Roam Research launched in public beta in Oct 2019 and calls itself “a note-taking tool for networked thought”. It’s been getting a lot of praise and attention from smart people on Twitter.
So let’s take a look. Does it have a compelling value prop, or is it a bloated MVP?
The Zettelkasten Method
Roam is inspired by the Zettelkasten method of note-taking. “Zettelkasten” is a German word for a box to store note cards in. The main idea of the Zettelkasten method is that instead of writing in a book of full pages, you write on tons of small rearrangeable index cards. The cards have bullets on them and use a special numbering system to enable cards to reference each other. LessWrong has a good in-depth explanation.
Roam’s hypothesis is that popular tools like Evernote, Google Docs and Notion are flawed for research and note-taking because they’re more like a book of pages than a pile of index cards.
Roam has two atomic building blocks: pages and bullets.
- A Roam page is like a Zettlekasten index card, not like a big, um, page. Maybe Roam should call these “cards” instead of “pages”.
- A Roam bullet is a smallish unit of text which lives on a Roam page and can be nested in an arbitrarily deep hierarchy.
What about Workflowy?
Workflowy is notable to mention because it’s built entirely out of bullets, which is similar to Roam.
Unlike Roam, Workflowy doesn’t have a separate concept of a “page”. Your whole Workflowy account is like one big page of bullets. If you want to make a new page, you just make a new top-level bullet instead. Then you can “zoom in” to that bullet point and it’s like you’re “on that page” seeing only its sub-bullets.
I personally use Workflowy pretty often, so I get why a bullet-based tool is quick and easy for taking notes and drafting ideas. The main question I had about Roam is how it’s incrementally better than Workflowy for my needs.
Roam’s MVP seems to already have feature parity with all of Workflowy, which is pretty impressive. (It’s not fully caught up yet on multi-user collaboration and scalability, but I’m sure it will be soon enough.)
What does Roam do that Workflowy can’t? Roam’s #1 differentiator is linking: You can easily link a bullet (and its page) to another page by putting the page’s name in a double square bracket or hashtag.
For example, if I’m editing a bullet on my “Marine Biology” page and I type
#SpongeBob, then I’ve linked that bullet to my “SpongeBob” page. What if I didn’t have a SpongeBob page? I do now, automatically. I can then go to my “SpongeBob” page and automatically see that “Marine Biology” references it, and maybe other pages reference it too.
While Workflowy encourages you to work within one top-level bullet at a time, Roam encourages you to split off lots of top-level “pages”. Even just putting a new word in [[brackets]] automatically makes it a page, so you find yourself dumping every concept-node in your brain as a page.
If you’re writing bullets in Roam, and you just add a couple extra keystrokes once in a while, pretty soon you get a cool linked graph of pages. Then you can go through all your pages and flesh them out with more bullet points, and those will inspire linking to new pages that you’ll want to flesh out afterward. The overall effect is that Roam feels like dumping your brain into a “second brain”.
I think Roam’s data model of pages/cards full of bullets is powerful, and has never been properly explored in another tool before. I’m excited about that because I personally want to use that data model. I think it’s an incrementally better version of Workflowy’s bullets-only data model.
What’s the Value Prop Story?
Being made of a novel set of simple, powerful building blocks is a very good sign. But can we get more specific about what use cases this could get market traction for? What’s a value prop story?
Joe drafts his blog posts in Workflowy and often gets ideas to write about related concepts. To keep track of those, he creates top-level bullets for those other concepts. But Workflowy isn’t that great at managing lots of top-level bullets, so Joe only creates a top-level bullet when he thinks his related concept is a promising candidate for a blog post. If Joe switches to Roam, he feels like his tangential ideas spill more easily out of his head without being prematurely filtered, because he can set them aside as linked Pages.
Jerry has been using the Zettelkasten method with physical index cards. By switching to Roam, he gets to keep his same Zettelkasten patterns plus get the benefits of a digital medium: unlimited storage, easy copying, cloud backup, integrating images and multimedia. (And if he had switched to another tool like Notion or Workflowy instead of Roam, there would be more friction in attempting Zettelkasten, especially in trying to link one card/page to another.)
