Kenneth Truong Took the Offer Nobody Understood - Then Waited for the Spreadsheet to Work

The email arrived while Kenneth Truong was leaving an Uber interview. His phone had been off. He turned it back on to summon a car and found a note from the recruiter for the other company, the mysterious one in Suite 420 with four people at the interview table and not much of a public footprint. Please call as soon as possible. Truong assumed he had failed. Instead, Sigma offered him the job.

There are neat career decisions and there are interesting ones. Truong had already worked inside Amazon, Qualcomm and Informatica. Airbnb, Snapchat and Uber were among the recognizable opportunities now in front of him. Sigma had not launched a product. Friends and family supplied the obvious arithmetic: famous companies were safer; tiny companies often vanished; free food was, at minimum, food.

Then investor Mike Speiser called. He spoke about co-founders Rob Woollen and Jason Frantz, the size of the problem and the kind of company they intended to build. Truong was impressed that Speiser had made the time. The pitch worked. When the other companies raised their offers, he had already decided. He wanted more impact than a large organization had given him. On June 13, 2016, he walked into Sigma as its first frontend engineer. Seven people were in the office.

A launch party with no guests

The new engineer was nervous. He was working beside people whose résumés made him question his own place in the room. His answer was neither an affirmation taped to a monitor nor a grand theory of confidence. He put in the hours and contributed where he could. For the next few months, the team built features and prepared to expose the unfinished thing to daylight.

September 1 was launch day. Truong stayed through the night polishing the landing page, caught a short nap on the office couch around six in the morning and waited for the first strangers to arrive. The site went live. The count of new users remained stubbornly, almost theatrically, at zero.

It is hard to find a tidier rebuke to startup pageantry. A launch button is only a button. Sigma had no marketing team, no sales team and, as the following months made plain, no crisp reason for customers to change their habits. Friends tried the product. Few saw enough value. The glamour of being first gave way to the administrative burden of continuing.

Truong began to burn out. Perhaps the sensible people had been right. He could have joined a company already moving toward a public offering; he joked later that he might at least have been eating well. Carol, a designer at Sigma and a close friend, urged him to give it another year. It was practical advice disguised as loyalty: he had already put in the time, so stay long enough to learn what the time meant.

The useful sentence at a weekend demo

The first hopeful signal did not arrive from a market report. During a weekend demonstration for friends, a business analyst remarked that Sigma could have helped at her former company. The introduction she offered would have to wait; that company had a deadline. Truong barely cared about the delay. Somebody had connected the product to an actual problem.

On Monday he delivered the news to the office, including an estimate that this single use case might save thousands of dollars a month. He kept going. Friends, networks, the other companies around Sutter Hill Ventures - surely there were more people with the same ache. His speech ran long enough for Carol to intervene with a compact editorial review: “ok! we get it!” Lunch could resume.

The team still had an architectural problem. Sigma was building a spreadsheet-like interface with its own query language, while also asking customers to store their data inside Sigma. A meeting with Snowflake exposed the awkwardness. Why would a cloud data warehouse move its data into somebody else's system merely to analyze it?

Frantz proposed the decisive change: put the interface on top of Snowflake instead. Let the warehouse remain the warehouse. Give business users a familiar surface without requiring them to write SQL. The team scrapped what it had built and spent a month producing another minimum viable product. Long nights returned, but this time the work had a more legible constraint.

On April 11, 2017, Woollen and Frantz presented the new version at Snowflake while the rest of the team waited. At 3:50 p.m., Woollen sent word that the demo had gone well and Snowflake would become the first alpha customer. Truong's written reaction years later was only two words, one of them unsuitable for a quarterly report: “Fuck Yes!” After almost ten months, the work had met a real user.

The interface becomes an institution

A founding frontend engineer is hired to make screens, but the screens are where a product's arguments become visible. Sigma's argument was that a person fluent in spreadsheets should be able to explore warehouse-scale data without first becoming fluent in database syntax. Every choice in the interface had to negotiate between familiarity and scale: the comfort of a cell, the unruly size of a cloud table.

Truong's later technical writing shows how unromantic that negotiation can be. Sigma tried Redux with Redux Thunk, a custom state system modeled on MobX with Firebase, MobX itself, and finally Apollo with GraphQL. The endpoint was not fashion. Apollo supplied polling, caching and optimistic responses the application needed. Then expected errors created another puzzle. If a user attached to a post had been deleted, should the application throw an exception, erase the post or represent the absence as data?

The pattern Sigma adopted used GraphQL union types so an expected error could sit where the missing object would have been. The interface could distinguish a known condition from a broken system and render it sensibly. It is a small example of the broader job: complexity is allowed backstage, but it should arrive in front of the user wearing a name tag.

Public patent records name Truong on multiple Sigma-assigned inventions in precisely this territory: dashboard loading from cloud-warehouse caches, editable tables, dimensional hierarchies, cross-account sharing, and systems for servicing database requests. A patent is not a personality test. Collectively, however, the records show years spent on the machinery behind fast, comprehensible interaction with enormous datasets.

Advice with a timer attached

When Truong writes for junior engineers, his advice is appealingly measurable. If stuck, research the problem for about 20 minutes before asking for help. Ask sooner than “days,” because self-reliance loses its charm when it becomes expensive. Break large pull requests into pieces that a reviewer can actually understand. Do not complain about inherited code as though you have inherited a haunted manor. Improve it.

He recommends finding the portion of a feature that delivers most of the benefit, then protecting it from the final portion that consumes most of the schedule. He advises keeping secondary work ready while code waits for review. These are not maxims designed for conference slides. They are queue management for people whose work depends on other people.

That same instinct appears in his long-running Stack Overflow account, where React and JavaScript dominate his answers. The public profile lists more than 4,000 reputation points, 64 answers and over one million people reached. The scale is oddly appropriate for someone working on analytics: many tiny acts of clarification, accumulated until they become a large number.

Risk, without the commemorative plaque

Truong has argued that startups can offer broader networks, faster learning, visible impact and financial upside. He has also warned that they are not interchangeable lottery tickets. He recommends inspecting founders, investors, product space and technical durability. His own experience gives the advice its tension. Sigma's first version did not simply need more polish. It needed to be discarded.

The distinction matters because his story could be flattened into an advertisement for boldness: decline the famous companies, join the tiny one, become right. But luck, timing and other people's judgment crowd every scene. Speiser took the call. Carol kept Truong from leaving. A friend noticed a use case. Frantz saw the architectural opening. Woollen carried the revised product into Snowflake. Truong's contribution included staying present long enough for all of those connections to become consequential.

Years later, the office photographs are almost comic in their ordinariness: fluorescent light, whiteboards, mismatched chairs, a life-size cardboard co-founder. In one birthday prank, Truong dressed to match Woollen and put cornstarch in his hair. Another year, colleagues used Woollen's face as their video-call background. Culture, in its early form, is often just the accumulated evidence that people survived a difficult week and were still willing to make a joke on Friday.

The clean moral would be to trust your instincts. Truong's own account offers a better, less decorative one. Instinct got him through the door. Evidence forced the product to change. Friends made doubt tolerable. Engineering turned a promising idea into a system other people could use. The spreadsheet did not work because somebody believed harder. It worked because a small group was willing to notice where belief had been wrong.