Xtravu Play • DS Techworks Solutions • 2025 to present
Replacing a signage system we could not trust with one we control
DS Techworks sells signage hardware, and for years it ran on Xibo, a free content system we did not control. Screens went black without warning and every failure landed on our support desk. How could we replace it with something a café owner could run in three steps and a restaurant chain could run on hundreds of screens?
Situation
A system we did not own kept taking our screens down
Hours before a major launch event, the screen one of our FMCG clients was counting on stopped playing. Nothing in Xibo explained why. We put a technician on emergency travel to reset the device on site, at significant cost.
Xibo on the left, the tickets it produced on the right.
That evening was the expensive version of a weekly event. I sorted 412 support tickets from the previous six months by cause:
- 38%Screen went black or content stopped
- 24%Device silently went offline
- 21%Could not find, assign or schedule a layout
- 11%Upload failed and had to be redone
Almost nobody used Xibo's depth. Around 94% of clients only ever did three things: add a screen, build a playlist, push it live.
Task
Replace it, without making signage harder
The goal
A signage platform a first-time user can run in three steps, on software we control end to end, so a black screen is something we can see coming and fix ourselves.
- Reliable on our own hardware
- Three steps for a first-time user
- Flexible enough for a chain
- No rise in hardware price
I owned research, scope and interaction design, and ran sprint planning and engineering handoff as associate project manager.
Action
Four decisions that shaped the product
1. Asked clients twice, and learned more the second time
We invited 20 clients before anything was built, and 11 came. They said they wanted different content at different times. Nobody, including me, asked how a schedule actually reaches them. With a working build, 19 came back, and the scheduler is where they got stuck.
- 11 of 20attended round one, before anything was built
- 19tested the working build in round two
- 7 of 19set a schedule correctly on the first calendar design
2. Cut the product to three jobs
I made the three things clients actually did the entire first version. Everything else Xibo offered was left out on purpose. A school noticeboard and a café were the first two pilots.
The same three jobs on a phone.
3. Rebuilt scheduling around rules, not dates
The first scheduler was a calendar: pick every date you want. Of 19 clients, 7 managed it, 8 needed help and 4 gave up. One, from Burger Singh, needed an offer every Wednesday until December and would not select each Wednesday by hand.
- 7set it correctly
- 8needed help
- 4gave up: the calendar could not express what they needed
Before: pick dates one by one. After: state the rule once.
I expected the rule builder to replace the calendar. It did not. Clients with one-off dates like Christmas kept picking dates, and every restaurant client, with several schedules a day, moved to rules. Both stayed.
What restaurants were actually asking for: the same screen, different menus by daypart.
4. Added Applets for everything that is not a playlist
SCERT asked for a staff-room noticeboard built from PDFs people could tap open. That did not belong in a playlist, so I proposed a separate category for anything that is not media playback. Live streams, always-on dashboards and webhook triggers later arrived the same way, without touching the three-step core.
And fixed our own black screens
Owning the platform meant inheriting its failures. Displays went black from cached media filling storage, and a live news feed filled 2GB on a Jaipur LED wall in 23 minutes. The team added a watchdog to purge old data and had the wall back within the hour.
Result
From 7 of 19 to 19 of 19, and a platform we can fix
- 19 of 19clients set a schedule unaided on the revised design, up from 7 of 19
- 122clients moved from Xibo to Xtravu Play
- ~180new clients, including Mithaas, Biryani By Kilo and McDonald's
- ~5,000screens running on the platform
- 80%fewer support tickets, from about 250 a month to about 50
- Oct 2025first version launched
- ~300clients in total
Some of the clients now running on Xtravu Play.
Reflection
What I would do differently
- 1Ship the small version soonerWe started in June, designed into September and launched in October. A rough build with scheduling in July would have surfaced the Wednesday problem months earlier.
- 2Ask how the work arrivesClients said "different content at different times" and I took it at face value. One follow-up about how schedules reach them would have shown they think in rules.
- 3Watch for the failure you already hadSilent storage build-up is why we left Xibo. We let it happen again before adding monitoring.
The product team, on Diwali.
