YouTube Studio's Insights Tab: What Replaced Analytics, and the Eight Saved Reports Worth Building

If you opened Studio recently and found that Analytics had been replaced by something called Insights, nothing is broken and nothing has been taken away. YouTube has been rolling out a redesigned Studio since July 2026, and its own documentation warns that the analytics page may look different than before. Some channels have had the new layout for weeks. Plenty still have the old one. A few have the new overview and none of the new cards.

The reflex is to treat this as cosmetic and wait it out. That would be a mistake, because the redesign encodes a view of what analytics is for. The top of the page is getting simpler, more grouped and — for eligible channels — narrated by AI summaries. The bottom of the page, Advanced mode, is getting considerably more powerful, including the first genuinely new mechanic in Studio analytics for years: saved reports. Which layer you spend your time in will decide whether your channel decisions get sharper or vaguer over the next year.

This is the practical version: exactly what moved, what the AI cards can and cannot know, what to build in Advanced mode while the interface is still settling, and the questions the new Insights page still cannot answer. If you want the funnel-reading fundamentals rather than the interface changes, the channel audit walkthrough covers those in order; this post assumes you can read a retention curve and want to know where YouTube moved it.

What actually changed, in one table

Two separate waves are involved, and conflating them is why creator forums are confusing right now. The July wave rebuilt the analytics surface. The September announcements at Made on YouTube added tools on top of it, mostly in the mobile Studio app.

Change	Where	What it means in practice
Analytics renamed Insights	Studio, left menu	Same data, new label and new grouping of related metrics and charts
Regrouped overview	Insights overview	Related metrics sit together instead of in one long scroll
Back button and simpler date picker	Insights overview	Faster navigation; full real-time and custom ranges move into Advanced mode
Click any chart to go deeper	Insights overview	Charts are now doors into Advanced mode rather than dead ends
Four AI insight cards	Insights, Ask Studio channels only	Channel Summary, Content Patterns, Audience Loyalty, Video Summary
Trends becomes Research	Studio	The idea-finding surface is separated from the performance-reading one
New sidebar	Advanced mode	Filters, breakdowns and metric selection in one place
Search, sort, multi-metric	Advanced mode	Find a video by title, sort by length or publish date, add several metrics at once
Saved reports	Advanced mode	Up to 50 saved configurations, reopened in one click
Research feed, Insights and Research destinations	Studio mobile app	Outlier-spotting and performance review on the phone
Aggregated AdSense earnings	Studio	Earnings across linked accounts visible in one view

Coverage of the July test — including write-ups at Social Media Today and NetInfluencer — put the AI cards and the Research change in the same experiment as the rename, rolling out to a small group first. The Advanced mode upgrades appear to be travelling more broadly, and YouTube has published a Help Centre page on Advanced mode that reflects the new interface.

Why a rename is worth paying attention to

"Analytics" names a pile of data you interrogate. "Insights" names conclusions handed to you. That is not marketing pedantry; it is the design brief for the whole redesign, and you can see it in every individual change. Metrics are grouped so you scan rather than read. The date picker is simplified, with the fiddly real-time and custom ranges pushed down a level. AI cards summarise patterns before you have asked a question. All of it reduces the effort required to form an impression of your channel.

For most creators most weeks, that is a real improvement. The old overview asked every creator to be their own analyst, and the honest outcome was that many looked at views, glanced at CTR, felt something, and closed the tab. A grouped layout with a plain-language summary at the top beats that.

The risk is narrower and worth naming: a summary is a compression, and compressions lose exactly the thing that distinguishes a real finding from a coincidence. Two videos down 30% look identical in a summary sentence and completely different in a breakdown — one lost browse impressions, the other lost its click-through rate on the same impressions. Those two need opposite responses. The summary layer will tell you the channel softened. Only the second layer tells you which lever to pull.

The four AI cards, and what each one can know

The cards under test are Channel Summary, Content Patterns, Audience Loyalty and Video Summary, available to creators who already have access to Ask Studio. If you have been using Ask Studio, these will feel familiar: the same reading of the same data, except it arrives unprompted instead of as an answer to a question. The full breakdown of Studio's AI tools covers what Ask Studio is good for; the cards inherit both its strengths and its ceiling.

It helps to be precise about what kind of statement each card is capable of making.

Card	The question it can answer	The question it cannot
Channel Summary	What moved across the channel in this window	Whether the movement is a trend or variance
Content Patterns	What your better-performing videos have in common	Whether that commonality caused the performance
Audience Loyalty	How much of your watch time comes from people who come back	Which of your choices made them come back
Video Summary	How one video performed, in context	How it would have performed packaged differently

The pattern in the right-hand column is not a flaw in the models. It is the difference between description and counterfactual. Your analytics contain one history: the videos you made, packaged the way you packaged them. Nothing in that history contains the version you did not publish, so no amount of summarising produces it. When a card says your best videos are the ones with a face in the thumbnail, it has found a correlation in a sample of perhaps forty uploads, in which the face-thumbnail videos were probably also the ones you cared most about.

The practical rule: treat card output as a hypothesis generator with a fast read time, and treat the source of truth as the breakdown underneath. When a card says something surprising, click through. The redesign makes that click easy on purpose.