There are at least two well-formed Value Prop Stories for Roam, and they’re not hypothetical — they seem to describe some specific people I’ve seen talking about Roam on Twitter. I myself am kind of a Story #1 Joe. I currently pay for Workflowy and I think that Roam is significantly better than Workflowy for my needs.
Is It Bloated?
Roam has a clear hypothesis to test: Will people prefer to structure their notes and drafts on top of a pages-&-bullets data model? The question is, did they build an MVP that efficiently puts that hypothesis to the test, or did they build a bunch of other bloated crap?
My verdict is… NOT BLOATED! Here’s why:
- No major non-core features. They held back from building important, tempting, but non-critical features into MVP. While they couldn’t resist including a graph visualizer, they resisted including any major features that would have slowed down their launch, such as realtime multi-user collaboration, because those features weren’t necessary for validating their data model hypothesis.
- Minimalist design is a good call. The UI design they launched with does a good job of being simple. The product isn’t particularly pleasant-looking, but it’s very usable. The hypothesis they’re looking to prove is that data model is their key differentiator, not UI design. So de-prioritizing UI design was a good call.
- Smart compromises on performance/scalability. Right now it feels like a ton of the application is running client-side, so that when you have a lot of data and you search for a common keyword, it freezes your comp for a bit. That’s good! They prioritized making a fast experience for people who just write a handful of reasonably-sized notes. Because by the time you’ve written a handful of reasonably-sized notes, you get hooked on why Roam’s data model is superior, and you’ve validated the key hypothesis of their MVP.
When I’d first read Roam’s White Paper, I’d figured their product was likely to be a bloated MVP because the paper is full of abstract claims about how people think and why software should be better at helping people think, but lacks specifics about use cases or product features. I changed my mind when I read the authors’ note at the end:
We wrote this White Paper in Winter of 2017/2018, and while it still reflects much of our vision for the tool, some of the features — particularly related to [[Bayesian Reasoning]], [[Argument Analysis]], and [[Prediction Markets]] we found to be in-fact much lower priority for the researchers and decision makers whose thinking we aim to assist — and risked making the tool too complex for the more significant use cases.
So the Roam team seems self-aware that their white paper isn’t usable as a product roadmap. I don’t see any other red flags that their MVP or their product roadmap is bloated.
Why I’m Excited
I’ve switched to Roam for the kind of content drafting I used to do in Workflowy. I drafted this post in Roam.
I’m sold on Roam’s data-model hypothesis: that pages and bullets are a superior data model for drafting notes and creative works of all kinds. Roam’s competitors, other editing tools, make you work in multi-paragraph units, without the kind of dynamic rearrangeable structure you get when you write on index cards Zettelkasten-style.
Roam is good product today. How much more awesome can it get than it already is?
Let’s look at its product roadmap. These features and improvements are obvious, and I believe they’re all in the works right now:
- Performance and scalability improvements
- Realtime multi-user collaboration
- Mobile app
- Developer API
- Third-party integrations and data import/export
- Bullet widgets — more goodies like the built-in widgets that let you use special bullet-text syntaxes to embed images, Tweets, todo items and range sliders. They’ll probably allow third-party widgets too.
None of these upcoming features seems like as big of a leap forward as the current Roam MVP’s introduction of a new pages-&-bullets data model. I suspect that’s the biggest innovation Roam will ever have to offer. So compared to Evernote, Google Docs, Notion and Workflowy, is Roam ultimately a huge qualitative leap forward, or an incremental one?
My guess is that it’s an Airtable-sized leap.
Airtable is a 5-year-old tool that I use for a bunch of stuff at work. All it does is let you read and write tables of structured objects, and it has become a highly successful company last valued at $1.1 billion.
Airtable’s innovation was also data-model-based: While traditional spreadsheets like Excel and Google Sheets let you edit flexible grids of cells, Airtable uses less-flexible lists of records with primary keys and structured column types. It’s actually similar to Microsoft Access, a tool for editing database data.
Airtable may not have any more big innovations up their sleeve to differentiate themselves further and further from Excel, and that’s okay. The company is on track to create $billions of additional value with an increasingly sophisticated product built around a data-model innovation.
So is Roam.