Arena: Market Conditions Before New Relic
Ruby on Rails developers and web operations teams · 2008 · Worldwide
Barriers to participationHigh setup and upkeep costsDisconnected workflows
In 2008 a developer whose Ruby on Rails application slowed down had to dig through log files to find which controller, database query, template or cache call was at fault NR1. The application-monitoring software sold to large companies was installed on the customer's own servers and needed months of consulting before it worked, which put it out of reach of a small web team NR2. Server health was watched separately, usually by operations staff running open-source tools such as Nagios and Graphite DD10. When a site went down, each side looked at its own tool and switched between them to find the cause DD10.
How each step happened
Step 1 of 5 · 2008–16
Started inside the application code
A lesson from Wily, applied to Rails
Lew Cirne had built Wily Technology, whose monitoring software carried, he says, the flaws of big-ticket, on-premise enterprise software NR2. New Relic began as his project to learn Ruby on Rails. RPM launched in May 2008 as a hosted service: a developer installed a free plug-in and saw the slowest controller actions and database calls within minutes NR2 NR1. Cirne picked Rails as a segment small enough to own NR2.
The choice put the product inside the application. Each agent ran within the application server process and measured the code's own methods NR16, so New Relic was generally installed only on application servers, where traces are generated DD10. Peter Offringa, who ran teams on it from 2012 to 2015, says its design assumed static servers DD10. The view from inside the code also depended on the layer below it. On Heroku, New Relic's queue-time metric came from headers set by Heroku's router; from 2011 to February 2013 it became inaccurate and "eventually useless", and customers could not diagnose slow requests NR14.
Monitoring for cloud hosts came by acquisition: New Relic bought Opsmatic in November 2015, and New Relic Infrastructure entered private beta in August 2016 DDX-1 NR21. "The notion of a static server is fading," Cirne said then NR21.
Rivals Datadog launched in 2012 with infrastructure monitoring for cloud hosts, which put its agent on nearly every host, because every host needs basic monitoring while not every host needs tracing DD1 DD10.
Developer-First StrategyFounder Domain Expertise
Step 2 of 5 · 2008–20
A closed agent for each language
The agents were built alongside the first product, so this step also starts in 2008. New Relic widened its reach one language and one hosting platform at a time. In October 2009 RPM 2 added a Java agent beside the Ruby one; both reported to a common console, and a REST API let other vendors use the data NR15. By 2011 there were agents for Ruby, Java, PHP and .NET NR16. Hosting platforms carried the product: Engine Yard offered RPM free at launch, and Heroku sold it inside its own console, which brought 15,000 sign-ups by 2011 NR1 NR16.
Each agent was New Relic's own code, written for one language and supported by its own engineers. Installation was quick inside the frameworks New Relic covered, and coverage grew only as fast as New Relic could write agents. New Relic released its agents and integrations under an open-source license in July 2020, twelve years after launch DDX-2.
Rivals Datadog built its agent in a public repository from December 2010 and added integrations for AWS, Azure and OpenStack by 2014 and Docker and CoreOS by 2015 DD31 DD24 DD26.
Distribution Partnerships
Step 3 of 5 · 2008–19
Priced per server, then sold to enterprises
Trial users, then inside sales
New Relic charged by the server. Three months after launch it added tiers at $200 a month per server, and in September 2008 a free Lite tier that Cirne says drove awareness and sales NR2. In 2009 it hired inside salespeople to call trial users, who converted better NR2. Accounts grew as customers monitored more servers: at the 2014 listing the typical customer paid 14% more each year NR13. Airbnb installed the Ruby agent in 2009 and in 2013 moved to a site license covering all New Relic products NR11.
Field sales for the enterprise
From 2014 New Relic turned to large enterprises. Erica Schultz, who joined that April when a quarter of revenue came from enterprises, moved selling from an inside team in San Francisco to field representatives and technical presales in local markets NR12. By 2019 enterprises supplied about 55% of revenue, and the annual filing credited growth mainly to more product use by existing accounts NR12 NR19. Expansion was slowing: late in 2019 the net expansion rate was 109 to 112 percent DDX-6 DDX-5.
Rivals Datadog charged $15 per host a month with a 14-day trial and no charge per user, kept adoption inside large customers bottom-up, and by mid-2019 had about 40% of its customers on more than one product DD24 DD19 DD1.
Product-Led SalesSubscription PricingFree Trial
Step 4 of 5 · 2014–20
Separate products until New Relic One
By 2014 New Relic sold seven products, including server and mobile monitoring beside APM NR11. Ryanair used APM and Insights, New Relic's analytics product, to show website performance beside live booking figures NR17. Infrastructure followed in 2016 and a Kubernetes integration in 2018 NR21 NR18.
The products were added one at a time, each on its own terms. Offringa says that after 2015 New Relic kept to a limited slice of APM while Datadog and Elastic built full observability, and the infrastructure and log features it added "felt bolted on and not fully integrated" DD10. Breadth also did not settle diagnosis: after an early-2012 incident, RP Data evaluated New Relic and chose AppDynamics, judging that New Relic would still require "quite a bit of further investigation" NR22. New Relic One, launched in September 2019, put logs, metrics, traces and infrastructure in one platform, and logs linked to traces across all its application agents in January 2020 DD10 DDX-4.
Rivals Datadog released tracing in February 2017 on the same tags as its infrastructure data, at one price per host for both, and bought Logmatic for logs in September 2017 DD23 DD22.
Multi-Product
Step 5 of 5 · 2019–23
A price reset, then taken private
The slowdown came before New Relic changed its prices. Its annualized net expansion rate fell to 115.9% in the quarter to March 2020 from 130.9% a year earlier, with less upselling and more customers cutting spend NR20. In July 2020 it rebuilt the offer around New Relic One: data at $0.25 per gigabyte ingested, full-platform users priced per seat, and a free tier of 100 GB a month, so customers "no longer have to sample which applications they monitor" NR3. Nancy Gohring of 451 Research questioned whether per-user charges would limit participation, and the merger proxy later recorded lost customers and downgrades during the change NR9 NR10.
Installed habits slowed departures but did not stop them. Cloudbeds' move from New Relic to Datadog was simple technically, but retraining developers stretched a planned two months to about six NR8. The new pricing would have tripled Cloudbeds' bill; its negotiator won nearly 90% off the new list price, and still chose Datadog for stronger logs and more active development NR8. In 2019 New Relic assembled the single platform Datadog had sold since 2017, but having begun in the application code (step 1), it had no agent on every host to build it on.
New Relic remained a large business. Revenue reached $599.5 million in the fiscal year to March 2020 NR20, and in November 2023 Francisco Partners and TPG took it private at an equity value of about $6.5 billion NR5.
Rivals Datadog's dashboards and alerts span infrastructure, tracing and logs in every team, so leaving it means rebuilding them and retraining engineers; one customer took about a year to move to Grafana DD14 DD13.
Usage-Based Pricing