Working Figure

I spend a lot of time thinking about a fairly ordinary analytics problem: good work is still too hard to turn into something useful for other people.

Not useful in the technical sense. The query can be right, the dashboard can load, the model can be sensible, and the chart can be accurate.

Useful in the practical sense: someone else can look at the work, understand what matters, trust the logic, and make a better decision because of it.

That step is still messier than it should be.

In my own work, the frustrating part is rarely the calculation itself. It is usually the last mile: getting from the analysis to the thing a stakeholder actually sees. The work starts with data, definitions, assumptions, caveats, comparisons, and judgement. By the time it reaches a meeting, it might be a slide, a PDF, a screenshot, a dashboard tile, or a short paragraph of commentary sitting next to a chart.

That translation sounds harmless, but it often changes the work.

A chart can be exported without the context that made it meaningful. A dashboard can answer the metric question without answering the “so what?” question. A slide can communicate the recommendation while losing the connection back to the data. The caveat can survive as a footnote, if it survives at all. The analyst then ends up re-explaining the logic in the meeting because the artefact itself cannot carry enough of the thinking.

I find this frustrating because it is not usually caused by a lack of effort or capability. It is often caused by the shape of the workflow.

The analytics industry has improved a lot of the production side of the work. BI tools are useful for governed metrics, shared definitions, and repeatable reporting. Notebook-style tools are useful when the analysis is still being explored. Code-based visualisation and data app frameworks are useful when the standard dashboard shape is too limiting.

The problem is not that those tools are bad. The problem is that the final version of the work often needs qualities from several of them at once.

A stakeholder-facing piece of analysis might need the freshness of a dashboard, the explanation of a report, the visual care of a custom chart, and the directness of a slide. In practice, those qualities are usually assembled through exports, screenshots, manual edits, and presentation software that has little relationship to the logic behind the work.

That is where the workflow still feels unnecessarily rough.

[IMAGE PLACEHOLDER: A simple diagram showing analytical work moving from data, code, BI, or notebooks into the common delivery layer: presentations, reports, PDFs, PNGs, and embedded visuals. The point is to show that the creation side of analytics has modernised faster than the presentation side.]

The common delivery layer inside companies has not changed as much as the creation layer. Most insight still lands in ordinary formats: a presentation, a report, an embedded visual, a PDF, a PNG, or something that can be shared before a meeting. People still copy charts into decks, export static images from live tools, rebuild analysis manually in presentation software, and use awkward add-ons to connect numbers into tools that were not designed for analytical work.

Each individual step is understandable. The whole workflow is clumsy.

I do not think this should be treated as a soft problem. Communication is not separate from the analytical work if the value of the work depends on someone else understanding it.

The way an insight is packaged affects how much of the original thinking survives. A static export, a live dashboard, a slide, and a written report each carry context differently, and each creates different failure points. A chart with the right comparison, a clear title, and the relevant caveat close by will be understood differently from the same chart pasted into a deck without context.

This is why I am interested in the presentation layer of analytics. Not presentation as styling, but presentation as the surface where analysis becomes usable.

A lot of valuable analysis does not need to become a full dashboard or a custom data app. Often the job is more direct: show what changed, explain why it matters, and give the audience enough confidence to decide what happens next. The final artefact might need to be short, visual, current, and easy to share. It might also need to retain a visible relationship to the underlying data and logic.

Current workflows make that combination harder than it should be.

AI is worth talking about here because it changes the cost of producing analytical material. If queries, charts, summaries, code, and draft commentary become easier to generate, the bottleneck moves closer to judgement, editing, framing, and delivery.

That is where I think analysts should be spending more time anyway.

The practical test for better tooling is whether it removes the steps where analysts currently rebuild, reformat, screenshot, annotate, reconcile, or re-explain work they have already done. Better tools should make it easier to move between data, visuals, narrative, and presentation without constantly breaking the connection between them.

That is the broad area I want to write about here.

Working Figure will be about data visualisation, analytics communication, AI-assisted workflows, open analytical formats, dashboards, reports, presentations, and the practical craft of turning analysis into something people can use. I expect some posts to be technical, some to be more about design, and some to be notes from building and thinking through these problems.

I am writing as both an analyst and a founder. The analyst part is tired of seeing useful work weakened at the point of communication. The founder part is interested in what better tools and formats might make possible.

That is the starting point for this blog: how to make analytical work easier to understand, easier to share, and easier to act on.