Interactive pulse dashboards

Day 3 · 11:15

A dashboard is not a gallery of every KPI you have. It is a one-screen argument with filters. Shiny (and Posit Connect, Streamlit, Power BI) are delivery vehicles. The grammar you learned yesterday still decides whether the screen tells the truth.

The ancestor of this session was a live-coded word cloud. Useful as theatre; rare as a management system. You will design a supply-chain pulse: OTIF, lead time, defect ppm, one screen, three filters (supplier, lane, last n weeks).

The one-screen rule

1. One question (*Are we leaking reliability, and where?*).
2. One scatter or violinbox that answers it.
3. A small table of the ten worst lanes.
4. Filters that a director can drive with a thumb on a phone.
5. A timestamp of the extract, in the footer, always.

A Shiny skeleton

library(shiny)
library(tidyverse)
 
otif <- read_csv("supplier-otif.csv")
 
ui <- fluidPage(
  titlePanel("Supplier pulse"),
  sidebarLayout(
    sidebarPanel(
      selectInput("supplier", "Supplier",
                  choices = c("All", sort(unique(otif$supplier)))),
      sliderInput("weeks", "Last n weeks", 4, 24, 12)
    ),
    mainPanel(plotOutput("scatter"), tableOutput("worst"))
  )
)
 
server <- function(input, output) {
  dat <- reactive({
    otif %>%
      filter(week >= max(week) - input$weeks + 1) %>%
      { if (input$supplier == "All") . else filter(., supplier == input$supplier) }
  })
  output$scatter <- renderPlot({
    dat() %>%
      ggplot(aes(lead_days, otif_pct, colour = supplier)) +
      geom_point(size = 3) +
      labs(x = "Lead time (days)", y = "OTIF %")
  })
  output$worst <- renderTable({
    dat() %>%
      group_by(lane, supplier) %>%
      summarise(otif = mean(otif_pct), .groups = "drop") %>%
      arrange(otif) %>%
      slice_head(n = 8)
  })
}
 
shinyApp(ui, server)

You will not productionise this in seventy-five minutes. You will leave knowing the shape: inputs on the left, a reactive dataset, a plot, a table of exceptions. That shape ports to Power BI and to a well-made Excel with Power Query — the difference is whether the pipeline is visible.