Journey of Analytics

Deep dive into data analysis tools, theory and projects

Tag: automated email with R

Email Automation for Google Trends

This blogpost will teach you set up automated email reports to view how search volumes i.e. Google Trends vary over time.

Email automation for Google Trends over time

The email report will also include important search terms that are “rising” or near a “breakout point”. This can be really useful as the breakout keywords indicate users across the globe have recently started paying great attention to these search terms. So you could create original content to attract people using these phrases in their Google searches. This is an amazing way to boost traffic/ sales leads, etc. by taking advantage of new trends.

Note, that the results are relative, not absolute search volume terms. But it is still quite useful, as you could narrow down your keywords list, and login to Adsense/ keyword planner to get the exact search volume. Export the list from this script to a .csv and simply upload in Adsense. It also gives you a clear indication of how search volumes compare across similar terms – classic SEO !

We will perform all our data pull requests and manipulations in R. The best part is that you do not need any API keys or logins for this tutorial!

[ Do you know what else is trending? My book on how to get hired as a datascientist. Check it out on Amazon, where it is the #1 NEW RELEASE and top 10 in its categories. Whether you are currently enrolled in a data science course or actively job searching, this book is sure to help you attract multiple job offers in this lucrative niche. ]

Without further interruptions – let’s dive right in to the trend analysis.

Step 1 – Load Libraries

Start by loading the libraries – “gtrends” is the main library package which will help us pull data for search results.

library (gtrendsR)
library(ggplot2)
library(plotly)
library(dplyr)
library(sqldf)
library(reshape2)

Step 2 – Search terms and timeframe

We need to specify the variables we need to pass to the search query:

  • search terms of interest. Unfortunately, you can only pass 5 terms at one time. Theoretically you could store the value and repeat the searches, but since numbers are relative, make sure that you keep at least one term in common and then normalize.
  • Remember, we get relative search volumes (max = 100); NOT absolute volumes!
  • timeframe under consideration, has to be in the format “YYYY-mm-dd” ONLY. No exceptions. End date = today, whereas start date will be the the first of the month, six months ago.
  • country codes – I’ve only used “US”, but you can search for multiple countries using a string array like c(“US”, “CA”, “DE”) which stands for US, Canada and Germany.
  • Channels – default is “web”, but you can search for “new” or “images”. Images would be good for those promoting content in Pinterest or Instagram.

date6mthpast <- Sys.Date() – 180
startdate <- paste0( substr(as.character(date6mthpast), 1,7), “-01”, sep = “”)
currdate = as.character(Sys.Date())
timeframe = paste( startdate, currdate)
keywords <- c(“Data Science jobs”, “MS Analytics Jobs”, “Analytics jobs”,
“Data Scientist course”, “job search”)
country=c(‘US’)
channel=’web’

We will use the gtrends() function to pull the data. This function returns a list of variables and dataframes. Of main interest to us is the “interest_over_time” data frame, “interest_by_region” (state names mostly). There is also a dataframe named “interest_by_city”. The first one does hold data with date value, while the others do not, so it will be aggregated over the entire data frame. I worked around this by pulling once for 6 months, and again for the last 45 days.

trends = gtrends(keywords, gprop =channel,geo=country, time = timeframe )

Step 3 – Search Volume over time

We will use the ggplot function to get a graphical representation of the search volume.

ggplot(data=time_trend, aes(x=date, y=hitval, group=keyword, col=keyword))+
geom_line()+xlab(‘Time’)+ylab(‘Relative Interest’)+ theme_bw() +
theme(legend.title = element_blank(), legend.position=”bottom”, legend.text=element_text(size=12)) +
ggtitle(“Google Search Volume over time”)

Google search volume over time
Google trends over time

Notice that the interest in term “job search” is significantly higher than any other keyword and spike in April and May (people graduating maybe?) This makes sense as it is a more generic term, and applies to larger number of users. On a similar vein, look at the volumes for “MS analytics jobs” . It is almost nil, so clearly a non-starter in terms of targeting.

Step 4 – Interest by Regions

Let us look at interest by region (state names in the US ) . This might be useful to target folks by region, or even location-based ads.

locsearch = trends$interest_by_region
plot2 <- ggplot(subset(locsearch, locsearch$keyword != “job search”),
aes(x=as.factor(location), y= hits, fill=as.factor(keyword) )) +
geom_bar(position=”dodge”, stat=”identity”) +
coord_flip() +
theme(legend.title = element_blank(), legend.position=”bottom”, legend.text=element_text(size=9)) +
xlab(‘Region’)+ ylab(‘Relative Interest’) +
ggtitle(“Google Search Volumes by Region/State”)

Note, that “data scientist course” has almost zero interest compared to broader terms like “analytics jobs” or “datascience Jobs”. Interestingly, “analytics jobs” seems to be a preferred term over “data science” only in NY, MA and Washington DC.

