The Situation

Nike.com search ran on legacy Endeca — thousands of manually curated rules, no ML, every result touched by a human. A decision was made to rebuild from scratch with an in-house ML system. I joined right as that decision was being made.

The Real Problem

The tech was the easy part. The real problem was organizational. Brand wanted control and storytelling — products placed for visual impact, full-price items front and center regardless of what consumers were actually searching for. Commerce wanted revenue and optimization. These teams had never had to agree on anything before. Now they had to agree on everything. And nobody trusted an algorithm they couldn't see into.

What I Did

Three moves that mattered:

The Numbers

The Situation

A 20-year-old US construction ERP product — used by hundreds of complex companies for payroll, project management, and equipment tracking — was sold into Australia and Canada before it was ready. I was sent in as the first person on the ground in both countries. The product worked in the US. It didn't fully work there yet.

The Real Problem

The product gaps were fixable. The trust wasn't — at least not easily. Buyers felt misled. They'd signed contracts based on what they were told the product could do, and what they got wasn't that. I was the first company representative many of them had met in person. I walked into rooms full of people who had every reason not to trust me or anyone I worked for.

The other problem: I had to figure out what actually mattered. Not everything on the gap list was equal. Some things were critical for their first tax season. Some things were nice-to-haves that could wait a year. Nobody had done this before for these markets, so nobody could tell me which was which. I had to learn it myself — from the customers.

What I Did

I showed up. In person. Multiple times. I sat with CEOs and CFOs. I sat with the people who would actually use the product every day. I asked what was keeping them up at night and I listened without defending the product.

Then I made a list of what absolutely had to work for their first tax season, their first major project close, their first regulatory deadline. Everything else went into a backlog with honest timelines. I stopped telling customers what they wanted to hear and started telling them what I could actually promise.

Internally I had to do the same thing — bring our product and engineering teams along on what actually mattered for these markets, and why some of what felt urgent in the US wasn't the right priority here. That required building credibility on two fronts simultaneously.

The Outcome

Our first Australian customer went live, got through their first tax season, and became one of our strongest advocates. That one customer helped unlock momentum in the market. The quarter we went live in Australia was the largest selling quarter in company history — not just in Australia, but company-wide.

I still keep in touch with the CFO from that first customer.

The Situation

Wayfair's search platform was running on a legacy system that couldn't keep pace with the business. The task: replace it with a neural network-powered engine, validate it with real revenue impact, and roll it out across five international markets and brands. No pressure.

The Real Problem

Replacing a search engine isn't a technical problem — it's an organizational and trust problem. Wayfair had merchandisers, SEO teams, category managers, and regional GMs who all had opinions about search results and a lot of anxiety about handing control to an algorithm. The technical work was challenging. Getting 60+ people across multiple countries to trust the new system enough to let it run was harder.

The other real problem: we didn't fully know what “better” looked like until we were in it. The nearest neighbor algorithm we launched with needed adjustment. We had to build in enough experimentation infrastructure to catch that quickly and fix it without losing momentum.

What I Did

Built the A/B testing framework before we needed it. That was the foundational decision — we couldn't make claims about revenue impact without a rigorous way to measure it, and we couldn't iterate on the ranking model without fast feedback loops.

Worked closely with data science on the ranking model itself — defining which signals mattered for relevance, where business goals like margin and newness should come in, and how to balance personalization against consistency across markets.

Partnered with each international market's merchandising team before rollout, not after. Understood the regional nuances — what “relevant” meant in Germany vs the UK vs the US isn't the same. Built those inputs into the model rather than treating them as edge cases to handle later.

On the US launch: no diving catches. We planned for things to go wrong and they didn't. The contingency plans never had to be used, which meant the international rollouts could follow on the timeline we'd promised.

The Numbers

The Situation

Jama was saying yes to everyone. Requirements management software for any company that would pay for it — small, large, simple, complex. I joined as their first PM when the company had fewer than 40 employees. We had around 10 large complex customers and 25-30 smaller ones. The CEO was still helping close deals. The sales team was immature. The product was being dog-fooded and shaped by whoever was loudest.

The Real Problem

The small customers looked like growth. They had shorter sales cycles and we were closing them. But when I did the analysis on who was renewing and what the unit economics looked like, the picture was different. Small customers churned. They cost almost as much to support as the large ones. And they were pulling the product in directions that didn't serve the complex use cases where we were actually winning.

The large customers — aerospace, defense, medical devices, automotive — were harder to close. But when we closed them, they stayed. And they had problems complex enough that our product actually mattered to them in a way it never would for a company that could just use Jira.

What I Did

Made the case for minimum deal sizes. That meant saying no to a category of customer we'd been saying yes to, which is uncomfortable for any early-stage company with revenue pressure. Had to build the argument from data, not intuition — churn rates, support costs, renewal rates, lifetime value by segment.

Then shifted focus to the segments where we had real defensibility. Automotive was the most interesting — Tesla was becoming a major player, companies like Faraday Future were promising full autonomy. The regulatory requirements for automotive electronics software were intense, and nobody in our space was set up to serve them well.

I identified ISO 26262 — a critical certification for automotive electronics development — as the key unlock. Competitors had it. We didn't. Prospects weren't always asking for it upfront, but it was becoming a deal-stopper at contract time. I got a quote from TUV SUD (the leading certification body, based in Europe), built the business case, and convinced the executive team and board to invest in the certification process.

We were the first agile SaaS company they had ever certified. Their process wasn't designed for how we worked — I had to work with them to adapt their surveys and audit approach. Then coordinated the internal audit process across the whole org.

The Outcome

I also launched a decision-tracking feature during this period — zero to one, built with our design and research team — that logged 10,000 decisions in its first month. Customers immediately started asking for it in other parts of the product.

The Situation

When I joined Workday as a Manager of Platform Product Management, I inherited the Interoperability team — three people, no shared direction, no product identity. Within a year I also took on the Delivery pillar, building what became a team of nine across two distinct product organizations.

No trust in the roadmap or in each other. The org had a name — UI Platform — but nobody could tell you what it stood for or why it mattered. My job on paper was to lead the Interoperability and Delivery pillars. My real job was to build a team that could do the work first.

The Real Problem

Platform PM is a hard job to hire for and a harder job to manage. The work is invisible — you're not shipping features customers can see, you're building the connective tissue that makes everything else possible. Without a clear story for why that matters, teams drift. Engineers lose faith in product. Product loses credibility with engineering. And everyone loses the plot.

That's where I started. Two of the original three people weren't working out. The team had no vision document, no operating model, no way of measuring whether what they were doing mattered. And Workday was simultaneously pushing hard to become a platform — which meant the stakes for getting interoperability right were only going up.

What I Did

Three things in parallel, and order mattered.

On the managed-out side: I had to make two difficult calls early. Both people left on good terms. Making those calls quickly was one of the most important things I did — it cleared the way for the people who could do the work.

The Outcomes

The Kernel platform went from an internal tool to the connective tissue for Workday's ecosystem play — universal header for acquisitions, mobile web view platform, first GenAI chatbot strategy established. WebView Extend launched to all customers after a successful early access program. CI test time cut 50% through code migrations. GitHub Actions migration reduced pipeline failures and freed up developer capacity.