A shelf is a small place to discover how the world works. It has edges. Products have dimensions. Someone has to decide which ones deserve the space. Wendy Chen, who once traded at Jane Street, has made those decisions her business. The founder and CEO of Omnistream brings the habits of quantitative finance to grocery and convenience retail: look closely, test the assumptions, and let the results argue back.
By January 2026, her description of the problem had acquired a memorable measure of time. The half-life of a good decision got shorter.
A store plan can be perfectly sensible when it is drawn and less useful by the time someone rearranges the shelves. Chen’s work lives in that interval. Her software helps retailers turn sales data and the particular characteristics of a store into decisions that can keep up with its customers.
There is something pleasingly unglamorous about this ambition. Technology can promise almost anything. A shelf demands an answer about Tuesday.
A computer lesson, disguised as an English lesson
Chen’s first encounter with programming was a family arrangement. After moving from China to Canada, she was learning English. Her father, who had a doctorate in computer science, wrote programs to help her learn. At six, she was being introduced to data structures alongside a new language. He explained computing through algorithms and system design.
She later studied Systems Design Engineering at the University of Waterloo, earning a Bachelor of Applied Science. That educational choice fits the childhood lesson: understand how the parts interact, then work out what to change. A system can be a learning program, a market, or a shop. Each has rules, constraints and feedback. Changing one piece can alter the behaviour of the rest.
Her early professional record includes Research In Motion, Barclays and Qualcomm. At Jane Street, where she worked as a trader from 2007 to 2012, she entered a world in which mathematical judgment had consequences. A confident explanation could still lose money. The opportunity to make decisions came bundled with the obligation to examine them.
Her own professional biography describes leading automated commodities trading desks in New York and Hong Kong. It emphasizes strategy, risk management and the work of keeping autonomous systems within changing regulatory requirements. Those are useful credentials for a founder selling automation: she had worked with systems whose mistakes could be expensive, and whose operation required more than an elegant algorithm.
“The market doesn’t care how you feel.”
Wendy Chen, recalling a lesson from trading
First the trader. Then the investor. Then the founder.
Chen calls her path a reverse career
. Trading came first, angel investing followed, and building a company came after that. An early investment in Instacart brought grocery into focus. It gave her a reason to look beyond the transaction on a screen and toward the decisions that make a retail business run.
The sequence matters. Investing lets someone consider a business from a little distance. Founding removes that comfortable gap. You become responsible for finding the customers, explaining the product and discovering which of your attractive assumptions survive contact with someone else’s daily work. Chen had already learned to take uncertainty seriously. Retail gave that habit a different setting.
Early public records put her in Hong Kong’s retail technology community. In July 2013, she appeared as co-founder and CEO of StampUp/Omnistream at a seminar about digital marketing for small businesses. A 2015 business-intelligence profile described Omnistream’s personalized marketing tools, mobile customer communication and predictive engine. The shelves would become a sharper focus, but retail and customer behaviour were already in the picture.
In September 2014, she spoke to students and faculty at the National University of Singapore after SAP’s Asia Pacific Japan president, Adaire Fox-Martin. Chen discussed her experience as an entrepreneur and her participation in SAP’s Start-up Focus program. Long before the current conversation about AI agents, she was working through the question of how business software reaches people who can use it.
The shop at the end of the spreadsheet
Omnistream moved its headquarters from Hong Kong to Singapore in 2018. Chen was pursuing Southeast Asian retail, where different languages, business practices and local conditions made a uniform approach difficult. The commercial proposition was unusually direct. In its early outcome-based model, the company took a share of the profit it helped generate.
That arrangement gave a buyer a practical reason to listen. It also gave the vendor a practical reason to be careful. A successful demonstration would have to become a successful operation. The retailer’s restrictions belonged inside the calculation: a recommendation that cannot be carried out has little value, however impressive the mathematics.
Chen’s approach combined a retailer’s own transactions with external information about the setting in which its stores operated. The goal was to choose a product range for each location. A chain’s logo might be identical above every door; the demand behind those doors could differ. Treating the chain as one average shop could conceal the very differences that offered room for improvement.
For the shopper, the consequences are ordinary and immediate. The item they came for occupies a real space, or it does not. For the retailer, every choice also competes with another use of the same shelf. Omnistream’s current work produces store-specific planograms, the diagrams that tell a shop how to arrange its merchandise. Chen’s quantitative background meets a problem you can point at with your finger.
A customer needs a reason to take the risk
Chen has also thought carefully about the person asked to buy the software. In her 2022 writing on customer adoption, she treats reluctance to change as rational. An existing process may be awkward, but it is familiar. A new one carries the possibility of failure, and someone inside the business will have to answer for it.
Her response is to make the first step manageable. Test a small selection of stores and categories. Compare their performance with control stores. Make the improvement visible to the people who use the system and the people who authorize it. Design the pilot so that expanding it does not require beginning again.
She also argues for working with existing infrastructure and keeping the initial costs low. These details reveal an operator’s understanding of persuasion. A customer has calendars, colleagues and older software to contend with. Adoption becomes easier when those realities are part of the product design. The sale continues after the signature, through implementation and evidence.
In June 2022, Omnistream announced a $7 million Series A led by SIG Venture Capital, with Delivery Hero Ventures, Spiral Ventures and Wavemaker Partners participating. At that time, the company reported deployments across more than 8,500 retail locations in Asia, Australia, Europe and North America. The figures describe its position at that funding announcement; they give a dated view of the scale Chen had reached.
June 2022
at the announcement
Learning with the people who already run the stores
For Chen, entering retail required respect for the experience already there. She has spoken about the need for humility when working with an industry that operated long before her company arrived. That stance makes room for the store operator’s knowledge alongside the model’s predictions. The software can calculate; the business still has to function.
She has been equally direct about the difficulty of maintaining an intellectually curious company as it grows. More experienced hires do not automatically preserve the habits of the original team. In discussing Omnistream’s expansion, she described making painful personnel decisions to protect its culture of honest inquiry. The principle has a cost when it involves colleagues rather than numbers.
There is a revealing distinction in her attitude to certainty. She wants assumptions examined, including her own, while accepting that leadership involves decisions other people may dislike. The trading floor taught her to question the relevance of an insult rather than absorb its premise. Her account of gender expectations is similarly plain: the pressure on women to remain agreeable can make difficult business choices harder.
Outside work, her strategy games offer a less expensive arena for competing ideas. She has named Settlers of Catan and Civilization among the games she enjoys, describing play as another way to think about an effective strategy and how a machine might learn it. Even recreation comes with a decision problem. A game is an orderly place to ask a question that runs through her work: which choice makes sense under these conditions?
The decision has an expiry date
The more recent public work keeps returning to the speed of change. In June 2024, Chen discussed retail planning with Anthony Lacavera on Beyond Innovation. That same month, she joined a Coresight Research commerce conference in New York, alongside retail and consumer-goods leaders, to discuss localized assortments and the ability to test changes at store level.
By January 2026, Omnistream was working with Nimble and YDISTRI on a multi-agent retail demonstration at NRF, with Microsoft collaborators. The participants brought different parts of the operation into a shared scenario: market information, inventory and assortment decisions. For Chen, it was another version of the system-design problem, with more pieces that needed to work together.
Her account of the shorter life of a good decision gives the work its urgency without requiring a futuristic setting. Demand moves. Stores have constraints. A plan needs another look. Chen’s career has carried her from programs written to help a child learn, through financial markets, to software that helps a retailer decide what belongs on a shelf. The setting keeps changing. The discipline of checking the answer travels with her.