Reproducible briefings

Day 3 · 09:30

A 47-tab workbook is not a briefing. Quarto (the successor to R Markdown) knits prose, code and charts into a PDF or HTML that rebuilds when the extract updates. That is the difference between a pack and a system.

The ancestor of this session used metal-band data to teach R Markdown. You will use a monthly commercial brief: campaign ROAS, a handful of sentences, and a table that cannot drift from the CSV. The skill is the same — literate programming — the audience is EXCO.

A document is a file with a header

---
title: "Aurelia Foods — trading update"
subtitle: "West Africa · Month 7"
author: "Commercial excellence, Lagos Business School cohort"
format:
  html:
    toc: true
    theme: [cosmo]
  pdf:
    documentclass: article
execute:
  echo: false
  warning: false
params:
  month: 7
---

Chunks

A chunk is a fenced block. Options in the header control whether it echoes, caches, or makes a figure of a given size. The prose around it is ordinary markdown: headings, bullet lists, links.

#| label: roas-table
#| echo: false
 
library(tidyverse)
campaigns <- read_csv("campaign-roas.csv") %>%
  filter(week <= params$month)
 
campaigns %>%
  group_by(channel) %>%
  summarise(
    spend = sum(spend_ngn_m),
    revenue = sum(revenue_ngn_m),
    roas = revenue / spend,
    .groups = "drop"
  ) %>%
  arrange(desc(roas))

When params$month changes, the whole brief changes. That is the point. No more ‘I updated the chart but not the bullet’.

Prose that earns its place

Inline R lets a sentence stay true:

Search remains the conversion engine at
`r round(search_roas, 1)`x ROAS. TV is not a failed
channel; it is a reach channel at `r round(tv_roas, 1)`x.
We will not rank them on one number.

Citations and numbers that do not type-drift

Quarto supports BibTeX. Use it for any external claim (Nielsen, NBS, CBN). For internal claims, the citation is the path: data/campaign-roas.csv, extract 12 August 2026. If you cannot point at a file, it is a rumour.