Template
Introduction
Looking at my journey to being hired by LinkedIn, using some less traditional visualisation methods, though hardly breaking the mold here!
library(tidyverse)clean_names <- function(x) str_replace_all(tolower(x), " ", "_")
kable_tally <- function(data, variable, cap = "") {
data %>% group_by_(variable) %>% tally %>% arrange(desc(n)) %>%
knitr::kable(caption = cap, format = "html") %>%
kableExtra::kable_styling(full_width = FALSE)
}
count_words <- function(word, corpus) sum(str_count(corpus, word), na.rm = TRUE)
count_exact_words <- function(word, corpus) sum(corpus %in% word, na.rm = TRUE)The data was generated manually in a Google Sheet as I was tracking all the applications I made over the course of several months.
if(!interactive()){
setwd("..")
}
interviews_raw <- readxl::read_excel("data/job_search_list.xlsx") %>%
select(-Email, -Contact, -Phone)
colnames(interviews_raw) <-
colnames(interviews_raw) %>%
clean_namesCleaning
summary(interviews_raw %>%
mutate_at(c("status", "stage_reached", "contact_gender"), as.factor))## company job_title status stage_reached
## Length:22 Length:22 ACCEPTED : 1 1st Round : 2
## Class :character Class :character INCOMPATIBLE: 3 2nd Round : 2
## Mode :character Mode :character NOREPLY : 8 Final Round: 1
## REJECTED :10 Late : 1
## None :16
##
##
## applied_date contact_gender salary
## Min. :2018-03-02 00:00:00 F : 2 Min. :28.00
## 1st Qu.:2018-03-27 06:00:00 M :13 1st Qu.:29.50
## Median :2018-04-27 00:00:00 UNKNOWN: 7 Median :35.00
## Mean :2018-04-17 04:21:49 Mean :33.25
## 3rd Qu.:2018-05-07 00:00:00 3rd Qu.:35.62
## Max. :2018-05-29 00:00:00 Max. :37.50
## NA's :14
## key_phrases link is_recruiter?
## Length:22 Length:22 Mode :logical
## Class :character Class :character FALSE:10
## Mode :character Mode :character TRUE :12
##
##
##
## glimpse(interviews_raw)## Observations: 22
## Variables: 10
## $ company <chr> "AON", "(Archer)", "First Data", "(Wallace Mye...
## $ job_title <chr> "Data analyst", "DATA ANALYST - ANALYTICS CENT...
## $ status <chr> "REJECTED", "NOREPLY", "NOREPLY", "REJECTED", ...
## $ stage_reached <chr> "1st Round", "None", "None", "2nd Round", "1st...
## $ applied_date <dttm> 2018-03-02, 2018-03-02, 2018-03-02, 2018-03-0...
## $ contact_gender <chr> "UNKNOWN", "F", "UNKNOWN", "M", "UNKNOWN", "M"...
## $ salary <dbl> NA, 37.5, NA, NA, NA, NA, NA, 37.5, NA, NA, NA...
## $ key_phrases <chr> "global, BI Technologies, Python, data mining,...
## $ link <chr> "https://www.irishjobs.ie/Jobs/Data-Analyst-81...
## $ `is_recruiter?` <lgl> FALSE, TRUE, FALSE, TRUE, FALSE, TRUE, FALSE, ...A few things come up from the initial summary:
- Inconsistent formatting of factor levels.
key_phrasesis really a character vector and should be nested.salaryuses inconsistent factor levels.contact_gendershould useNAfor missing values.
interviews_clean <-
interviews_raw %>%
mutate_at(c("stage_reached"), toupper) %>%
mutate(salary = ifelse(salary == "?", NA, salary))
interviews_clean$key_phrases <-
(str_split(interviews_raw$key_phrases, ",") %>%
map(str_trim) %>%
map(clean_names))Exploration
Titles were mostly data analytics roles.
interviews_clean %>% kable_tally("job_title", "Job Title Tally")| job_title | n |
|---|---|
| Data Analyst | 8 |
| Junior Data Analyst | 2 |
| Data analyst | 1 |
| DATA ANALYST - ANALYTICS CENTRE (JNR-MID) | 1 |
| Data Analyst - Dublin North. | 1 |
| Data BI Developer | 1 |
| Data Engineer | 1 |
| Data Scientist | 1 |
| Graduate Software Developer | 1 |
| JUNIOR DATA ANALYST - PHARMA SUPPLY CHAIN | 1 |
| QA Graduate Programme | 1 |
| Robotic Process Automation (RPA) Analyst (036660) | 1 |
| SQL Data Analyst? | 1 |
| SQL Developer (Capita) | 1 |
Mostly even split between explicit rejection and applications where I recieved no reply. “INCOMPATIBLE” means I didn’t meet the requirements for the position.
