How to Analyse Your YouTube Channel with Claude and Ooty Iris

YouTube Studio gives you all the data you need. Views, watch time, subscriber changes, revenue, traffic sources, demographics, impressions, CTR -- and that is before you start looking at individual video performance and audience retention curves.

The problem is not the data. It is the synthesis. YouTube Studio presents information in separate report sections, each with its own filters and time ranges. Understanding why your channel is performing the way it is means visiting several sections, mentally combining the information, and drawing your own conclusions. A thorough analysis can take 2-3 hours.

This tutorial shows a different approach: using Claude with Ooty Iris to bring your YouTube data into a conversation. You ask questions. Claude pulls the data through Iris tools. The analysis happens in real time, in plain language.

The YouTube Analysis Workflow

Eight areas to cover in a complete channel analysis session

1

Channel Overview

Views, watch time, subscribers, revenue -- period over period

2

Top Videos

What content works and what patterns repeat

3

Traffic Sources

Search vs suggested vs external -- where growth comes from

4

Audience Profile

Demographics, geography, and when they watch

5

Video Deep-Dive

Retention curves, drop-off points, CTR diagnosis

6

Video SEO Audit

Metadata, tags, descriptions, search optimization

7

Competitor Intel

What is working in your niche that you are not doing

8

Trending Topics

Rising trends that fit your audience and format

What You Need

• An Ooty account with Iris activated -- ooty.io
• Claude Desktop (or any MCP-compatible client) with Iris connected via MCP endpoint and license key
• Google/YouTube account connected via OAuth in your Ooty dashboard (Iris > Connect > Google)

First-time setup: go to your Ooty dashboard, navigate to Iris, and connect your Google account through OAuth. Then add Iris to Claude using the MCP endpoint URL shown in your dashboard. About 5 minutes total, one-time task.

What Iris Gives Claude

When Iris is connected, Claude has access to 20+ YouTube analytics tools including:

• get_channel_overview -- channel-wide metrics with period-over-period comparison
• get_top_videos -- best performing videos ranked by your chosen metric
• get_traffic_sources -- breakdown of where viewers come from
• get_audience_demographics -- age, gender, geography of your audience
• get_audience_activity -- when your audience is most active
• get_video_stats -- detailed metrics for specific videos
• get_retention_curve -- audience retention data showing where people drop off
• get_video_seo_score -- SEO audit for individual videos
• get_competitor_stats -- public stats for competitor channels
• find_content_gaps -- topics competitors cover that you don't
• get_trending_topics -- rising topics in your niche

You do not need to know which tool does what. Ask Claude what you want to understand, and it picks the right tool.

C

Prompt to Claude

“My latest video published 2 weeks ago and underperformed. Pull the analytics and help me understand what went wrong. Compare it to my channel averages, then check the retention curve -- where exactly are people dropping off?”

Iris runs:get_video_statsget_channel_overviewget_retention_curve

Step 1: Get Your Channel Overview

Start every analysis with a broad view of where things stand. This provides context and flags anything unusual before you drill into specifics.

Prompt:

"Give me an overview of my YouTube channel performance for the last 28 days compared to the previous 28 days. What changed? Are we moving in the right direction?"

Claude calls get_channel_overview and returns your key metrics with period-over-period deltas: views, watch time, subscribers gained, likes, comments, estimated revenue (if monetised). More importantly, it frames the changes in context: "Your views are up 12% but watch time is down 8%. This suggests you are getting more impressions but viewers are watching less of each video -- average view duration dropped from 4:20 to 3:45. That pattern often indicates title/thumbnail expectations are not matching the content."

That diagnostic framing is the difference between seeing a number and understanding it.

The Four Metrics That Matter

View count is a lagging indicator. These four metrics predict future growth.

Impression CTR

How often people click after seeing your thumbnail

Benchmark:4-10%

Leading indicator of thumbnail/title effectiveness

Avg View Duration

How long people actually watch

Benchmark:50%+ of video length

Primary signal YouTube uses for recommendations

Subscriber Conversion

New subscribers per 1,000 views

Benchmark:2-5 per 1K views

Indicates content builds lasting audience

Traffic Source Mix

Search vs suggested vs browse vs external

Benchmark:Varies by strategy

Reveals which growth lever is actually working

Step 2: Find Your Top-Performing Videos

Understanding what works is the foundation of content strategy. You cannot replicate success if you do not know what success looks like on your channel specifically.

