Venus

Presentation

Venus is an all-in-one UX analyser that scores a website on its usability. It uses an AI that is based on machine learning models as well as in-depth analysis algorithms.

Here is the pitch introducing the project

Features

We perform more than 10 tests on the website based on powerful Algorithms and Machine Learning models. Thus one analysis can take up to 3 minutes.

UX analysis

We made in-depth analysis algorithms to evaluate:

1. The consistency and clutter of a website. Our algorithms score the website based on the quantity of possible user interactions. The more interactions there are, the lower the score. These algorithms are defined here and base their score calculations on a self-generated dataset based on various sites found with google search automated queries.
2. We also analyzed the page length and the number of items shown. This KPI is important to evaluate the complexity of the website, and therefore the trouble a user has to go through to get the information.
3. The load speed of the website (on desktop and mobile). A slow website implies a ruined user experience. We put this metric in the core of our analysis to evaluate the performance of the website.
4. The order of information. We think that the more a user scrolls on your website, the less the information he encounters should be important. Important pieces of information should be placed right at the beginning of each of your website page. In order to check this, we run two checks :
   • first we analyze the header consistency of each of the website pages : if a page mixes small and big titles in an unordered way, we reduce its efficiency score. This algorithm is defined here
   • then we perform keyword analysis on the top part of each page. We analyze their relevance and importance to the user. These keywords are then displayed on the frontend of the project after the analysis. This part uses computer vision algorithms, whose implementations can be found here

UI analysis

1. The visual coherence of the color choice. Our algorithm defined here produces statistics about the most used colors and evaluates the coherence of this choice based on color complementarity.
2. The visual cluttering : based on statistics produced in step 1, we analyse the most used colors on certain areas of the screen and determine if the website layout is not too complex and has sufficient blank spaces.
3. Accessibility : with the color analytics, we also check if the website is usable for all types of color-blind people. This metric is important because it can significantly impact the user experience of color-blind people. This algorithm is defined here
4. Mobile compatibility : all of our algorithms are also used on the mobile version of the target website, allowing us to check its accessibility from a smartphone.

Security analysis

The user experience is also impacted by the security of the page, otherwise your web browser will show warnings and reduce the quality of the user journey. We check the validity of the ssl certificate here.

Transfer learning

Everytime someone makes a request to analyze a website, we improve our convolutional learning model. The purpose of this model is to evaluate the various KPIs we currently analyze with algorithms from a simple screenshot of the website.

The more requests are made, the more our model becomes efficient in evaluating your website KPIs.

You can find the model definition here.

Try it out

Access

A server is currently running on our end, so access is simple!

If you want to use our tool, just click here!

Or go to this link https://ai4ux.poc-innovation.com/

Host it yourself!

Clone the repository and launch the following command from the root:

git checkout master
docker-compose up -d --build

A popup should then appear.

If it doesn't, go to your localhost:3000 for the frontend and localhost:5000 for the backend.

Made with ❤️ by PoC