Data Analysis & Interpretation
From raw data to real insight — descriptive, inferential, regression and multivariate work.
Read more
Dashboards, charts and reports that make a finding visible in one look.
A good figure does something no table can: it makes the reader see the finding before they have finished reading the caption. A bad one hides the same finding behind a rainbow of colours and a legend nobody can match to the lines.
Visualisation is not decoration added at the end. The chart type is determined by what you are showing — comparison, composition, distribution, relationship or change over time — and choosing the wrong one misleads even when every number in it is correct. Pie charts with eleven slices, dual axes that manufacture a correlation, and truncated axes that exaggerate a difference are the three most common offenders.
We build figures for two different audiences. For a thesis or a journal the requirements are strict: greyscale-legible, colour-blind safe, vector where possible, sized for a single column or a full page, with fonts that survive the printer. For an organisation the requirement is different — an interactive dashboard that answers the five questions people keep asking, updated from a live file.
The six areas on the poster cover both: data collection and management, exploratory analysis, visualisation and dashboards, statistical analysis and modelling, machine-learning insight, and actionable reports with recommendations.
From a single publication figure to a full working dashboard.
The most important property of a figure is that it does not overstate. That means starting a bar axis at zero, showing the spread and not only the mean, marking sample sizes where groups differ in size, and displaying confidence intervals or error bars whenever an estimate is being compared with another estimate.
It also means resisting the temptation to add a third dimension for effect. A perspective bar chart makes values harder to read, not easier — which is why the 3D illustrations on this website are used for decoration and never for reporting your actual results. Your figures are flat, plain and legible.
Where a relationship is genuinely uncertain, the figure should show that. A scatter plot with a fitted line and a shaded confidence band tells the truth about a weak association far better than a bare trend line, and reviewers respond well to authors who present their evidence without inflating it.
What decision or claim does this figure need to support? Everything else follows from that answer.
Before choosing a chart we look at distributions, outliers and missing patterns, because these determine what can honestly be shown.
Two or three alternatives for each finding, so you can see which one communicates it fastest.
Palette, labelling, axis treatment, annotation and sizing for the medium the figure will actually appear in.
Where the work is ongoing rather than one-off, with a refresh path from your source file.
Editable source files, the export in the format you need, and a short session on how to update it yourself.
Most dashboards fail for the same reason: they show everything the data contains rather than the handful of things somebody actually decides on. The first conversation we have is not about charts, it is about who opens this and what they do differently after looking at it. If there is no answer to that, the dashboard should not be built.
Once the questions are clear the layout follows: the headline number where the eye lands first, the trend beside it, the breakdown below, and the detail available on demand rather than displayed by default. Filters that people will actually use, and none that they will not.
We also make sure it can be maintained. A dashboard that only works while its author is available is a liability. You get the source file, the data-preparation steps written down, and a walk-through so that someone in your team can keep it running.
| What you receive |
|---|
| Vector figures for thesis and journal submission |
| Raster exports at the resolution your publisher requires |
| Editable chart source files |
| Interactive dashboard with documented refresh steps |
| Data-preparation script or workbook |
| Style sheet so future figures match |
| Written commentary on what each figure shows |
| A handover session for your team |
Most ask for 300 dpi for photographs and 600 to 1200 dpi for line art, or vector EPS and PDF, which is better because it never pixelates. We supply vector by default and raster exports at whatever the specific journal's guide requires.
Yes, and it is worth doing even for colour journals because readers print. We distinguish series by shape, line style and fill pattern as well as colour, and we check every figure in greyscale before handing it over.
It can, if the data has a stable source — a shared workbook, a database or an export that always looks the same. We set up the refresh, document it, and show you how to fix it if the source format ever changes.
Often. Screenshots of SPSS or Excel default charts are the most common reason a submission gets sent back. Send us what you have along with the underlying numbers and we will rebuild them properly.
Yes, and it is worth doing. We build a style sheet — fonts, sizes, palette, line weights, axis treatment — and apply it to every figure so the document looks like one piece of work. You keep the style sheet for any figures you make later yourself.
From raw data to real insight — descriptive, inferential, regression and multivariate work.
Read more
Reliable analysis, accurate results — from raw responses to a defensible results chapter.
Read more
Bibliometric mapping that turns a literature pile into a defensible research gap.
Read moreSend your topic, your dataset or one draft chapter. We will tell you honestly what it needs — before you pay anything. The first consultation is free.