Technical judgment · Enterprise
Enterprise performance for large accounts
Boards with large tables were getting slow for the customers who rely on them most.
For large customers, speed is a product feature. Fixing the cause, by drawing and sending less, did more than speeding up individual actions, and it kept working as tables grew.
Why
Enterprise boards increasingly rely on Tables for large sets of structured data. At that scale performance degraded for some boards.
The company-wide effort to make Miro faster gave us the support to fix the underlying architecture, not just patch symptoms.
My contributions
- Made the case with evidence. Conducted user interviews and combined them with what customers were hitting: boards that needed thousands of items and tables that slowed down as they grew.
- Set the priorities and the order. Ranked the work in three tiers. First came smoother panning and zooming on heavy tables, and sending each view only the data it needs. Next came adaptive detail rendering and separating storage from rendering, with the first scoped as a proof of concept so we could prove the approach before committing the team to it. Server-side calculation came last. This order meant each step built on the one before, so the tables could keep scaling technically as customers added more data.
- Aligned the teams and the sponsors. Involved the leaders, the team that owns the board itself and the platform team from the start, and tied the work to the company goal of a fast, smooth board, inside the company-wide effort to make Miro faster, which gave us the backing to fix root causes.
- Drove culling to release. Engineers built it; I was the PM. A large table has thousands of cells, and the browser used to draw all of them, even the ones nobody could see. Culling draws only what is on screen, with a lightweight placeholder for the rest and a simplified outline when you zoom out. Memory stays in check as a table grows, and tables load faster. It now covers tables, timelines and kanban boards.
Who benefited
Customers with the largest boards can keep working smoothly as their tables grow.
The platform team got a clear priority order and a better way to handle incoming requests.
Result
How we measured it: we compared the same set of boards before and after the change. Boards with tables got 41% faster. Boards without tables got 6% faster over the same period, so most of the gain on the table boards is likely down to the table work.
Want to talk about this? Get in touch →