Prompt:

"What are my top 10 videos by watch time in the last 90 days? For each one, give me the title, views, average view duration, and subscriber conversion rate. Are there patterns in what makes the top videos successful?"

Claude calls get_top_videos and returns the ranked list, then reasons about patterns: "Your top 5 videos by watch time all share a common format -- they lead with a specific problem statement in the first 30 seconds rather than an intro. Your tutorial-style videos are averaging 6:20 view duration vs 3:50 for discussion videos. Highest subscriber conversion rates are in tutorials, not commentary."

Those cross-video patterns would take significant time to surface manually in YouTube Studio.

Step 3: Understand Your Traffic Sources

Where viewers come from determines which growth levers you can actually pull. A search-driven channel grows differently from one powered by the algorithm.

Prompt:

"Break down my traffic sources for the last 28 days. How are people finding my videos? What percentage comes from YouTube search vs suggested videos vs external? Has this changed compared to last month?"

Claude calls get_traffic_sources and returns the breakdown. Common patterns:

• Search-heavy channels get consistent, predictable traffic but can struggle to hit algorithm-driven spikes
• Suggested-video-heavy channels grow faster when videos perform well but traffic can drop sharply when recommendations stop
• External traffic (websites, newsletters, social) is often undervalued -- it indicates a more loyal audience since they sought you out

If your search traffic is low but your content is clearly searchable (tutorials, how-tos, specific topics), that is an actionable gap: your metadata probably needs work.

According to YouTube's Creator Academy, search and suggested videos together account for over 70% of watch time on the platform.

Step 4: Know Your Audience

Understanding who watches and when helps you make better decisions about content format, video length, and publishing schedule.

Prompt:

"Who is my audience? Give me demographics -- age groups, top countries, gender breakdown. Also tell me when they are most active. I want to know the best times to publish new videos."

Claude calls get_audience_demographics and get_audience_activity, then synthesises: "Your core audience is 25-44 years old (67%), primarily from the US (42%), UK (18%), and Australia (9%). They are most active between 6pm-10pm local time, with Tuesday and Thursday showing the highest engagement. Your current publishing schedule (Monday mornings) does not align well with peak activity -- consider shifting to Tuesday or Thursday evenings."

YouTube's own research confirms that videos published when a channel's audience is most active tend to perform better in the first 48 hours, which influences longer-term distribution.

Step 5: Diagnose a Specific Video

When a video underperforms -- or overperforms -- understanding why is more valuable than just knowing that it did.

Prompt:

"My video [paste title] published 2 weeks ago and I am disappointed with the performance. Pull the analytics and help me understand what went wrong. Compare it to my channel averages."

Claude calls get_video_stats and compares against your norms. It then identifies which specific metric is responsible.

Diagnostic Patterns

What different metric combinations tell you -- and what to do about them

High impressions, low CTR

medium priority

Why: Thumbnail or title is not compelling

Fix: Test new thumbnails, rewrite title

Good CTR, low watch time

high priority

Why: Content does not match the expectation set by title/thumbnail

Fix: Align intro with promise, restructure content

Low impressions overall

high priority

Why: YouTube is not distributing the video

Fix: Check metadata, tags, category, topic cluster fit

Good watch time, low subscribers

low priority

Why: Content delivers value but does not create a reason to subscribe

Fix: Add a subscribe CTA, create series content

Sudden retention drop

medium priority

Why: Specific moment loses viewers (topic shift, pace change)

Fix: Review the exact timestamp, restructure future content

Views up, watch time down

medium priority

Why: Getting more impressions but holding attention less

Fix: Title/thumbnail expectations may not match content

Follow-up:

"Pull the retention curve for that same video. Where are people dropping off? Are there specific moments where we lose more viewers than usual?"

