NHSBSA DALL Playbook

Within the Data Science team, the work that we conduct aims to tell a ‘story’ to the customer, providing them with the insights that the raw data alone would not.

Output formats

There are a range of outputs that may be agreed with the customer:

written reports
data sets
dashboards
interactive reports
slide decks (always required)
code/models

Output structure

A typical example of how a report could be structured is:

Introduction.
Contents.
Key findings (since some may not have time to read the whole content).
Methodology.
Assumptions, limitations, definitions and uncertainty.
Results.
Recommendations and next steps.
Appendices (if applicable).

Supporting text

Most outputs will feature at least some text. General tips for writing includes:

putting the key findings at the front as people might not read the whole thing
including caveats and assumptions, and for particularly important ones mention them up front
write numbers as numbers, not words
using bold tag lines for emphasis of important points
developing an interesting narrative (sometimes difficult to do)
if appropriate, including recommendations and ideas for further exploration
avoiding sentences longer than 20 words
writing in simple terms aimed at a lay audience, such that even a 12-year-old would be able to understand
if multiple people have contributed text, assigning one person in the final stages to act as an editor and rewrite in a coherent style
leaving the writing of any text that supports visuals such as charts or tables until the team is almost certain they are finalised

A very useful tool for written output is Hemingway Editor. If you can get your text to mostly have no issues in this, you know it is readable!

Slide decks

Generally, it is recommended that most initiatives should have a report, which could be in the form of a slide deck, as well as an additional slide deck used to present the story and any other agreed outputs.

Customer summary

This is to give the customer an easily shared and concise report on the outcomes of an initiative. The summary should:

state the problems or questions the initiative looked at, with examples
give the most important assumptions or caveats, and if appropriate issues encountered
summarise the most important findings
give brief recommendations and next steps
provide a mix of visuals and text, with more text preferred
use technical language, if you think the customer is familiar with it

Short presentation

This is intended to be used by the Data Science team to present to others. It also serves as a quick reference of the outcome of an initiative, helpful to people looking back at it in future.

A good presentation will:

show examples of the why the initiative was done
give important assumptions and caveats which are necessary to ensure understanding
summarise the most important findings
give brief recommendations and next steps
limit the amounts of explanatory text and make the most of visuals
optimal: use animations to help present the content
take around 10 minutes to present (depending on situation)
assume audience is non-technical
focus on up to three findings that you would like the audience to remember (applicable for large initiatives with lots of components and findings)
use a good presentation style such as the newsreader style
have an up to date final slide containing feedback and use cases
give people the opportunity to provide feedback- if you are not getting feedback ‘organically’, reach out to get enough for at least one slide!

There is a course Present with Impact available for NHSBSA staff, so ask your line manager if interested.

Accessibility

If your output is to be available online, it must be tested for accessibility and include (or link to) an Accessibility Statement. Full details are in the internal Data Science Wiki.

Try to create outputs with accessibility in mind; for example using the shiny template.
Don’t test for accessibility too early; do it only once reasonably confident the non-text output is finalised.