Show me the thing I came for.

One dataset, five views, linked

Select an account in the graph and the timeline shows its transfers, the map shows its counterparties and the grid scrolls to its rows. One selection model across four surfaces, with no synchronising code in between.

Community detection over a payments network

Louvain over the transfer graph, with the community driving colour. A structuring ring is a cluster that moves money between its own members more than it moves it anywhere else, and the clustering finds it before anyone writes a rule for it.

Betweenness, to find the intermediaries

Centrality ranked and mapped straight onto node size. The account that sits on the path between two clusters gets bigger, which is the one worth a subpoena.

Min-cut on a layering chain

The smallest set of transfers whose removal disconnects the source from the destination — the chokepoint, found rather than argued about.

Money mule and structuring detectors

Six domain detectors over the same graph: mule accounts, structuring, velocity, cycles, beneficiary chains and an aggregate score. Part of the finance domain pack.

Shortest path across an asset graph

Given an entry point and a crown-jewel asset, the route between them. The path is the thing to break, and it is a query rather than a meeting.

Attack paths at WebGL scale

An enterprise asset graph is a WebGL graph. Same API, same layouts and same events as the thousand-node view the analyst started on — only the renderer changed.

Analysis-driven styling

Run the algorithm, then style from the answer: styleBy maps a result straight onto size, colour or visibility with a 400ms transition. No server round trip, so the analyst can change the question without waiting.

Structural layout on an unfamiliar network

Groups nodes by the role they play in the topology, so hubs sit with hubs and leaves with leaves. It is how you see the shape of a network you have never seen before.

Lens layout

Magnifies around a focus node and compresses the rest, so context stays on screen while the detail is readable.

Tweak layout under a live feed

Re-runs the layout while keeping existing positions stable. When the data updates, the user's mental map survives it.

Combos: group, expand, drill down

Collapse a set of entities into one node and open it again. The grouping is computed over your data rather than baked into it, so a different grouping is a different call and not a different dataset.

Incremental updates

Nodes arrive on a slow feed; links animate in and the layout accommodates them. Beside it, the same feed with redraw instead of animate — which is visibly worse, and is the argument.

Renderer switch: SVG, Canvas, WebGL

One graph, three renderers, switched in place. The DOM element count falls from tens of thousands to one as you cross from SVG to Canvas, which makes the point faster than a paragraph.

Metabolic pathway layout

A pathway is not a force-directed graph that happens to be biological. This layout exists because the work asked for it.

Kinship tree and dendrogram

Lineage where the relationship has a direction and a generation, drawn as a tree rather than settled by a force — and a dendrogram for hierarchical clustering results.

The chart wall: one dataset, many idioms

The same underlying figures drawn as twenty different chart types, each real, each reading the same data shape and taking the same options. Learn the grammar once.

Sankey and chord flows

Where the volume went, and between whom. Flow charts that share a theme with the graph beside them.

Candlestick and OHLC

Financial series drawn properly — bodies, wicks and a price scale — in the same package as the graph engine rather than as a second vendor's component.

Small multiples across twelve regions

One chart per category on a shared scale, which is the honest way to compare twelve things and much better than twelve lines fighting over one axis.

Theme tokens: your design system, not ours

Point the charts at your existing colour, spacing, radius and typography variables and they stop looking like a charting library. Light and dark are two token sets, not two code paths.

Zoom from a century to a second

One continuous gesture across the whole range, with the detail level following it — aggregated bands when zoomed out, individual events with labels when zoomed in.

Multi-track: twelve accounts aligned in time

One track per entity, sharing an axis. An investigation of twelve accounts is twelve rows, and the sequence is the story.

Playback across linked views

Step a cursor through time and watch the graph, the map and the grid follow it. This is how a sequence of transfers becomes a narrative.

Clustering by zoom

Points collapse into counts as you zoom out, so a dense port is one marker labelled 4,182 rather than 4,182 overlapping markers.

Edge bundling on flows

Parallel routes grouped into visible corridors. Bundling runs in projected world space, so it stays correct at every zoom instead of re-bundling on each pan.

Tracked movement with trails

Interpolated strictly between known samples, with a trail behind. We do not invent a position beyond your data, because on a map an extrapolated position is a false claim about where something was.

Offline tile pack, no internet

A vector basemap drawn from a tile pack shipped with the application. We wrote the protobuf reader, the tile decoder and the renderer, so a basemap adds no dependency and needs no tile server.

Linked selection: graph to grid

Click a node and the grid scrolls to its row. Filter the grid and the graph dims what the filter removed. There is no synchronising code because there is nothing to synchronise.

Grouping, pivot and in-cell sparklines

Row grouping with aggregation, multi-column sort, an advanced filter builder and sparklines in the cells — sharing a theme with the charts beside them.

Annotations for a briefing product

Narrative annotation over the chart, then export to PNG, SVG or PDF. Vector export that is really vector, because a briefing pack is read at whatever size the reader chooses.