About me

Interactive dashboard using Google Analytics, Github Actions and Dc.js

⏳ 6 mins read time

Google analytics allows to measure a website performance. The information collected by this module allows to improve the content of a website by giving the owner insights about:

  • what are the most relevant parts of the website ;
  • how people reach the website ;
  • how many time people spend in the website ;
  • and also the path that they follow in the website.

All this information is available in the google analytics website, where you can find a sort of predefined charts. Sometimes you want to personalize this charts and use them at your own will. In this article I will show you how to:

  • connect to the google analytics api to fetch your analytics data ;
  • make interactive charts using the popular javascript library d3.js and dc.js ;
  • deploy your dashboard automatically in github pages serveless architecture by using github actions.

Interactive dashboard from my google analytics data

Google analytics

In order to fetch your analytics data, you have to:

1) Create project in Google Developers Console

2) Create service account and download json key file

3) Enable Google Analytics API

4) Grant service account to access Analytics account

All this information is also available in analytics developer guide.

After downloading the json service account key, you can use it in python or nodejs API to fetch your data. I used the example shown in the google documentation. I modified the example to:

  • save results in a csv file
  • use a service account key in a dict form instead of reading it from a file. This will allow us to read the token from an environment variable.

Create a dashboard

Javascript libraries like d3 allow us to create interactive visualizations, where the reader is able to define his own story by making his own data explorations.

  • import the js libraries in the html
  • create a script section in the html for visualizations
  • define the div place holder in the html where the figures will be drawn.

For the ones that like to see some code, here I present a simplified extraction from my js code to show you how it is simple to create the interactive figures.

// Reading data
d3.csv("data.csv").then(function(data) {
  // Create a Crossfilter instance
  var ndx = crossfilter(data);

  // Parse date
  var dateFormatParser = d3.timeParse("%Y%m%d");
  data.forEach(function(d) {
		d.dd = dateFormatParser(d.date);

  //Define Dimensions
  var countryDim = ndx.dimension(function(d) { return d["country"]; });
  var dateDim = ndx.dimension(function(d) { return d.dd; });
  var deviceDim = ndx.dimension(function(d) { return d["device"]; });
  var sourceDim = ndx.dimension(function(d) { return d["source"]; });
  var pageDim = ndx.dimension(function(d) { return d["pagePath"]; });
  var allDim = ndx.dimension(function(d) {return d;});

  //Group Data
  var countryGroup = countryDim.group().reduceSum(function (d) {
    return d["sessions"];
  var dateGroup = dateDim.group().reduceSum(function (d) {
    return d["sessions"];
  var deviceGroup = deviceDim.group().reduceSum(function (d) {
    return d["sessions"];
  var sourceGroup = sourceDim.group().reduceSum(function (d) {
    return d["sessions"];
  var pageGroup = pageDim.group().reduceSum(function (d) {
    return d["sessions"];
  var all = ndx.groupAll();

  var countryChart = dc.rowChart("#country-chart");
  var timeChart = dc.barChart("#time-chart");
  var deviceChart = dc.pieChart("#device-chart");
  var sourceChart = dc.rowChart("#source-chart");
  var pageChart = dc.rowChart("#page-chart");
  var dataTable = dc.dataTable("#data-table");
  var numberRecordsND = dc.numberDisplay("#number-records-nd");

  // Count the number of records
    .valueAccessor(function(d){return d; })

  // Country chart
    .margins({top: 10, right: 50, bottom: 30, left: 40})
    .ordering(function(d) { return -d.value })

  // ....

  // Draw all the charts

In a few words:

  • Read data from a csv file (other formats can be used csv, json, tsv, ..) ;
  • Create a crossfilter instance from the data. Using this, the filter used in one graph would be applied to all of the graphs that use the same crossfilter ;
  • Create dimensions using different available variables, like date, device, etc ;
  • Group data using dimensions an reducing using a defined parameter. I used the number of sessions.

Github Actions

Deploying the code in github pages allows to deploy static page and removes completely the server dependency. In addition github actions allows to have a complete CI/CD that can be configured in a yaml file like the following:

    # * is a special character in YAML so you have to quote this string
    - cron:  '57 23 * * *'

    name: Publish github pages
    runs-on: ubuntu-latest

      - uses: actions/checkout@v1
      - name: Set up Python 3.7
        uses: actions/setup-python@v1
          python-version: 3.7
      - name: install requirements
        run: make venv
      - name: get data
        run: make data
          TOKEN: ${{ secrets.TOKEN }}
      - name: build html
        run: make build
      - name: deploy
        uses: docker://peaceiris/gh-pages:v2.3.2
        if: success()
          PUBLISH_BRANCH: gh-pages
          PUBLISH_DIR: ./dist

In this way, I will:

  • fetch the data everyday in order to keep an updated dashboard and copy it to the dist folder ;
  • render my html and copy it to the dist folder ;
  • copy my js script to the dist folder ;
  • copy my dist folder to the gh-pages branch so that github can publish it in github-pages. I use @peaceiris github action to publish to github-pages.

This file is put in the .github/workflows folder. It’s automatically taken into account if you have github actions active for you account. It’s still in beta phase but it will be released to the public in November 2019.


Github actions are a great way to publish content in a serverless architecture. It’s schedule option opens great opportunities like the ones published in their repository.

The interactive dashboard uses only div elements in the html page. The frontend part can be controlled from libraries such as bulma in order to create responsive designs.

From my dashboard I can conclude that people visit mostly:

  • my image rotation article is one of the most popular articles. It’s because I wrote the article to solve a stackoverflow question and put the link to my article there. Stackoverflow is a very popular website!
  • my ssd yolo article is very popular too. It’s because deep learning applications are a big hype right now, so I get a lot of organic traffic coming mostly from google. My website is well ranked by google search algorithm, I think AMP standard helps a lot.
  • People visit mostly form desktop devices, but that may change in the future.
    I let you have your own conclusions by exploring the chart by yourself at this page

As always, the code is available at github. Don’t forget to 🌟.

In relation with 🏷️ d3.js, dc.js, google-analytics, github-actions:

Interactive data exploration using Dc.js

Interactive visualizations allows the reader to explore data for differents points of views. Javascript library Dc.js allows to combine multi variable firlters to the analysis.

Writing notes with Vimwiki and Hugo static generator

Vim is a simple and ubiquitous text editor that I use daily. In this article I show how to use Vim to take and publish diary notes using Vimwiki and Hugo.

Muse: Mopidy web client with Snapcast support

Services like Mopidy and Snapcast are ideal to create a multiroom streaming audio player using devices like the RaspberryPi or android telephones. This posts presents a web interface that uses the state of the art web technologies and integrates nicely with Mopidy and Snapcast.

Algolia search engine in Jekyll static pages

This article shows how to implementent Algolia searching engine in a static website and update the entries using Github Actions.

Subscribe to my newsletter πŸ“°

and share it with your friends: