Using Python and Data Science Practices in SEO Analysis of Data Benj Arriola 85SIXTY Speakerdeck.com/benjarriola @benjarriola 

Without Python and Data Science Practices ● SEO Tools ● Crawling Tools, Audit Tools, Keyword Research, Content Analysis, Crawlability, Indexability, Rank Tracking, Web Analytics, Social Listening, Backlink Analysis, etc. ● Analysis of Data ● Data often ends up in a spreadsheet. ● You sort, filter, join with Vlookups, summarize with Pivot tables, visualize with charts and graphs ● Recommended Action Items ● Update content ● Update code ● Update server setting ● Try to get other sites to update their sites ● Implementation ● SEO / Marketers / Content Managers / Writers with CMS Access ● Developers with Backend Code Access ● IT / Network Admin / Server Admin with Server Access 

With Python and Data Science Practices ● SEO Tools ● Crawling Tools, Audit Tools, Keyword Research, Content Analysis, Crawlability, Indexability, Rank Tracking, Web Analytics, Social Listening, Backlink Analysis, etc. ● Analysis of Data ● Data often ends up in a spreadsheet. ● You sort, filter, join with Vlookups, summarize with Pivot tables, visualize with charts and graphs ● Recommended Action Items ● Update content ● Update code ● Update server setting ● Try to get other sites to update their sites ● Implementation ● SEO / Marketers / Content Managers / Writers with CMS Access ● Developers with Backend Code Access ● IT / Network Admin / Server Admin with Server Access Export as Spreadsheets, or Use APIs Export Results of Analysis Integrate with CMS for Automated Updates Converting your manual analysis into steps your Python script will do Typically, the biggest challenge is: (1) converting your manual analysis into distinct steps and (2) turning these steps into Python instructions. Python Analysis of Data and Integration with CMS’ for Automated Updates may take time to setup, but once completed, everything will be faster 

Why We Use Python and Data Science Practices? Handling Large Amounts of Data Repetitive Process Faster Gathering Data, Analysis & Implementation 

Crawling / Audit Tools Keyword Research Content Analysis Crawlability / Indexability Ranking Tracking Web Analytics Social Listening Backlink Analysis Sources of Data 

Vector Embeddings ● Some NLP Vector Embedding Factors ● Semantic Similarity ● Gender and Role Relationships ● Tense and Parts of Speech ● Polarity and Sentiment ● Formal vs. Informal Language ● Topical Relevance ● Lexical Hierarchy ● Contextual Dependency ● Sentential Structure ● Emotional Tone or Connotation ● SEO Tools Known to Use Vector Embedding ● Screaming Frog ● inLinks ● WordLift ● MarketMuse ● MarketBrew 

Examples Applications ● Reporting Dashboards and Analysis from Various Tools ● Determining Revenue Per Keyword ● Quantifying the Effect of Pagespeed Updates on Traffic and Revenue ● Keyword Research ● Targeting the Best Keywords ● Vector Embedding ● Finding Duplicate Content ● Identifying Internal Linking Opportunities ● Updating Image Alt Text as Scale ● Writing Title Tags and Meta Descriptions for Thousands or Millions of Pages 

Speakerdeck.com/benjarriola @benjarriola Keyword Research Python & Data Science Use Case 

Common Keyword Research Process ● Objective ● Determine the best keywords to target and optimize ● General Process ● Start with some initial seed words ● Use keyword research tools to get more keyword ideas ● Analyze the data to determine which keywords to target ● Finalize the primary target keywords Seed Words Keyword Exploration Keyword Analysis Primary Target Keywords 

Common Keyword Research Process ● Seed Words ● Common Sense ● Current Ranking Keywords (GSC, SEMRush) ● Main Navigation ● Keyword Exploration ● Competitor Keywords – SEMRush ● Keyword Analysis ● Python Script ● Target Keywords ● Exported from Python Script Seed Words Keyword Exploration Keyword Analysis Primary Target Keywords Keyword Exploration Keyword Analysis Primary Target Keywords Seed Words Keyword Exploration Keyword Analysis Primary Target Keywords 

Keyword Analysis ● Analysis Stage: Converting your manual analysis into steps your Python script will do ● More popular industries ● More competitors ● More current ranking keywords ● More website pages ● More product lines / services More Keywords to Analyze Breakdown into steps to narrow down hundreds or millions of keywords to a smaller number of keywords to target. 

