README.md

metadata

license: cc-by-sa-4.0
language:
  - en
size_categories:
  - 10K<n<100K

SEO Data Dive 2025: Discovering Patterns in Search (Data Challenge)

IMPORTANT NOTE: This is an exploratory analysis challenge, not a traditional machine learning competition. Winners will be selected exclusively by an expert jury based on the quality of the submitted analyses. By participating, you explicitly agree to this evaluation process.

Welcome & The Mission

Welcome to the SEO Data Dive Data Challenge! Your mission, should you choose to accept it, is to decode the secrets of search engine rankings through curiosity, analysis, and data storytelling.

Your Goal: Create a compelling, public Hugging Face Space that tells a data-driven story. The Grand Prize: The creators of the best analyses will win an invitation to present their work as speakers at the RAT Community Meeting 2025 in Hamburg, Germany, supported by travel grants and cash awards.

If you have any questions about the data or the rules, please don't hesitate to start a new discussion in the "Community" tab. We will be actively monitoring it and are happy to help!

How to Participate & Submit

1. Explore the Data: Download the data from the "Files and versions" tab.
2. Create Your Analysis: Develop your analysis and visualizations locally in a notebook or script.
3. Build a Hugging Face Space: Create a public Hugging Face Space (e.g., using Streamlit or Gradio) to present your findings. Your Space must contain:
   • Your complete data analysis (e.g., embedded as a notebook or presented through interactive elements).
   • Your key visualizations and charts.
   • Your data story and conclusions.
   • A link to your summary presentation (max. 10 slides, e.g., PDF, Google Slides).
   • Or send us your data story in a format of your choice to sebastian.suenkler@haw-hamburg.de
4. Submit Your Entry: Go to the "Community" tab of this dataset. Post a new discussion topic with the title "Submission: [Your Project Title]" and include a link to your public Hugging Face Space.

Avenues for Exploration

This dataset opens up numerous avenues for exploration. Here are a few ideas:

1. Feature Engineering & Modeling:
   • Develop creative new features (e.g., URL complexity, brand vs. non-brand queries).
   • Build simple predictive models (e.g., for position) to conduct a feature importance analysis.
   • Apply NLP techniques to textual fields (title, description, h1) to extract topics or sentiment.
2. Creative Data Visualization & Storytelling:
   • Create interactive dashboards for exploring the data.
   • Develop insightful charts that illustrate complex relationships.
   • Tell a compelling “story” based on the patterns you discover. For example: "How does DuckDuckGo really differ from Google?" or "The Anatomy of a #1 Ranking in 2025".

Evaluation Criteria

A jury will evaluate the submitted Hugging Face Spaces based on the following criteria:

• Insightfulness & Originality (40%): Are new, non-obvious patterns uncovered? Is the approach creative?
• Analytical Rigor (30%): Are statistical methods appropriate and correctly applied? Are conclusions supported by data?
• Clarity & Storytelling (30%): Is the story communicated clearly and persuasively? Are visualizations meaningful and aesthetically pleasing?

Prizes

Winning this challenge is about showcasing your groundbreaking work to a community of experts. The centerpiece of each prize is the invitation to present your findings as a speaker at the RAT Community Meeting 2025 in Hamburg, Germany. The cash awards and travel grants are provided to support and reward you for accepting this unique opportunity.

The prize packages are awarded as a whole and are non-severable. Acceptance of the prize constitutes a firm commitment to present at the conference.

The top three ranked teams will be offered the following prize packages:

• 1st Place Prize Package:
  • An invitation to present as a featured speaker at the RAT Community Meeting 2025.
  • A travel grant to support attendance at the conference.
  • A $500 USD cash award.
• 2nd Place Prize Package:
  • An invitation to present at the RAT Community Meeting 2025.
  • A travel grant.
  • A $300 USD cash award.
• 3rd Place Prize Package:
  • An invitation to present at the RAT Community Meeting 2025.
  • A travel grant.
  • A $200 USD cash award.

Crucial: Should a winning team be unable or unwilling to accept the invitation to present, they forfeit the entire prize package. This includes the travel grant and cash award. The complete prize package will then be offered to the next-highest-ranked team as determined by the jury.

Timeline

• Competition Launch: 07/07/2025
• Entry Deadline: 08/31/2025
• Judging Period: 09/01/2025 - 09/05/2025
• Winners Announced: 09/06/2025

Competition Rules

Sponsor: German Research Foundation (DFG), Project No. 55736635 Host: SearchStudies

By entering this competition, you agree to these Official Rules.

