Reports#
Reports and dashboards are overlapping tools/endpoints which share many elements.
Which to make and what to call your end product is dependent on the project and organisation.
Dashboards vs Reports#
Dashboards are often updated live
Mostly graphs, plots, etc.
Key performance indicators (KPIs)
Current trends, accumulation in minutes, hours, days, weeks, …
Reports can be:
Summaries of activities, economy, etc. up to a point
Investigations into a subject
Suggestive of future path or action
Often printable, but digital reports can contain interactive graphics
(ChatGPT)
Reproducibility#
In many projects, e.g., those based on Data-Driven Desision Making, transparency, traceability and reproducibility are important.
Ideally, this means being able to recreate report contents at any time.
Markdown, Jupyter Notebooks, RMarkdown and similar frameworks are popular open alternatives for such reports.
Power BI Report Builder is the Power BI equivalent with options similar to Power BI Desktop, but optimised for paginated output, e.g., PDF without interactivity.
Redo datahandling, calculations and visualisations at the time of rendering/compilation.
Underlying data must be available, possibly using a local copy.
Copy-pasting from various sources is a no-go.
Version control can be used to trace any changes to data, code or text.
Even though the project may continue and the report evolve, a snapshot from a certain timepoint is always available.
Graphics#
“A picture is worth a thousand words” is an old cliche, but still valid.
Eye catching, yet informative graphics with descriptive captions can carry most of the message in anything from a glossy paper brochure to a research article.
In a report, text is inevitable as context, background, interpretations and conclusions are hard to convert to graphics that are universally understandable.
A digital report can retain some interactivity, e.g., Plotly interactions.
With some effort, also more advanced interactivity can be included, e.g., in an .html formatted report or an online report in services like Power BI.
Interactivity can engage the reader more, partially by just being more fun, but also because it feels more tangible and personal.
Numbers#
Numbers can be communicated in various ways depending on the type and function.
Key Performance Indicators (KPI) are single figures that show some important summary, e.g., gross income, increase in sales since last period, a competitive index, etc.
These are typically highlighted in a report possibly grouped together and placed visibly.
Tables can be needed for increased precision or as background material for graphs.
Formatting can highlight what we want the reader to focus on.
Pivoting, grouping and summarising can give a much easier interpretation than large tables.
Audience#
Do not overestimate your readers!
Simple graphics and simple messages are most likely to have an impact in most situations.
The higher up in an organisation a report is aimed at, the less time one can expect will be used at interpretation (boldly speaking).
(ChatGPT)
Background knowledge and technical expertise is likely to be different from your own.
Footnotes, linkes to sources etc. are useful if technical terms must be used.
…
Consider cultural background and language fluency when writing.
Relatively short sentences.
Avoid using:
artifical or fancy words, e.g., foreign words to appear interesting
slang that can be misunderstood or not understood
Consider carefully the use of buzzwords, e.g., “synergy”, “big data”, “data science”, “prompting”, as overuse makes text sound clichéd.
This is especially important if the report is supposed to be used for many years as terms become obsolite quickly, especially in the tech business.
Layout#
The perfect theme for a report does not exist and is a moving target.
Looking back a couple of decades, sharp division using lines and strong background colours were popular.
Organisation and grouping using alignment, white space and soft colours is more typical as of 2024.
“Air”, i.e., open surface around elements, is less stressful than an over-crowded page.
Use paragraphs to group sub-themes in the text and keep them short.
Tools/software#
The above mentioned tools are just examples. The possiblities are virtually endless.
A surprising amount of business reports are made as presentations, e.g., PowerPoints, or PDF exports of presentations, thus violating or stretching the reproducibility principles.
Spreadsheets, e.g., Excel, and text documents, e.g., Word, are also popular.
If the topic and demands are simple, do not overengineer the solution.
AI usage#
Automised report generation is possible in some frameworks.
Reproducibility might not be conserved.
AIs will hallusinate about topics they do not know well.
Fine-grained control might be reduced/lost.
But the tools are getting better every day.
Julius.ai is an example of a an AI that combines a chat with full data analysis capabilities, plotting, problem solving and report generation.
A compromise, using AI in the process, can:
increase reporting efficiency,
retian reproducibility and fine-grained control,
keep hallusinations in check.