarching vision to be part of it at every step, right?”
This theory, that a magical experience emerges from a singular vision, isn’t out of place in tech. But as Rand continues to talk in this vein, his eccentricity shows again: “It’s very, very hard for a disconnected team without that vision and that leadership and oversight to see it through. One of my favorite examples is…”
And at this point, he’s perfectly teed up to say Steve Jobs, right?
“Disney World. You have Walt Disney, and he wanted every aspect of every piece of the park to provide this magical experience. He worked with a massive team to execute it, but he was part of the vision and the approval in every step.”
In re-animating Disney, Rand grins so hard his eyes nearly shut and the wings of his mustache fan out. He really wants search engine optimization to feel like Space Mountain.
For the past few years, Rand has worked in this limbo between the big picture and control freak-ness. Having been CEO of Moz from its inception until 2014, Rand now acts as an “individual contributor.” He’s occasionally referred to this new role as “product designer,” only to qualify it as “not the UX/UI kind, but the strategic ‘this is what we’re gonna build and why’ kind.”
But even if this is what we’re going to build and why steers the ship, it occasionally has to alter course for little things. Rand cares deeply for details, too—which doesn’t make his job any less difficult.
Being Walt Disney, in the weeds, is tricky business.
E pluribus unum
By 2013, Moz had made a name for itself in the marketing tech world, but it was stumbling. For five years consecutively, it had grown 100% year-over-year, before dropping to a growth rate around half of that. There were many reasons for this, including delayed launches and troubled recruiting, but Rand readily admits his own struggles as a CEO, people-managing.
At the beginning of 2014, he stepped down from his executive role. In the context of tech, this would usually signal an ousting; a show of no confidence; a founder banished from something that grew faster than him. But Rand actually made this choice, and he’s still at Moz, and he prefers it this way.
As CEO, he was too far removed from product-building, which hampered the Disney approach of vision and approval. He could express his vision, but he wasn’t really able to fine-tune and approve things.
“Having been very hands off in a leadership role with a product that really tanked for us years ago, now I’m at the weekly meetings,” he says. “I’m in Slack watching all the conversations. I’m approving every design and UX element. I’m testing every data source in a backend dev environment. It makes a big, big difference.”
And if Rand says the word “vision” too often, it’s because he earnestly means it. It occurs to me that there’s a paradox in getting so close to the product while harping on vision. Does vast oversight risk being data-averse? Rand thinks of it as complementary. While big data doesn’t tell Moz to create this tool or that, it certainly helps when making the vision visible to others.
“I want to build an amazing park where people dream about going,” he says. “Being data-driven means saying, ‘The lines for this particular ride are too long. What can we do to help alleviate that? How many people can we let into the park at a time?’ It’s very much at the on-the-ground, tactical level.”
A fair amount of people want into the Moz park, and the newest attraction is Keyword Explorer. But such demand is relatively new. Keyword research had long been territorialized by Google, and when Rand first built the product, nobody was feeling the magic. In the prototyping phase, his vision was imperceptible.
Here’s why he built it anyway: For most of Moz’s history, its empire had been built on its web crawlers, which scour the internet, trying to mimic Google, and give customers a competitive edge in terms of link research. But the company had never devoted many resources to the SEO discipline of keyword research.
After moving into the individual contributor role, Rand wanted to productize this problem. He saw a twofold opportunity. For starters, Google’s AdWords tools were becoming less accurate. The search giant was giving oddly specific search volumes that didn’t register as correct to the Moz team. Apparently, Google was calculating a range for search volumes and then selecting an arbitrary number within each range to represent them.
“We uncovered this through research, by bidding on a bunch of keywords and then analyzing them,” he says. “Google’s numbers actually meant a bucket average range with some elements of randomness and standardization put in.”
On top of this, no SEO had ever developed a satisfactory workflow for keyword research.
“I watched dozens of SEOs do their keyword research,” Rand says. “Just the process that people had of exporting to CSV, putting it in Excel, deleting all the rows they didn’t want, adding in all the columns they did want, then going out and fetching the data for those columns by copying and pasting — man, that sucks, right? The entire workflow piece was completely missing.”
Between Google’s worsening functionality and the cumbersome processes SEOs were using, Rand envisioned a solution.
By combining Google’s data with clickstream data, Moz could offer more comprehensive search volumes than AdWords. By estimating an accurate range of searches, instead of Google’s needlessly precise numbers, Keyword Explorer could give SEOs more actionable information. And by developing a centralized hub to research, list, and compare keywords, Moz would solve the long-ignored problem of messy workflows.