Step 5 – Trending Terms

The original list also returns a dataframe titled “related_queries”. Some of these can have tags as “rising” or “breakout”. These are the terms we would like to know as they occur.

breakoutdf = reltopicdf[reltopicdf$related_queries == ‘rising’,]
volbrkdf = breakoutdf[1:10,]
row.names(volbrkdf) = NULL
volbrkdf

trending seach terms related to keywords of interest
trending seach terms related to keywords of interest

Step 6 – Create pdf for email attachments

We will use the pdf() command to create an attachment with all the graphs and add to the email we want to send/receive on a daily basis.

filename2 = paste0(“Google Search Trends – “, Sys.Date(), “.pdf”)
pdf(paste0(“Google Search Trends – “, Sys.Date(), “.pdf”))
plot1
plot2
print(volbrkdf)
dev.off()
dev.off()

If you want to embed one of these figures in the email itself, instead of attachment then please see my post on automated reports.

Step 7 – Send the email

We use the RDCOM library to send out emails. You do need to have Outlook mail client on your PC.

library(RDCOMClient)
OutApp <- COMCreate(“Outlook.Application”)
outMail = OutApp$CreateItem(0)
outMail[[“To”]] = “an**@gmail.com”
outMail[[“subject”]] = paste(“Google Search Trend Report -“, Sys.Date())
email_body1 <- “Write email body content with correct html tags”
outMail[[“HTMLBody”]] = email_body1
file_location = paste0(paste0(abs_path,”/”),filename2)
outMail[[“Attachments”]]$Add(file_location)
outMail
outMail$Send()

The first time you run this, you might need to “allow” RStudio to access your email account. Just add this script to a task scheduler, and choose frequency of delivery, and you are all set! Note, a dummy email address is used, so remember to change the recipient address.

Try it out and let me know if you have any questions, or run into errors. The script is available on the Projects page – link here

Last but not the least, if you are interested in getting hired in a data science field, then please do take at my job-search book. Here is the Amazon link.

Happy Coding!

Automated Email Reports with R

R is an amazing tool to perform advanced statistical analysis and create stunning visualizations. However, data scientists and analytics practitioners do not work in silos, so these analysis have to be copied and emailed to senior managers and partners teams. Cut-copy-paste sounds great, but if it  is a daily or periodic task, it is more useful to automate the reports. So in this blogpost, we are going to learn how to do exactly that.

The R-code uses specific library packages to do this:

  • RDCOMClient – to connect to Outlook and send emails. In most offices, Outlook is still the defacto email client, so this is fine. However, if you are using Slack or something different it may not work.
  • r2excel – To create an excel output file.

The screenshot below shows the final email view:

email screenshot

email screenshot

As seen in the screenshot, the email contains the following:

  • Custom subject with current date
  • Embedded image
  • Attachments – 1 Excel and 1 pdf report

Code Explanation:

The code and supporting input files are available here, under the Projects page under Report automation. The code has 4 parts:

  • Prepare the work space.
  • Pull the data from source.
  • Cleaning and calculations
  • Create pdf.
  • Create Excel file.
  • Send email.

The workflow should look something like this. For really simple reports, where you do not need much formatting, you can do everything within the first script itself.

Prepare the work space

I always set the relative paths and working directories at the very beginning, so it is easier to change paths later. You can replace the link with a shared network drive path as well.

Load library packages and custom functions. My code uses the r2excel package which is not directly available as an R-cran package. So you need to install using devtools using the code below.

It is possible to do something similar using the “xlsx” package, but r2excel is easier.

Some other notes:

  • you need the first 2 lines of code only for the first time you installation. From the second time onwards, you only need to load the library.
  • r2excel seems to work only with 64-bit installations of R and Rstudio.
  • you do need Java installed on your computer. If you see an error about java namespace, then check the path variables. There is a very useful thread on Stackoverflow, so take a look.
  • As always, if you see errors Google it and use the Stack Overflow conversations. In 99% of cases, you will find an answer.

Pull the data from source

This is where we connect to an Excel CSV (or text) file. In practice, most people connect to a database of some kind. The R-script I am using connects to a .csv file, but I have added the code to a connect to a SQL database.