Target Keywords – Narrow Down ● Narrow Down Rules ● A good balance of long tail and head terms ● Why? Get the low hanging fruit and monitor the journey to the general head terms ● A balance of transactional, informational, navigational ● Why? ● Transactional for more sales ● Informational, for new customer discovery and link bait purposes ● Navigational, for customer support, ORM ● Balance of high KEI and high Search Volume ● Why? Low-hanging fruit to gain benefits, more sales quicker, while working on the higher search volume in the process ● Common keywords between competitors ● Why? ● Quicker way to narrow down industry-relevant keywords. ● Quicker to exclude competitor brands. ● A good balance of different topics ● Why? High search volume or high KEI can make keywords too focused on 1 thing. It might be 1 product, 1 service, 1 product trend, ● Include and exclude list filter ● Long tail keyword pattern analysis Whether this is 1,000 keywords, or 1,000,000 keywords, the Python Script will run all rules and narrow down to whatever number of keywords you set this out to. Different SEOs may have a different process, as long as this process can be articulated into distinct steps, then it can be transformed into Python code. 

Implementation ● Having the targeted keywords alone is good enough for keyword research. ● Implementation: ● Use keywords in key areas of a page. ● Title Tag ● Meta Description ● Heading Tags ● Main Content ● Image Alt Text ● URLs ● Schema Markup ● Etc. 

Implementation Automation ● Categorize Targeted Keywords ● Use the ChatGPT API ● Cost for running this for 100 keywords OpenAI's pricing for the GPT-4 model is $0.03 per 1,000 prompt tokens and $0.06 per 1,000 completion tokens. ● Prompt Cost: (150 tokens / 1,000) * $0.03 = $0.0045 ● Completion Cost: (500 tokens / 1,000) * $0.06 = $0.03 ● Total Cost: $0.0045 + $0.03 = $0.0345 

Implementation Automation ● Assigning a category and subcategory to a page ● Uses Screaming Frog’s Vector Embedding ● The category and subcategories will look for the highest similarity match. 

Implementation Automation ● Rewriting Title Tags if the main target keyword is not used ● For each category and subcategory, find the highest search volume keyword. ● If the keyword is not used in the title tag, use the ChatGPT API to rewrite the title and use the keyword. ● Similar rules can be done for the meta description ● Doing a manual QA check would be a smart thing to do in the initial runs 

Speakerdeck.com/benjarriola @benjarriola Getting Started & Learning More Other Information & Resources 

Potential Challenges ● The Learning Curve ● Not everyone learns at the same pace ● The Technical SEO team does it for the rest of the team ● Great internal training program ● Create an application with a user- friendly interface (Streamlit, Dash, Flask, Django) ● Hire people ● Partner with other companies ● Tool Ownership ● You or your employer? 

Other Resources ● Learning Python and PANDAS Basics ● YouTube ● Coursera ● Udemy ● ChatGPT ● SEO Application ● Jacky Chou - Indexy ● Greg Bernhardt – ImportSEM ● JC Chouinard 

Speakerdeck.com/benjarriola @benjarriola Welcome to San Diego! Thanks 

Benj Arriola Spoken at 48 conferences in 4 different countries, 21 cities since 2007. Won major prizes in 7 international SEO competitions from 2005 to 2009 that include cash and a brand new car! Python Usage 

Speakerdeck.com/benjarriola @benjarriola Questions?