What it does today.

The Sifter beta is focused on the first hard problem in any data project: getting the right data, in the right shape, without weeks of plumbing.

The pitch on sifter.com says it plainly — describe the dataset you want and Sifter builds it. Under the hood that means natural-language and structured workflows for collection, transformation, and management — producing artifacts you can read, fork, and trust.

  • Describe, don't assemble. Tell Sifter what you need; review and refine the dataset it produces.
  • Versioned, forkable datasets. Each dataset is real, versioned data — branch it, fork it, share a snapshot, or roll back when something changes.
  • Cell-level lineage. Every value in every dataset traces back to its source. No black-box numbers, no trust-me data — just the receipts, down to the cell.

Where it's going.

Data collection is step one. The longer-term goal for Sifter is an end-to-end workspace that takes a question all the way to the story you tell with the answer.

  1. 01

    Collect & manage

    Build the right dataset, keep it fresh, and trust where it came from. Available in beta

  2. 02

    Analyse

    Explore, model, and pressure-test the data with a query layer that keeps analysts moving without the usual platform tax. Next

  3. 03

    Visualise

    Charts, dashboards, and exploratory views — built on the same runtime so they stay in sync with the data underneath. Next

  4. 04

    Publish

    Interactive publications that turn analysis into narrative — so the people you're trying to convince can play with the data, not just squint at a screenshot of it. On the roadmap

The runtime underneath.

Sifter is built on a runtime we designed from scratch — a WebAssembly client that does the heavy lifting in your browser, and a backend service that handles the work that has to happen elsewhere. We chose to build it ourselves because the existing options forced trade-offs we weren't willing to make: either fast and locked-in, or flexible and slow.

That runtime is what makes the rest of the roadmap possible. Each phase — collection, analysis, visualisation, publication — runs on the same foundation, so a dataset you build today flows straight into the chart you'll publish from it tomorrow.

It's also what makes Sifter more than a single product. Over time, we expect the same runtime to back custom analytics platforms built on top of it — not just Sifter itself.

Built by Spillsoft — a small studio in Airdrie, Alberta.

Want a look at Sifter?

We're letting people into the beta as we go. If you've got a dataset problem that fits, we'd love to hear about it.