How charts work

Charts sit after the question and the finding. They show observed or derived values that already exist in the pack. They do not invent a trend, a midpoint, or a success score.

Every chart keeps its title, geography, period, unit, source, evidence basis and one limitation. Titles sit below the section heading. Plot labels and values use caption size and wrap in full. Time series fill the plot width. Axes show the years and values in between the ends, not only the first, middle and last mark. Tap, drag across the plot or use the arrow keys to move a hairline, the selected value is shown above the chart, not as a hover key elsewhere on the page. Crime, news and map venues stay descriptive. Register-survival probabilities stay withheld.

How a chart is chosen

The default chart is chosen for the metric. On a time series you can switch Line, Columns and Table. Columns stay available only when there are few enough periods to read; a dense monthly series stays a line. Category charts keep Chart and Data table. Area is not offered, because a filled band can look like interpolated values. Long category labels use a horizontal bar. Two published years across several sectors use a slope or dumbbell only when every row shares those endpoints. An illustrative linear trend stays hidden until you ask for it. A 100% stacked bar appears only when the parts form the published total. A heat map is optional and row-normalised, colour cannot be compared between rows.

Comparisons

Values are compared only when period, geography, SIC and denominator match. Otherwise they are shown separately. Rate differences are percentage points, not “net growth”.

Guest preview

Without an account you see observed facts, an evidence-coverage matrix, and a bar chart only when two or more facts share a unit. Charts, history, maps, saves and decision cases stay in the full report.

Signed-in reports

Business idea and place reports add sector bars, a common-period slope when endpoints match, and ONS openings versus closures.
Property reports pair annual price and volume on the same years, drag, tap or the period control keep both charts in step, without a second y-axis. HPI stays on its own index line. Planning counts are a time series you can read by quarter or by complete year, not a funnel.
Company reports keep financial facts in a table unless two comparable filed periods exist.
The place map is a point map at local-authority resolution with a stated radius. It is not a choropleth of England.

City profiles

Public city pages show a horizontal bar of leading SIC sections. The city list compares registered-business density with a lollipop from zero. That comparison is not a ranking of where to invest.

Decision cases

Closing cash is a line per named alternative. Sensitivity uses horizontal bars of effect size. Stored-run comparison uses a dumbbell. Uncertainty can show a p5–p95 band with a p50 marker when those percentiles exist. Alternatives are not ranked.

What charts must not do

Combine unrelated dimensions into an opportunity, attractiveness or demand score.
Treat crime, news or venue pins as inputs to a decision score.
Fill a map as if street-level demand were known.
Mix price and transaction volume on one dual axis.
Use pie, donut, gauge or radar charts as the default vocabulary.