Claude calls get_retention_curve and identifies drop-off points: "Your retention curve is mostly normal until the 4:30 mark, where you lose about 22% of remaining viewers in 30 seconds. That is the point where you shifted topics -- the viewer who came for X may not want to hear about Y."

That is specific, actionable feedback. According to YouTube, average view duration is the single most important metric for the recommendation algorithm.

Step 6: Audit Your Video SEO

YouTube is the world's second-largest search engine. How well your videos are optimised for search affects both discovery and ranking.

Prompt:

"Run an SEO audit on my last 5 published videos. Which ones are well-optimised for search? What are the most common issues? What should I fix first?"

Claude calls get_video_seo_score for each video and identifies patterns: missing or thin descriptions, tag mismatches, titles that don't match how people search, missing chapters. Instead of auditing each video separately in YouTube Studio, you get a cross-video view of your SEO hygiene.

Step 7: Research Your Competition

Understanding what works in your niche -- not just on your channel -- provides intelligence for content decisions.

Prompt:

"Look at [competitor channel name/ID]. What are their metrics, and what content topics are performing best for them recently?"

Claude calls get_competitor_stats and synthesises: "This channel has 87,000 subscribers and averages 45,000 views per video. Their top-performing content in the last 90 days is weighted toward [topic category]. Videos with [specific element] consistently outperform their channel average."

Follow-up:

"Find content topics that this competitor covers successfully that I am not making yet. Show me my content gaps."

Claude calls find_content_gaps and returns a prioritised list of topics -- filtered by what works for competitors but is absent from your library. Directly actionable for content planning.

Step 8: Find Trending Topics

Timing content to rising trends can significantly boost performance, especially for search-indexed content.

Prompt:

"What topics are trending in [your niche] on YouTube right now? Which of these trends fit my channel, and which should I skip because they do not match my audience?"

Claude calls get_trending_topics and reasons about fit: "Four trending topics in your niche are currently rising. The first two fit your typical format well. The third trends toward a younger demographic than yours. The fourth is peaking -- you would need to publish within a week to catch it."

Tips for Better Analysis

Provide channel context upfront. "My channel covers personal finance for UK millennials, 45,000 subscribers, goal is 100K in 12 months" -- that context shapes every recommendation Claude gives.

Ask why, not just what. "What do my top videos have in common, and how do I apply that pattern to future content?" produces more useful output than "What are my top videos?"

Compare across timeframes. Single-period data can mislead. Always ask Claude to compare to a prior period so you see change, not just state.

Do not just look at views. Views are a lagging indicator. Impression CTR (thumbnail/title effectiveness), average view duration (attention retention), and subscriber conversion rate (casual-to-follower conversion) predict future growth better than raw view count.

Use audience activity data. Publishing timing is an overlooked lever. Aligning your schedule with when your audience is active improves early video performance, which YouTube uses to determine whether to push the video further.

Common Mistakes

Optimising for views at the expense of subscribers. A video getting 10,000 views with 500 new subscribers is more valuable for growth than a viral video with 100,000 views and 50 subscribers.

Ignoring the retention curve. Watch time matters less than average view duration as a percentage. A 5-minute video where 60% finish is healthier than a 15-minute video where 15% finish. YouTube explicitly prioritises percentage-based retention in its recommendation system.

Chasing trends that do not fit your audience. Trend-chasing can spike short-term views but alienate your core audience. Always filter trending topics through "does this match who already watches my channel?"

Not checking traffic sources when diagnosing problems. A video underperforming can have completely different solutions depending on whether the issue is low impressions (distribution/SEO), low CTR (thumbnail/title), or low watch time (content). Traffic sources tell you where in the funnel the problem sits.

What This Replaces

Doing this analysis manually means visiting multiple YouTube Studio sections, noting numbers, building comparison spreadsheets, and then reasoning about what it all means. A thorough channel analysis: 2-3 hours.

With Iris and Claude, a complete analysis covering all eight areas takes 30-45 minutes. Not because the analysis is shallower, but because data retrieval and initial synthesis happen instantly rather than through manual report navigation.

The time you save goes into actual thinking: deciding what content to make, what to improve, what to stop doing. That shift from data retrieval to strategic thinking is the real value.

Get started with Ooty Iris