Hej! Let's talk.

Product Design Lead in Copenhagen. Happy to chat about work, design or a good coffee spot.

Stefan Cioroiu's resume: Product Design Lead with over 10 years of experience, based in Copenhagen. Senior Product Designer at Veo Technologies, earlier Senior Product Designer at Novo Nordisk, Senior then Lead Product Designer at Planday, and Product Designer at Tradeshift, Founders and Zigna, starting as a UI and graphic designer at Me-Mover. Studied at KEA, Copenhagen School of Design and Technology.

6 min read Web app · Data design · Onboarding · Launch

Analytics 2: The 30-Minute Coach and Proactive Team Setup

Veo’s AI reads every player’s stats and shirt number. But a number isn’t a name. I designed the roster, lineups and jersey linking that turn “#7” into a player, for coaches with thirty minutes a week.

Context
Veo • Web / Analytics
Role
Senior Product Designer
Scope
Roster, lineups, jersey linking
Status
Shipped Summer 2026

Background: Veo’s Analytics 2 film, by Veo’s brand team

The problem

The stats need a name. Nobody had set one up.

Analytics 2 promises every player their own numbers. The camera only sees shirt numbers, and turning “#11” into Brayan needs a roster and a lineup. At launch, most teams had neither.

The work falls on coaches with a day job. So: how do we get the thirty-minute coach to do the setup that unlocks everything else?

Needs a roster and lineup: who is #11?Per-player stats, tracked from video
The promise, from Veo’s launch film. The name in that card only appears if the team is set up.
  • Feel like coaching, not admin: a pitch and a formation, not a form.
  • Do it once: a default lineup on the team, reused for every match.
  • Never show wrong numbers to cover for missing setup.
  • Ask at the right moment, where the missing data hurts.

Shared words

First, agree on what a lineup is

Product, engineering and data each meant something different by “lineup”. We agreed five terms before drawing a screen.

Roster
Everyone on the team. At grassroots level, jersey numbers change from week to week.
Starting lineup
The eleven on the pitch at kick-off, for one match.
Default lineup
Saved on the team, so it isn’t redone every match.
Formation
The shape, like 4-3-3. You can set one without naming any players.
Snapshot
A roster snapshot per match, so old matches stay as they were.

The setup flow

From empty to 11 of 11

The shipped flow, one decision per step.

Team default lineup empty state with a Get started buttonThe formation dropdown grouped by 11, 9, 8, 7 and 5-a-sideAdd player with two picker options side by side; the chosen one lists players not in the lineup firstMember details with a jersey number field and a warning that the number is takenChanging the primary team captain, with a warning that past match data won’t be affectedA complete 4-3-3 team default lineup
1 / 6

The lineup builder

Setup that feels like picking a team

Coaches think in positions, so the lineup is a pitch, not a form. This is my tested prototype: drop players onto the pitch.

InteractiveDrag players onto the pitch. Tap an empty spot to add someone. Try another formation.

Players in the team 24

Drag players onto a position to complete the lineup

The formation

0/11 positions completed

Data you can’t fully trust

Three kids, one number seven

Unlinked jerseys can silently merge different players across matches. With engineering, I weighed the options on trust, adoption and launch scope.

InteractiveSwitch between the options and watch the #7 row.

What the camera recorded

MatchJerseyLinked toDistance
1#7Leo Hart9.1 km
2#7Sam Ortiz7.2 km
3#7nobody yet8.0 km
1–3#9Jonas Weber8.4 + 8.8 + 8.1 km
1–3#20Kofi Mensah10.2 + 9.6 + 10.5 km
3#14nobody yet6.9 km

What the coach sees · distance, last 3 matches

    On the whiteboard first

    Start from the player, or from the jersey

    Coaches don’t always know who wore what, but the video does. A spot can start from a player, a detected jersey or a typed number.

    Every conflict gets a plain warning before anything changes.

    One player, every state

    A face, a number, or an open spot

    Player Avatar, from our design system, draws a player the same way everywhere.

    InteractivePick who it is, change the kit and marks. Hover the avatar, click it to select it.

    Completed
    Content
    Size
    Number
    Team
    Marks
    Substitution
    With Label
    State

    Move it upstream

    Prevent unknowns instead of cleaning them up

    Instead of cleanup tools after the match, I pushed to catch missing setup where it’s cheap to fix:

    On the stats page

    Ask where the gap hurts

    Coaches notice missing names in their stats, so we ask there: one sentence, one button to the roster.

    The team stats page with a banner: Get more accurate data. Complete roster, or dismiss.
    The nudge on the team stats page

    When a player claims their number

    Fix it once, at the right scope

    Linking a number asks how far the fix reaches: this match, from now on, or all past matches, with a warning for rewriting history. It works like “claim jersey” in the player app.

    InteractivePick a scope and see which matches change.

    Link #7 to Leo Hart

    Prototyping the hard cases

    Merge, split and roll back

    Diagram: merging two jersey tracks, only the parts that don't overlap move
    Merging #2 into #1: only the parts that don’t overlap move
    Diagram: rolling back a fix also clears every fix made after it
    Rolling back: undoing a fix also clears every fix after it

    Launch

    Shipped with Analytics 2, summer 2026

    Team setup launched with Analytics 2 and the new Match Report, which I reviewed. Lineup setup shows at 0:08 in the launch film, the roster at 0:16.

    Veo’s launch film for the Match Report and player stats, by Veo’s brand team.

    Outcome

    Coaches did the setup

    6.5×

    completed lineups per week vs. before launch, and it held

    4.6×

    coaches building lineups each week

    100×

    views of the Lineup tab vs. before launch

    Relative numbers, setup only. Correlation, not proof.

    Looking back

    Honest data earns more trust than full data

    The big call was showing less: no number looks more certain than it is. Clear “unassigned” rows and nudges kept that honest view useful.

    Next time I’d agree on success metrics before building. Tying the signals to revenue took too long.

    Thanks for reading!

    Or see all projects