Which countries had the most contest winners? And what brands were most commonly offered as prizes?

data %>%
  add_count(Country, sort = T) %>%
  mutate(Country = fct_reorder(Country,n)) %>%
  ggplot(aes(Country, fill=Brand)) +
  geom_histogram(stat="count") +
  coord_flip() +
  labs(x="",y="") +
  scale_fill_manual(values = ghibli_palette(18, name = "MononokeMedium", type = "continuous")) +
  theme(axis.text = element_text(size = 14),
        legend.position = c(0.75,0.3),
        legend.text = element_text(size=6)) +
  guides(fill = guide_legend(ncol = 2, keywidth = 1, keyheight = 1))

From the histogram, we see that nearly half of the winners are from France and about 20% are from the UK. However, with 18 brands, it’s very difficult to answer my second question above. I’m still curious about how the brands are distributed among countries so I’ll experiment with network analysis in the next section.

What is the relationship between countries and brands?

network_data <- data %>% select(Country, Brand) 

simpleNetwork(network_data, 
              fontSize = 14, fontFamily = my_font, 
              opacity = 0.8, linkDistance = 100, 
              charge = -10, zoom = F)

This diagram certainly helps identify connections between countries and brands, but it’s not ideal.

Can I build a cleaner visualization of this relationship?

I think a Sankey diagram would work well here, but I’ve never successfully built one before.

# building dataframes to help construct the Sankey diagram
nodes <- data %>%
  count(Country, sort = T) %>%
  mutate(node = seq(0, n_distinct(Country) - 1)) # this basically numbers each row 0 to 7

targets <- data %>%
  count(Brand, sort=T) %>%
  mutate(node = seq(n_distinct(data$Country), n_distinct(data$Country) + n_distinct(Brand) - 1)) # numbers rows 8 to 25

links <- data %>%
  select(Country, Brand) %>%
  left_join(nodes, by = "Country") %>%
  left_join(targets, by = "Brand") %>%
  select(source = node.x, target = node.y ) %>%
  mutate(value = 1) %>%
  arrange(source)

names <- data.frame(node = seq(0, n_distinct(data$Country) + n_distinct(data$Brand) - 1)) %>%
  left_join(nodes, by = "node") %>%
  select(node, Country) %>%
  left_join(targets, by = "node") %>%
  mutate(name = if_else(is.na(Country),Brand,Country)) %>%
  select(node, name) 

# create sankey chart using NetworkD3
sankeyNetwork(Links = links, Nodes = names, Source = "source",
              Target = "target", Value = "value", NodeID = "name",
              colourScale = JS("d3.scaleOrdinal(d3.schemeCategory20);"),
              fontSize = 12, fontFamily = my_font, nodeWidth = 50, nodePadding	= 2)

I’m glad I put the extra effort into learning how to create this chart. I definitely learned a lot!

I think the simple Sankey diagram above does a very nice job of clearly visualizing the distribution of winners by country and brand. We clearly see that most of the winners were from France, the UK, and Belgium. We also see that products from Louis Vuitton, Omega, Carl F Bucherer, and Bovet were among the more popular brands offered as prizes. Additionally, brands that were offered more than once were distributed equally among France and the UK, with the exception of Louis Vuitton, which was distributed among France, the UK, and Belgium.