That code snippet is commented out, so feel free to substitute your own sql database links. The code will also work for Amazon EC2 cluster.

Some points to keep in mind:

  • If you are using sqlquery() then please note that if your query has an error then R sadly shows only a standard error message. So test your query on SQL server to ensure that you are not missing anything.
  • Some queries do take a long time, if you are pulling from a huge dataset. Also the time taken will be longer in R compared to SQL server direct connection. Using the  Sys.time() command before and after the query is helpful to know how long the query took to complete.
  • If you are only planning to pull the data randomly, it may make sense to pull from SQL server and store locally. Use the fread() function to read those files.
  • If you are using R desktop instead of R-server, the amount of data you can pull may be limited to what your system configuration.
  • ALWAYS optimize your query. Even if you have unlimited memory and computation power, only pull the data you absolutely need. Otherwise you end up unnecessarily sorting through irrelevant data.

Cleaning and calculations

For the current data, there are no NAs, so we don’t need to account for those. However, the read.csv() command creates factors, which I personally do not like, as they sometimes cause issues while merging.

Some of the column names have “.” where R converted the space in the names. So we will manually replace those with an underscore using the gsub() function.

We will also rank the apps based on categories of interest, namely:

  • Most Popular Apps – by number of Reviews
  • Most Popular Apps – by number Downloads and Reviews
  • Most Popular Categories – Paid Apps only
  • Most popular apps with 1 billion installations.

Create pdf

We are going to use the pdf() function to paste all graphs to a pdf document. Basically what this function does is write the graphs to a file rather than show on the console. So the only thing to remember is that if you are testing graphs or make an incorrect graph, everything will get posted to the pdf until you hit the “dev.off()” function. Sometimes if the graph throws an error you may end up with a blank page, or worse, with a corrupt file that cannot be opened.

Currently, the code I am only printing 2 simple graphs using ggplot() and barplot() functions, but you can include many other plots as well.

Create Excel file.

The Excel is created in the sequence below:

  • Specify the filename and create an object of type .xlsx This will create an empty Excel placeholder. It is only complete when you save the Workbook using the saveWorkbook() at the end of the section.
  • Use the sheets() to create different worksheets within the Excel.
  • The  xlsx.addHeader() adds a bold Header to each sheet which will help readers understand the content on the page. The r2excel package has other functions to add more informative text in smaller (non-header) font as well, if you need to give some context to readers. Obviously, this is optional if you don’t want to add them.
  • xlsx.addTable() – this is the crucial function that adds the content to Excel, the main “meat” of what you need to show.
  • saveWorkbook() – this function will save the Excel to the folder.
  • xlsx.openFile() – this function opens the file so you can view contents. I typically have the script running on automated mode, so when the Excel opens I am notified that the script completed.

Send email

The email is sent using the following functions:

  • OutApp() – creates an Outlook object. As I mentioned earlier, you do need Outlook and need to be signed in for this to work. I use Outlook for work and at home, so I have not explored options for Slack or other email clients.
  • outmail[[“To”]] – specify the people in the “to” field. You could also read email addresses from a file and pass the values here.
  • outmail[[“cc’]] – similar concept, for the cc field.
  • outmail[[“Subject”]] – I have used the paste0() function to add the current date to the subject, so recipients know it is the latest report.
  • outMail[[“HTMLBody”]] – I used the HTML body so that I can embed the image. If you don’t know HTML programming, no worries! The code is pretty intuitive, you should be able to follow what I’ve done. The image basically is an attachment which the HTML code is forcing to be viewed within the body of the email. If you are sending the email to people outside the organization, they may see a small box instead of the image with a cross on the top left (or right) of the box. Usually, when you hover your mouse near box and right click, it will ask them to download images. You may have seen similar messages in gmail, along with a link to “show images” or ‘always show images from this sender’. You obviously cannot control what the recipient selects, but testing by sending to yourself first helps smoothing out potential aesthetic issues.
  • outMail[[“Attachments”]] – function to add attachments.
  • outMail$Send() – until you run this command, the mail will not be send. If you are using this in office, you may get a popup asking you to do one of the following. Most  of these will generally go away after the first use, but if they don’t, please look up the issue on StackOverflow or contact your IT support for firewall and other security settings.
    • popup to hit “send”
    • popup asking you to “classify” the attachments (internal / public/ confidential) Select as appropriate. For me, this selection is usually  “internal”
    • popup asking you to accept “trust” settings
    • popup blocker notifying you to allow backend app to access Outlook.

That is it – and you are done! You have successfully learned how to send an automated email via R.

Facebook
LinkedIn