Designing a feedback loop people would trust

At Mixbook, customer feedback was painstakingly gathered through cumbersome one-off surveys and collected in spreadsheets, then distilled into a superficial handful of positive and negative comments for a Monday meeting. On the Mosaic team, direct customer feedback was the lifeblood of the product. It revealed needs clearly enough to shape our next product decisions.

The tools available at the time made that connection difficult. Surveys were long, deployment was manual, response rates were poor, and the questions often reflected what a company wanted to quantify rather than what a customer wanted to say.

My two co-founders and I started Delighted around a simple idea: make feedback dramatically easier to collect, make the experience good enough that people would answer, and deliver their words directly to people who could act.

The initial flow asked for only what was necessary: make the survey recognizable, add the people who should receive it, verify the import, and send. Advanced paths remained available through properties, the API, and scheduled delivery without turning setup into a configuration project.

The company paying for Delighted needed useful information. The person receiving a survey needed an experience that felt respectful and worth completing. As Delighted grew, we also had to protect the channel itself from fatigue and spam reports.

One of the most persistent requests was the ability to send reminders to people who had not responded. We declined for years. Reminders could increase a single survey's response count while making recipients less willing to answer future surveys. That was a poor trade for customers, their customers, and Delighted.

We also throttled each email address across surveys and organizations within an account so no person could be over-surveyed by accident.

Larger customers regularly asked for legal language, embedding options, and more visual control. The underlying system supported far more variables than the primary interface exposed. We handled exceptional needs case by case, then watched to see which patterns repeated.

When a custom capability repeatedly proved useful, it could become a first-class product. Website visitor surveys followed that path. Other capabilities remained behind feature flags because making them public would have complicated the product for everyone else.

The three founders jointly owned roadmap and feature priorities. We classified incoming feedback inside Delighted using tags and sentiment, which made repeated needs visible. Once we chose a problem, I explored several design directions, we selected an approach together, and then reduced it to the smallest useful release.

Delighted did not grow primarily through conventional marketing. A small “Powered by Delighted” link appeared at the bottom of each survey. Someone would receive a thoughtful survey from a company they respected, click the attribution, create an account, and begin sending their own.

That loop meant recipient trust was not a secondary concern. The quality of the survey experience directly determined the quality of our acquisition channel.

This customer case study followed Bonobos from the craft of making clothing to the moment a customer answered a Delighted survey and the team read that response. It made the abstract feedback loop tangible from both sides.

We hosted a customer-success dinner for existing customers with Fred Reichheld, creator of the Net Promoter System and a Bain Fellow at Bain & Company. The goal was simple: bring thoughtful customer leaders together to compare how they worked in practice.

We also held a separate fireside chat. I designed every part of both experiences, from menus and printed collateral to notebooks, laser-cut badges, packaging, and the takeaways guests brought home.