The quiet $1.2B company that taught enterprise IT how to stop drowning in its own alerts.
It's 3:14 in the morning at a Fortune 100 retailer. A core payment service starts wobbling. Twenty years ago, the on-call engineer would have woken up to 1,800 pages and a Slack channel screaming in twelve different fonts. Tonight, she gets one notification. One incident. With a root cause guess attached and a list of the three recent code deploys that probably did it.
That is, more or less, BigPanda's product description. It is also their pitch. The company sells software that takes the howling, overlapping, often contradictory output of every monitoring tool a giant company owns - Datadog, New Relic, Grafana, CloudWatch, Splunk, the lot - and squashes it down into something a human being can actually act on.
The pitch is not glamorous. Nobody puts an event-correlation engine on a T-shirt. But the customer list quietly reads like a who's-who of organizations that cannot afford to be offline: Intel, PayPal, Workday, Gap, News Corp, Caesars, Wix, Turner Broadcasting. In 2022, investors decided this list was worth $1.2 billion.
The number of monitoring tools isn't the problem. The number of alerts is.- BigPanda's pitch, repeated since roughly 2014
Most of them yell at the same time, about the same thing, in incompatible vocabularies. A failing database might generate alerts in nine tools at once. Each tool is technically correct. Together they are useless. Engineers learn to ignore most pages, which is fine until the one that actually matters arrives at 4:07 a.m. and looks identical to the 60 false ones from yesterday.
The industry has a term for this. It's "alert fatigue," and it is the reason a third of enterprise outages get noticed first by customers on Twitter rather than by the team paid to notice them. If you've ever sat in a war room and watched twelve smart people stare at fourteen dashboards, you know exactly what BigPanda was built to fix.
The cruel joke is that the more tools you buy to prevent outages, the more noise you generate, and the harder it gets to spot the real one. Modern observability is a chorus of tools all clearing their throats at the same time.
In 2012, Assaf Resnick was a Sequoia Capital investor with a finance degree from Berkeley's Haas School and a front-row seat to the dysfunction of every IT operations team his portfolio companies tried to scale. The thesis he kept hearing back was the same: the tools were fine, the alerts were drowning the humans.
He co-founded BigPanda with Elik Eizenberg, a computer scientist with a data-science background. There is, in the way of these things, a small piece of folklore: the company's first product wasn't AIOps at all. They started in ad-tech, then pivoted hard once they realized the correlation problem they'd been solving for advertisers was, in fact, the same problem haunting every enterprise NOC. The name stuck. The product completely changed.
In 2015, an acquirer came knocking. They said no. Resnick, by his own telling, wanted to see his baby "see the light of day." This is the kind of decision that either ages very well or very badly. In this case it aged into a unicorn.
We started in a radically different space doing advertising technology.- Assaf Resnick, on the pivot that made BigPanda
CEO. Ex-Sequoia investor. Berkeley Haas. Spends his weekends talking to NOC leads at companies most people have never heard of.
Chief Scientist. The data-science half of the founding duo. Built the first correlation engine that became the company's moat.
Sequoia got in at seed and never left. Insight led the Series C, D and the $190M Series E.
BigPanda is, at its core, a giant translation layer. Alerts come in from every monitoring tool a company owns. The platform normalizes them - reads the messy human-typed fields and decides what's actually happening. Then it correlates them. A CPU spike, a database timeout and an HTTP 500 storm all hitting the same service within 90 seconds? That's one incident, not three hundred.
From there, it adds context. Recent code deploys. Recent config changes. Topology data from the CMDB. By the time a human sees the page, the page already has half the investigation done.
Correlates and deduplicates raw alerts into a smaller number of high-fidelity incidents.
Adds change context, topology, and AI-generated summaries to each incident.
MTTR, MTTA, noise reduction, top-noise services. The KPIs ops leaders show to their CFO.
Drafts war-room updates, summarizes the timeline, suggests next steps. The TL;DR you wanted at 3 a.m.
Bi-directional sync with ServiceNow. Tickets enrich themselves.
Reads from the monitoring stack you already own. Writes to the on-call tool you already pay for.
You don't need more dashboards. You need fewer incidents.- Paraphrased from a BigPanda field marketing deck
Source: Crunchbase, BigPanda press releases. Series E led by Advent International at a $1.2B valuation. Cumulative: approximately $340M.
The stated vision is that one day, the platform handles the boring incidents on its own. The pager only rings when a human is actually needed. Self-driving operations for the data center, if you'll forgive the metaphor.
This is harder than it sounds. Self-driving cars at least drive on roads. IT environments are bespoke, fragmented, and full of legacy systems that someone's uncle wrote in 1998 and nobody is allowed to touch. To automate that, you need a system that can read the unwritten rules of every customer's stack and make decisions that don't break anything.
BigPanda's bet is that the data is already there. Every alert, every ticket, every postmortem is training data. The platform's job is to learn the topology and the human conventions, then quietly take over the dull bits. Mean time to resolution drops. Engineers sleep more. SLA penalties stop showing up in board decks.
Most outages aren't a tooling problem. They're a correlation problem.- The thesis, in one sentence
The interesting question is not whether AIOps works. It does. The question is what happens to the on-call engineer when the platform gets good enough to triage 80% of incidents without human help. The optimistic answer is that engineers focus on harder problems - capacity planning, architectural decisions, the things humans are actually good at. The cynical answer is that the on-call roster shrinks.
BigPanda is betting on the first scenario, and the timing might be on their side. Generative AI gave the entire category a new language and a new interface. Where you once had to learn the BigPanda correlation rules, now you can ask the assistant, in English, what just broke. The category is moving from dashboards to conversations.
The competitors know this. Moogsoft got acquired. Splunk got acquired by Cisco. ServiceNow built its own AIOps module. PagerDuty bolted one on. The AIOps space is consolidating fast, and BigPanda is the rare independent left standing with both scale and conviction. Whether that ends in an IPO, a strategic acquisition, or another decade of quiet enterprise compounding is anyone's guess.
Every minute of downtime is a minute that pays for the product.- Internal mantra, BigPanda go-to-market
It is, once more, the middle of the night at our Fortune 100 retailer. The payment service is wobbling. Twenty years ago this is a fire drill. Five years ago it is still a fire drill, just with better dashboards. Tonight it is a single notification, a half-finished diagnosis, and an engineer who, sometime around 3:23, hits a button and goes back to bed.
BigPanda did not invent IT monitoring. It did not invent machine learning. It noticed that the problem with operations was not a lack of signal but an overabundance of noise, and it built a company around the boring, useful work of turning one into the other.
That is the company. That is the bet. The pager, mercifully, has stopped ringing.