1. Eligibility: The competition is open to all individuals who are of legal age of majority. Employees, interns, contractors, and officers of the Sponsor are not eligible for prizes.
2. Team Limits: Teams may consist of a maximum of five (5) members.
3. Submission: The final submission is a public Hugging Face Space. The link must be posted in the "Community" tab by the Entry Deadline. The Space must contain a link to a summary presentation (max. 10 slides).
4. Judging: A jury appointed by the Sponsor will evaluate submissions based on the published Evaluation Criteria. The decisions of the jury are final.
5. Data License: The provided data is licensed under CC BY-SA 4.0. External, publicly available data is permitted, provided the source is cited in your submission.
6. Winner License: As a condition of receiving a prize, winners agree to license their winning work (code, text, visualizations) under the MIT License (for code) and CC BY-SA 4.0 (for non-code elements).
7. Prize Acceptance and Obligations: The prize packages are awarded in their entirety. The invitation to present at the RAT Community Meeting 2025 is an integral and mandatory component of the prize. By accepting a prize, a winning team formally commits to preparing and delivering this presentation. Failure to fulfill this presentation commitment may result in the forfeiture of the entire prize package.
8. Disclaimer: This competition is organized and sponsored solely by the German Research Foundation (DFG), Project No. 55736635. Hugging Face, Inc. is not a sponsor, is not involved in the administration of the competition, and is not responsible for the selection of winners or the awarding of prizes.

Dataset Description

The dataset for this challenge is provided in the file below, which you can find in the "Files and versions" tab.

• RAT_Search_Results_Trends_SEO_2025.csv - The Main Dataset: This file contains the complete dataset of approximately 90,000 search engine results from Google, Bing, and DuckDuckGo. Each row is a unique search result, enriched with over 50 detailed SEO metrics from our SEO classifier.

The search queries were sourced from trending topics on Google Trends in Germany between May 21 and May 28, 2025.

Data Fields / Column Descriptions

Column Name	Explanation
result	Unique ID for the search result (row in the CSV).
url	The exact URL of the search result.
main	The primary domain of the URL (e.g., https://de.wikipedia.org/).
position	The ranking position of the search result on the SERP.
title	The title of the search result, as displayed on the SERP.
description	The description of the search result, as displayed on the SERP.
searchengine	The search engine from which the result originates (e.g., DuckDuckGo_DE).
query	The search query that led to this result.
created_at	Timestamp of when the search result was collected.
seo_score	The overall SEO score calculated by our classifier (0-100).
analysis_explanation	Textual explanation of how the SEO score is derived.
canonical	1 if a canonical tag points to the URL itself, 0 otherwise.
category_content_quality	Score (0-100) for content quality.
category_meta_elements	Score (0-100) for meta elements (title, description, etc.).
category_technical_seo	Score (0-100) for technical SEO aspects.
category_user_experience	Score (0-100) for user experience.
content_length_score	Score (0-100) based on content length.
content_type	The content type of the resource (e.g., HTML, PDF).
description_score	Score (0-100) for the quality of the meta description.
description_text	The actual text of the meta description from the webpage.
exclusion_reason	Reason why a page might have been excluded from full SEO scoring.
external_link_count	Number of external links on the page.
h1	The text of the first H1 tag on the page.
heading_structure_score	Score (0-100) for heading structure (H1, H2, etc.).
https	True if the page uses HTTPS, False otherwise.
image_optimization_score	Score (0-100) for image optimization (e.g., alt tags).
internal_link_count	Number of internal links on the page.
keyword_optimization_reasons	Reasons for the keyword optimization score (often a JSON array string).
keyword_optimization_score	Score (0-100) for keyword optimization.
link_quality_score	Score (0-100) for link quality.
loading_time	Page loading time (e.g., in milliseconds).
micros	Detected microdata/structured data (often a JSON array string).
navigation_score	Score (0-100) for navigation quality.
nofollow	Number of nofollow links on the page.
og	Open Graph meta tags found (JSON string with og_tags, twitter_tags).
robots_txt	1 if a robots.txt file is present, 0 otherwise.
seo_classification	SEO classification (e.g., probably_optimized).
sitemap	1 if a sitemap is referenced, 0 otherwise.
social_tags	Details about social media tags (often a JSON array string).
social_tags_score	Score (0-100) for social media tag optimization.
title_score	Score (0-100) for the quality of the page title.
title_text	The actual text of the <title> tag from the webpage.
tools_ads	Detected advertising tools (JSON array string).
tools_analytics	Detected analytics tools (JSON array string).
tools_caching	Detected caching tools (JSON array string).
tools_seo	Detected SEO tools (JSON array string).
tools_social	Detected social media tools/widgets (JSON array string).
url_length	Length of the URL in characters.
viewport	1 if a viewport meta tag is present, 0 otherwise.
wordpress	1 if the page was built with WordPress, 0 otherwise.

Important Note on Data Types: Several columns (e.g., og, tools_analytics, micros) contain JSON data stored as strings. You may need to parse these to use them effectively.