Research that changes what gets built.

SAP Ariba wanted a guided buying experience for people who buy things at work but aren't procurement professionals. Someone in Marketing ordering carpet cleaning, say. Rather than design it first and test it after, I ran a workshop to build it with the customers themselves.

First, storyboarding. We asked a room of customers to sketch their own ideal buying experience from scratch, one step at a time, before showing them anything of ours. No prompts from our designs, just what they pictured when they imagined the whole thing working the way it should.

Second, card sorting our early designs. In small teams, people took a set of screen cards we had prepared and physically arranged them on the wall into the flow that made sense to them. It told us less about which individual screens people liked and more about how they expected the experience to unfold from start to finish.

To keep the signal honest, we didn't let the loudest person in the room win. Everyone voted individually with happy face and sad face stickers on what worked and what didn't, then explained each vote on a sticky note before the group discussed the results together.

It matched a lot of our thinking, and corrected some of it. People wanted the flow to feel conversational, not form heavy. They wanted help and the ability to loop in a colleague built into every step, not added later. And if we were going to point them toward preferred suppliers, we needed to let them go around that list too, as long as they could explain why.

Ariba was growing 30 percent year over year, and every new customer meant more support tickets. The instinct was to hire more support staff. I wanted to know first whether this was actually a capacity problem, or a findability problem.

Three layers. A Qualtrics survey to external buyers and suppliers on how they look for help and where they get stuck. Workshops with twelve internal stakeholders across support, engineering, design, sales, marketing and customer success, to get everyone aligned before we built anything. And ten long interviews and remote contextual inquiries, watching people search for help in the moment instead of describing it after the fact.

One persona stuck with me. Nic, a business process specialist who managed support tickets for his team, told me exactly how the product failed him.

That distrust, more than any missing feature, was the real problem. We rebuilt the help experience around where his confidence actually dropped rather than where we assumed it would, and shipped in-product help that reached people at the moment they got stuck instead of after they filed a ticket. It freed the support team to spend their time on the harder cases instead of the findable ones.

Guided Buying's lack of accessibility was stalling deals and go lives in the EU. Customers were escalating, and fixing it wasn't optional. I put together a team of developers and designers and worked directly with SAP's global accessibility testers and with customers who had motor and visual impairments, rather than auditing against a checklist on our own.

Some of it was obvious once someone actually tried to use the product with a screen reader or a keyboard only. Buttons and links were built as clickable divs instead of real interactive elements, so assistive technology never registered them. Color was used as the only signal for status, which fails for anyone with low vision or color blindness. Fixing it meant real interactive elements, keyboard first navigation, and never relying on color alone to carry meaning.

It sounds small. Does splitting a keyboard for thumb typing actually work. But Windows 8 was Microsoft's bet on tablets, and the keyboard was the thing people would touch the most.

Four days, 12 participants, and I iterated the layout live between sessions as pain points came up. The clearest one was wasted space. The visual gap down the middle of the split felt awkward, even before anyone typed a word. That led us to try filling it with a number pad.

Was that number pad actually useful, or just visually reassuring? Eye tracking on 7 participants, after they had an hour to get used to the layout, gave a clean answer. It didn't help their typing at all, but it did ease the visual emptiness people had flagged. Two pieces of evidence pointing at two different truths, and both mattered for the design decision.

A month of comparative testing. 28 people, paired up, four consecutive days each, a 2×2 within-subjects design against the competing keyboard, counterbalanced so device order couldn't skew the results. That's where we saw how performance and confidence actually shifted over time, not just on day one.

The eye tracking footage ended up as video highlights that marketing used almost immediately, a good reminder that research output isn't only for the product team.

Whatfix was a horizontal platform trying to grow past being a mile wide and an inch deep. I came in as the company's first vertical product hire, reporting directly to the founder and CTO, while the company moved from Series C to Series D. My job was to pick one vertical, prove it out, and build the playbook for the next ones.

I ran internal and external discovery to find the customer profile with the most urgent need, and landed on P&C insurance claims teams. From there it was long interviews with claims professionals to understand what actually slowed them down. Training new hires, staying current on shifting regulation, getting accurate intake data the first time. I built the product, Whatfix Smart P&C, directly around that, working with 12 of the top 25 US insurers to shape it.

Half the job. It meant going through the systems they already ran on, not asking them to add another tool. I opened the line's first ISV integrations into Guidewire, Duck Creek, and Majesco, along with SI partnerships with Accenture and LTI, then briefed the analysts and trained sales to carry the story themselves.