Step 1 of 5 · 2009–15

Correct first, three years late

A better database, finished later

RethinkDB was pitched at Y Combinator's demo day in 2009, before it had any software RE1. Slava Akhmechet's team wanted "an elegant, robust, and beautiful product" and chose three standards: correctness, with strict guarantees kept "religiously"; a simple interface, with the complexity absorbed inside the database; and consistency across the query language, drivers, cluster setup and documentation RE1. Each standard made the database better once it existed and slower to finish. After five years of development and 16 major releases, RethinkDB declared version 2.0 ready for production in April 2015 RE2.

Akhmechet later called these "the wrong metrics of goodness for most users" RE1. Developers wanted a database that existed when they needed it, felt fast on the tests they actually ran and was a good way to do a specific job RE1. "That put us three years behind the market," he writes RE1. The years spent finishing the database were the years in which developers made their choice. They chose MongoDB, and its early weaknesses were fixed one by one while it was already in use RE1. In Akhmechet's words, "MongoDB turned regular developers into heroes when people needed it, not years after the fact" RE1.

Rivals MongoDB dropped its application platform, wrote a driver for each popular language and launched publicly in 2009 MDX-1 MDX-2. Akhmechet lists the flaws of early MongoDB, from a global lock to weak sharding, that developers accepted in exchange for speed RE1.

Product LeadershipDeveloper-First Strategy

Step 2 of 5 · 2009–16

Free open source with little to sell

The business model ran alongside the build from the start, and on its face it was MongoDB's: a free, open-source database developers could download and run without a purchase RE1 RE5. The difference was what came after the download. Akhmechet says users saw RethinkDB as an open-source developer tools company, "because that's what we really were", and he calls that "one of the worst markets one could possibly end up in" RE1. Thousands of people used RethinkDB, often at work, but most would pay nothing at all RE1.

RethinkDB had little to sell them. Commercial services, meaning training, development help and production support from its engineers, arrived only with version 2.0 in 2015 RE2. Akhmechet estimates that a developer tools startup must work through ten times as many leads as other business-software startups to make one sale, which "demoralizes the team" and makes investment and hiring harder RE1. A slow sales funnel meant less money for product and distribution, and so less chance to recover the delay from step 1.

Rivals MongoDB chose its open-source licence as a route to paid products, demonstrated the database at language meetups city by city, and grew from 2,563 downloads in October 2009 to more than 30 million by 2017, with existing customers spending more each year MDX-1 MDX-3 MDX-2.

Developer-First Strategy

Step 3 of 5 · 2014–15

Always asked how it differed from MongoDB

By the time RethinkDB was ready to recommend for production, the comparison had been set. "Almost everyone was asking 'how is RethinkDB different from MongoDB?'", Akhmechet writes, and explaining the value of correctness did not change their answer RE1. MongoDB had become the document database that developers' code and skills assumed MDX-2.

In mid-2014, once it was clear RethinkDB could not compete head-on, the team looked for a reason to adopt that MongoDB did not offer RE1. It found realtime push: changefeeds let an application tell the database to send live updates whenever data changed, instead of repeatedly asking for them RE2 RE3. Version 2.0 was presented as a database for realtime apps, version 2.2 in November 2015 made changefeeds more reliable, and hundreds of startups and Fortune 500 companies ran RethinkDB in production RE2 RE7. The feature was real, but it put RethinkDB against Meteor and Firebase, which had been solving realtime for years; Akhmechet says it was again three years behind RE1. And the difference was a feature, which MongoDB could add later, not a different interface that developers' code depended on.

Rivals MongoDB had been the most popular modern database on DB-Engines since 2013, and more developers listed it as a LinkedIn skill than nearly all other non-relational databases combined MDX-2. Firebase and Meteor had built their realtime layers on top of MongoDB RE1.

Focus Strategy

Step 5 of 5 · 2017–19

Its remaining users moved to MongoDB

The software outlived the company. The Cloud Native Computing Foundation bought the rights to the code and gave it to the Linux Foundation in February 2017 under the permissive Apache licence, and volunteers kept releasing versions RE5. But users were now choosing a database with no company behind it.

signageOS, a digital signage company, shows how the realtime difference from step 3 wore out. It had used RethinkDB from the start because it could listen for changes to data in real time, and its user interface depended on that RE9. During a 2019 move to AWS, replication slowed the whole RethinkDB cluster; the team also faults its own migration plan RE9. It then replaced RethinkDB with MongoDB for three reasons: MongoDB also offered real-time change listening, it was "actively developed by a large company, 1000+ employees strong", and it had handled heavier load well RE9. The feature RethinkDB chose in 2014 to set itself apart was now available on the database developers already knew, from a supplier that would keep developing it.

Rivals MongoDB kept its users through the applications and skills written to its interface; Microsoft and AWS built MongoDB-compatible services rather than ask developers to rewrite MDX-8 MDX-5.