interviews_clean %>%
kable_tally("status", "Application Status Tally") %>%
kableExtra::row_spec(4, bold = T, color = "white", background = "green")| status | n |
|---|---|
| REJECTED | 10 |
| NOREPLY | 8 |
| INCOMPATIBLE | 3 |
| ACCEPTED | 1 |
interviews_clean %>%
mutate(month = format(applied_date, "%B")) %>%
group_by(month) %>% tally() %>%
knitr::kable() %>%
kableExtra::kable_styling(full_width = FALSE)| month | n |
|---|---|
| April | 7 |
| March | 7 |
| May | 8 |
Analysis
Sankey (or alluvial) plot of how far I got in each round and the ultimate result of the application.
stage_levels <- c("FINAL ROUND", "2ND ROUND", "1ST ROUND", "LATE", "NONE")
status_levels <- c("ACCEPTED", "REJECTED", "INCOMPATIBLE", "NOREPLY",
"CONTACTED","APPLIED" )
interviews_clean %>% group_by(status, stage_reached) %>% tally() %>%
ungroup() %>%
mutate(stage_reached = factor(stage_reached, levels = stage_levels)) %>%
ggplot(aes(weight = n, axis1 = stage_reached,
axis2 = status, fill = status)) +
ggalluvial::geom_alluvium( width = 1/12) +
ggalluvial::geom_stratum(width = 1/12, fill = "white", color = "grey") +
annotate("text", x = 1, y = c(21.5, 20, 18, 16.5, 8), label = stage_levels,
size = 3) +
scale_x_continuous(breaks = c(1, 2), labels = c("Round Reached", "Status")) +
theme_void() +
ggtitle("Round Reached vs Status of Application") +
scale_fill_brewer(palette = "Dark2")## Missing alluvia for some stratum combinations.
## Missing alluvia for some stratum combinations.
A simple timeline of when I applied for each position.
successful_application_date <-
interviews_clean$applied_date[interviews_clean$company == "LinkedIn (CPL)"]
interviews_clean %>%
ggplot(aes(x = applied_date, y = 0, colour = status)) +
geom_abline(slope = 0, intercept = 0) +
geom_point(size = 2) +
theme_minimal() +
theme(axis.text.y = element_blank(),
axis.ticks.y = element_blank(),
axis.title.y = element_blank(),
panel.grid = element_blank(),
axis.text.x = element_text())+
scale_colour_brewer(palette = "Dark2")
When looking at repetitions of keywords the answers aren’t that surprising. Entry level positions are focussed on data sourcing, cleaning and visualisation so there are tools repeatedly mentioned that are good for these tasks. Ability to communicate effectively is also a very highly prized skill, we need to be able to tell people what insights we’ve found!
key_words <- unlist(interviews_clean$key_phrases)
word_count <-
data.frame(word = unique(key_words),
n = map_int(unique(key_words), count_words, key_words)) %>%
arrange(desc(n))
word_count_exact <-
data.frame(word = unique(key_words),
n = map_int(unique(key_words), count_exact_words, key_words)) %>%
arrange(desc(n))
word_count_exact %>% top_n(7) %>%
knitr::kable(caption = "Top 7 Key Words") %>%
kableExtra::kable_styling(full_width = FALSE)## Selecting by n| word | n |
|---|---|
| sql | 15 |
| excel | 10 |
| tableau | 7 |
| python | 6 |
| communication | 5 |
| reports | 3 |
| bi | 3 |
| r | 3 |
| qlikview | 3 |
If we go further down the list we find stuff like r and, bi (business intelligence), SAS, SAP, SCALA etc. So apparently if you want a job in data analytics in Ireland in 2018, you should probably be learning SQL, Excel, Tableau and Python. The other tools and languages are probably more of a “nice to have” than anything.
Conclusions
- A lot of recruiters will tell you if you’ve been rejected, at the same time a lot won’t.
- Finding your first job in a field can take a long time.
- You should probably learn SQL, Excel, Tableau and Python if you want a data analytics job.
A work by Cormac Nolan
cormac.nolan85@gmail.com