Case Study — UnitedArts
Digitalizing a theater, one spreadsheet at a time.
Sole product owner and builder for a live theater's entire back office, from a stack of untracked spreadsheets to a platform 150 people across every department now rely on.
One ask: "digitalize our spreadsheets." No brief beyond that.
Divadlo Bravo is a theater running a full production operation, administration, finance, technicians, performers, maintenance, on tens of Excel sheets with thousands of rows, plus physical printed invoices. UnitedArts brought me in as sole product owner and builder, with a deceptively simple brief: digitalize the spreadsheets. No problem statement, no defined scope, just a pile of data and a business that needed it to work.
Invoices got paid twice. Or never.
200–300 invoices a month, tracked by hand, across tens of spreadsheets. None of this came from carelessness, it's what happens when 2 people process that volume entirely manually, with no shared system tracking what had already been paid. Staff resorted to printing invoices just to check them off one by one, and it still wasn't reliable enough to catch duplicates or missed payments. The same lack of structure spilled into everything downstream: technicians' timesheets often didn't match their own invoices or the shift plans they were meant to reflect.
Double payments
Invoices paid twice, or never.
Paper tracking
Printed just to track what was owed.
Timesheet mismatches
Didn't match shift plans or invoices.
Rate confusion
Hourly vs. show rate, unclear.
Interviews with every stakeholder group, plus a full audit of the existing spreadsheets.
With no problem statement to work from, discovery ran on two tracks at once: reconstructing how data actually flowed, and understanding why it was built that way in the first place. The goal wasn't just cataloguing what existed, but understanding why, why a sheet was structured a certain way, why a workaround existed, what each group actually needed to trust the numbers they were looking at.
Spreadsheet archaeology
Years of data, audited line by line.
Parallel interviews
Directors, finance, production, technicians.
Cross-referencing
Mapping how data actually moved between people and sheets.
Chasing the "why"
Not just what existed, but why it was built that way.
Three pillars. One deliberate order.
Pillar 2 held the highest value, but none of it could be trusted without a solid data foundation underneath it first. So the platform was built in this order, on purpose: timesheets and rates first, budget and finance second (the feature that alone cut a full day of admin work per week), and show production last, adding real but comparatively smaller value.
Timesheets, Employees, Rates
The data foundation everything else depends on.
Budget & Finance
Semi-automatic invoice processing, payment, and budget completion.
Show Production
Performances and scheduling, additional, comparatively smaller value.
Two decisions that mattered most.
Of everything designed, these two decisions did the most to make the system trustworthy, because they removed the chance for manual error at the exact points where it used to happen.
Invoice & Budget Automation
Timesheet Entry & Approval Flow
Linked, not typed
Invoices tie to pre-approved timesheet entries, or get OCR-read straight from email.
Pick a role, not a rate
Workers select their work type; the system calculates the correct rate.
Split before approval
Every invoice is split into budgets by item/show before it can be approved.
Attendance = timesheet
A production manager's attendance record auto-generates the entry.
Validated before payment
Bank integration checks all invoice and supplier data before sending payment.
Leaders approve
Department leaders, who actually assigned the work, approve it.
Blocked, not caught later
Amount mismatch on upload is blocked immediately, not flagged later.
No approval, no invoice
Workers can't invoice without an approved timesheet, so they chase it themselves.
Leadership backed the change immediately. Frontline staff took more work to bring along.
The UX evolved through a continuous feedback loop rather than being finalized upfront.
Workshops
Direct sessions with stakeholders.
Prototype feedback
Gathered before anything was built.
Live demos
Showing real progress along the way.
Production testing
Adjusting as real usage surfaced friction.
One less day of admin work. Every week.
There's no pre-launch baseline to compare against, the clearest evidence is what stopped happening. Automation removing manual entry, validation catching errors at the point of input, and a single source of truth replacing tens of disconnected spreadsheets.
The biggest lesson: even more time on discovery, especially agreements and rates.
Discovery felt thorough at the time. But partway through, new requirements around work agreements and rates surfaced that hadn't come up in the early stakeholder conversations, they broke the existing design, and that part had to go back to the drawing board.
Designed the solution based on early understanding.
Around work agreements and rates, not raised in early conversations.
Back to the drawing board for that piece.