As an individual contributor, Rand couldn’t force anyone to work with him. He had to go to specific developers, designers, and product managers, and rally them behind his cause. Any time they put into Explorer would come outside ongoing projects. With limited resources, the team was able to develop a workable product. But after Rand rushed the prototype to a few of his most-trusted SEOs, he received a reality check.
“It was what I would describe as a very classic minimum viable product,” he says. “Very minimal in terms of the feature set. And the feedback was, I would say actually, worse than not good. These are mostly friends and colleagues. I’m showing them this product, and they’re clearly underwhelmed.”
But the early feedback wasn’t all bad news. It served to reinforce another one of Rand’s core beliefs: “A minimum viable product is not viable.”
Maximum viable product
This is another way Rand goes against the grain. Maybe he’s been in the Northwest long enough to avoid the fad, but throughout the Valley, minimum viable products are accepted as a shibboleth of thrift, customer-centricity, and market readiness. It’s the nature of SaaS business and the developer style du jour.
But if you’re trying to build Disney World, these won’t work. Rand’s super-small beta proved why MVPs are antithetical to his approach.
“It’s fine to launch MVPs privately,” he says. “If you’re going to launch publicly — especially if you’re an already existing brand and you have a community — you want people not to be mildly impressed or say, ‘I can see the promise of this product.’ You want them to be blown away.”
However risky, he postponed the January launch date. While the cancellation upset his executive team, whom he was now answering to instead of leading, Rand needed the Explorer to be a magical experience. He couldn’t release a tool if it wasn’t going to blow his peers away.
“None of them were thinking, I’m going to switch away from my existing process to this tool. It was minimal in terms of the feature set. It had the data in there, but it didn’t really have a lot of the keyword filters. It didn’t have the graphs that show you distributions. It was missing a lot of the features that now set Keyword Explorer apart and put it above and beyond a lot of the competitors out there.”
He didn’t despair that the tool was coming at the wrong time or unwanted. Every day Google seemed to be reducing its features for SEOs. But like a ride with a long wait, Explorer was still too encumbered for his audience to reach a certain kind of joy in it.
“You collect information and use data to inform your vision and make you alter course slightly,” he says. “But a vision is much less based on what’s going on right now and what we should tweak.”
Rand drew up seven features that would be necessary to get Explorer to be a viable product and the engineering organization bargained that he could have five of them. Then they went to work.
In May, Keyword Explorer launched, and Rand was feeling a little more confident. He returned to the confidantes who had initially pooh-poohed the product: What do you think?
Change the world?
Dan Shure of evolvingseo.com, one of the dismayed early reviewers, sent Rand an email after the launch: You nailed it. This is it. I’m actually going to switch my keyword research to Moz— Dan.
This is it, was the confirmation Rand needed. It was more like his language. To release a product calculably superior? That was always possible, but it wouldn’t provide much of an experience. When the ultimate version of Keyword Explorer arrived, it did so with a Twittersphere avalanche, the kind of positive resonance that distinguishes a product that is awesome from a product that is simply better.
But Rand’s still trying to fine-tune it to perfection, to get it a little closer to being magical. And listening to his user base, there are evermore changes to make, like adding international search data. This doesn’t mean he’s failed. It means he’s succeeded. SEOs want more.
As with many of the impressions around Rand, his post-launch data stack is a bit more eccentric. It’s a quartet made up of Twitter, analytics, blog comments, and help desk tickets. He’s even gone so far as to personally email users who bought Explorer separate from the rest of the Moz suite, to get their input.
“I’m not using a software program to automatically send them out,” he says. “I’m custom-crafting each message.”
It’s a bit laborious, but it nets him the extra data. And it puts a Randian touch on things. Maybe a smaller startup could’ve iterated its way to adequacy, but he has a sense, even if he isn’t running Moz any longer, that he still has to live up to its standard.
“We’re not here to change the world,” he says. “We’re not Telsa. We’re not building rocket ships. We are helping marketers do better marketing. I like marketing. I hate shitty marketing. I love helping people get more traffic to their sites and their projects (at least most people). That is a big enough vision for me.”
. . .
Will you be in Seattle on July 21st? Stop by Mixpanel’s Office Hours at Altstadt, where DocuSign will reveal the twelve essentials for building a data-driven culture. Use the code SIGNAL for free entry—on us.