Web-reputation study : systematic analysis across all internet

Objective

Assess web-reputation of the brand Anonymized for electrical equipments :
- On public blogs such as “installation-renovation-electrique.com”.
- Respectively to its competitors.

Approach

Identify the main blogs dealing with questions around this industry ;
Identify users circuits i.e. how did they arive on this blog, and where did they go after ? ;
Extract all text data from those websites ;
Quantify popularity of each brand on these blogs.

Visualization of momentum and audience around the internet

TOP 8 blogs in terms of Nb visits gather around 500 000 unique visitors / month.
Find bellow the resuts and hypothesis:

Starting point: the point of entry is a famous blog everyone use as a reference: installation-renovation-electrique.com.

Then we can use an iterative process to find out the main similar blogs in terms of semantic similarity:

• According to several metrics:
  => Number of visitors / month (source: SimilarWeb)
  => Average time spent on website (source: SimilarWeb)
  => Rating in France in terms of influence: this metric is the most precise as it is averaged on 3 sources.

• Combining the results of the 3 leaders in the field of semantic similarity:
  => SimilarWeb
  => Alexa
  => SEMRush

NB : size of points is proportional to Nb users / months (source: SimilarWeb)

These blogs are most often discovered from GOOGLE : via what key words ?

In 80% of situations, users arrive via ** GOOGLE research**.

Where do users go after ?

This can visualized in SimilarWeb:
- paypal.com
- conrad.fr
- ebay.fr
- amazon.fr

Using R and basic “fuzzy matching” to take into account syntactic variations, we counted occurences

Appendix: more about text extraction from the target websites

We used dexi.io. This tool allows to set a robot able to:
- “Crawl” sites and sub-sites.
- “Scrap” all the visited parts of sub-sites.

We could analyze 30 000 pages ! (32 Mo of html text).