Logging a quiz night in under a minute

I started Quizrundan for my own pub quiz group in Stockholm, and other groups use it too. A group of friends, usually three to eight, logs each quiz night, rates the pub, and works through a round of pubs. The moment that matters is at the table, right after the results are read out.

The spec had four personas, and I added a fifth: Maja, who finds the app in a social post and has no group yet. They aren’t research. The agents use them to check a flow.

One goal came first: logging a night at the table takes under 60 seconds. It does.

Groups never see each other. Only the pub list is shared.

Free tiers only, which ruled out Sign in with Apple.

No photos from pubs or Google. Only one of the 19 pubs had an openly licensed photo, so each gets a drawing.

The agents write the code, so my design work happens in the calls: my answers to their questions, and what I approve, reject, or pick from their reviews.

One call that changed a build: the first build put the Overall rating last. QA flagged it, I moved Overall to the top, and it was rebuilt the same day.

The night: the pub, with tonight’s quizzes first, the date, who came, and the team name. A planned quiz fills most of it in.

Ratings: one score for the whole team, Overall first, the score the ranking counts. Every category can be skipped.

Result: placement, number of teams, and points. Anything left out can wait as a draft.

Saving ends on a screen that reacts to the result and suggests three untried pubs, so the next one is picked at the table.

One rating per night, agreed at the table. A rating from each person would turn a quick decision into a survey.

No league between groups. Stats ranks the group’s own pubs by its own ratings.

A group picks its own round. The spec gave every group the same 19 quizzes, so a group that finished had nothing left to aim for. Since 13 October a group can pick its own pubs, and since 16 October a theme, the original 19 among them.

I named it Quizrundan, “the quiz round”, with a nod to doing the rounds of every pub. The spec set the voice: the funniest friend in your quiz team, quick, warm, and a bit cheeky. The base is ultramarine, and from a studio’s site I picked flat colour blocks and two type voices. Sunshine yellow on that blue read too much like the Swedish flag, so blossom pink became the accent.

After the first release the app felt grey to me and short on personality, with no clear main action. I picked one per screen, booking the next quiz first. I tried DynaPuff for headlines and Inter for text, and kept both. DynaPuff is playful like a pub quiz, gives the app personality, and suits the drawings.

Later that day it still felt dated to me, and I asked why. A review found the brand on about eight of 60 screens. I approved a new direction, with the round as a board of 19 stamps on Home. Visited pubs get a check, and every pub keeps its full colour.

In dark mode the blue head on Home was too bright, so I asked for every block one step darker.

The drawings follow the style of Okra Livia, who is in my quiz group and made the references. An illustrator agent draws them in SVG, some from sketches I made with AI. I wasn’t happy with the first set, and the rounder second set was worse: the first set’s squared shapes suit the app and SVG better. I approved a sample of six on 19 October. The same day the illustrator agent redrew the mascot, portraits, avatars, group marks, empty states, and app icon from simple shapes. They ship with the next release.

The spec planned four releases. By 11 October the second had 134 tasks and a 1,050-line questions file. I asked for a set way of working. The answer was a roadmap of briefs, three product roles to shape them, and releases named by date.

Quizrundan is live.

The first release came on 6 October, the day after the first commit. The latest, on 16 October, added a pub map.

Feedback from quiz groups shaped the pub list and the map, but most changes came from using it myself at quiz nights.

It also pushed my own library. By 16 October the app had filed 155 entries for Zabi Components: missing pieces, bugs, and requests.

I’d plan by briefs from the first day, not by four releases. A brief is small enough to say yes to before it’s built. A release that big stopped telling me what was in flight.