EverOps

There is a familiar problem in enterprise technology, which is that the people who tell you what to do and the people who actually do it are almost never the same people. You hire a consulting firm, it produces a deck, the deck describes a beautiful future state, and then the consultants leave and you are holding the deck. Separately, you hire staff-augmentation contractors, who are individual bodies with no shared playbook and no particular stake in whether the thing works. EverOps, founded in San Francisco in 2012 by engineers Chris Lahey and Tim Fewkes, is a bet that there is a third option, and that the third option is worth paying a premium for.

The third option is called a TechPod, which is a name that sounds like something out of a startup that raised too much money but describes something fairly sensible. A TechPod is a small team - two to four senior engineers - that embeds directly inside a client's organization. They use the client's tools. They attend the client's standups. They get paged when things break. Crucially, they come with what EverOps calls proven playbooks, meaning the pod is not reinventing incident response or CI/CD pipelines from scratch on the client's dime, and they operate with continuity and redundancy, so the whole engagement doesn't collapse when one engineer goes on vacation.

What EverOps actually does inside these engagements is the deeply unglamorous, load-bearing work of modern software companies: running production environments across AWS, Azure, and GCP; building platform engineering and GitOps workflows on Kubernetes; wiring up observability with tools like Datadog; hardening security and compliance; and - the part that tends to get executives' attention - finding money. The company reports an average of $1.7 million in cloud savings per engagement, which is a polite way of saying that most companies are wildly overpaying for cloud and nobody internally has the time or mandate to fix it.

The economics of this are interesting. This is not a project business, where you sell a fixed deliverable and move on. It is closer to a managed-services model, where pods stay embedded on ongoing engagements, which is the kind of arrangement that only survives if the client keeps deciding, month after month, that the value is still there. EverOps reports 95% client retention, and in a services business that number is the entire ballgame. Clients in services leave the instant the value stops being obvious. A firm whose clients don't leave is a firm whose clients can't easily do the work themselves, or don't want to - and either way, that's a moat.

The client list is a decent proxy for the difficulty of the work: Chime, Coinbase, Docusign, Life360, Notion, Peloton, Rigetti, Snowflake, Whatnot, Zendesk. These are companies where downtime is measured in reputation and regulatory attention, not just refunded SLAs, and where the phrase "good enough" is not a viable operating philosophy. EverOps says it has supported seven client IPOs and been part of the infrastructure story behind $186 billion of combined client market value, which - while a somewhat impressionistic figure - points at the same thing: this is the plumbing behind companies that made it.

For most of its life, EverOps did all of this quietly and without outside money. It was founded in 2012 and operated for a full decade before taking any growth capital, which is either admirable discipline or a good reminder that services businesses can fund themselves from revenue in a way that software businesses often cannot. That changed in February 2022, when Stephen Koza - then a managing partner at Blue Ono Capital - led a growth investment into the company, in partnership with Riviera Capital, California Bank of Commerce, and Spring Capital. And then Koza did the thing investors occasionally do when they really believe the diligence: he took the CEO job himself. The stated goal of the capital was to professionalize operations and build a proper go-to-market engine, which is what you do when you have a good product that has been under-marketed for ten years.