Advanced mode is where the redesign actually pays off

While the top of the page was being simplified, Advanced mode was rebuilt with the opposite intent. The changes, per YouTube's documentation and the coverage of the update, are all about getting to a specific view faster:

• One sidebar. Filters, breakdowns and metric selection now sit together on the left instead of being scattered across the top of the table.
• Search by title. You can find a video by typing part of its name rather than scrolling a catalogue. On a channel with three hundred uploads this is the difference between checking a hunch and not bothering.
• Sort by video length or publish date. Two sorts that make format questions answerable in one screen: do my twenty-minute videos hold watch time better than my eights, and has anything changed since I switched?
• Multi-metric selection. Add or remove several metrics at once instead of one click per column.
• Searchable breakdowns. Geography, content type, revenue source and the rest are findable by typing rather than by remembering which menu they hide in.

None of that is glamorous. Together it removes most of the friction that stopped creators from looking past the overview, which matters because the overview is where the least actionable numbers live.

Saved reports: fifty slots, and how not to waste them

The genuinely new mechanic is saving a configured report — your filters, breakdowns and metrics — and reopening it later. Coverage of the update puts the ceiling at 50 saved reports per channel, which is far more than any channel needs and exactly the sort of allowance that invites clutter.

The discipline that makes this feature valuable is simple: a saved report is a question, not a view. Name it as a question. If you cannot phrase what the report is for in a sentence, do not save it. Six reports you read every week are worth more than thirty you built once in enthusiasm and never open, because the second set slowly trains you to skim a menu instead of reading data.

It is also worth deciding, once, which metrics your saved reports are built on. Since the August 2026 change to what counts as a public view, views tick up faster than they used to and historical counts were not restated, which makes any year-over-year comparison on views partly a comparison of definitions. Build your standing reports on engaged views, watch time and returning-viewer metrics, all of which are continuous across that change.

Eight saved reports worth building

These are the configurations that repay a slot. Each one is phrased as the question it answers, because that is how you should name them.

Report	Breakdown	Metrics	Read it
Which surface is feeding me?	Traffic source type	Impressions, CTR, views, watch time	Weekly
Did packaging or distribution soften?	Traffic source type, last 28 days	Impressions, CTR	When views drop
Does length still work for me?	Video, sorted by duration	Average view duration, average percentage viewed	Monthly
Who is actually coming back?	New vs returning viewers	Views, watch time	Monthly
Are subscribers watching, or just subscribed?	Subscription status	Views, average view duration	Monthly
Is the living room growing?	Device type	Views, watch time	Quarterly
Where is the audience I am not making for?	Geography	Views, watch time	Quarterly
How did the last five launches open?	Video, sorted by publish date	Impressions, CTR, engaged views	After each upload

Two of these deserve a note. The traffic-source report is the one to build first, because a click-through rate only means something inside a single traffic source — a 4% CTR on browse and a 4% CTR on search describe two unrelated situations. The traffic sources explainer works through why, and what counts as a good CTR covers the benchmark question that everyone asks and nobody can answer channel-agnostically.

The launch report is the one creators skip and then regret. Publishing is the only moment where your packaging is tested against a cold audience under roughly repeatable conditions, and if you are not capturing the first-days numbers in a consistent shape, you have no basis for comparing this thumbnail approach to last month's.

Groups: the report shape almost nobody sets up

Alongside saved reports sits an older feature that becomes far more useful now that a configuration can persist. A group is a collection of your own videos, playlists or channels — YouTube's documentation puts the cap at 500 items per group — that you can analyse as a single unit.

Groups turn vague format questions into measurable ones. Put your tutorials in one group and your commentary in another, and the difference in average view duration stops being a feeling. Group by thumbnail approach — face versus object, text versus no text, your current template versus the one before it — and you get something close to a natural experiment across your own catalogue. It is not a controlled test, because the videos differ in more ways than the thumbnail, but on a catalogue of any size it is enough to disqualify a bad habit.

Three groups worth having from the start: format, thumbnail style, and the videos you consider your best work regardless of performance. The third is the one that teaches you the most, because the gap between what you are proud of and what performs is the actual packaging problem.

Reading a packaging question in the new layout

Say a video opened badly. In the new layout the sequence is short, and it is worth committing to memory before the interface tempts you into a longer one.

1. Impressions first. If YouTube never showed the video, no thumbnail decision is implicated. You have a distribution problem — or a video that lives on search and has not accumulated yet.
2. Then CTR, inside one source. Impressions delivered but clicks low, on browse specifically, is the one pattern that genuinely indicts your packaging.
3. Then the first thirty seconds. Clicks delivered but retention collapsing early usually means the thumbnail and title promised something the opening did not pay off. That is still a packaging finding; it is just a mismatch rather than a weak image.
4. Then the traffic mix against your own baseline. A video that got a different mix from your usual one was shown to a different audience, and its numbers are not comparable to your catalogue average.

What the new layout genuinely improves is step two, because clicking the CTR figure now takes you straight into the breakdown instead of leaving you to find the equivalent view yourself. What it does not improve is the judgement in step three, which is not in the data at all.

What the redesign does not fix

Every version of Studio analytics shares one limit, and it is the limit that matters most for the work this blog is mostly about. Studio reports what happened to the thumbnail you published. It has no opinion on the one you did not. A card that says your CTR fell is describing an outcome with at least four plausible causes — the image, the title, the topic, or who YouTube happened to show it to — and it cannot separate them, because separating them requires two versions running against comparable audiences at the same time.

That is what Test & Compare exists for, and it is the only part of Studio that produces a genuine comparison rather than a history. The guide to thumbnail testing in Studio covers how to run one so the result means something, including the sample-size problem that makes most informal tests worthless.

Two smaller things the redesign also leaves untouched. First, the data still needs time to settle: the real-time figures cover a short recent window and are explicitly estimates, so a report read at hour six of a launch is reading noise. Second, summaries do not know your intent. A video that underperformed on purpose — an experiment, a favour, a format probe — will be summarised as a problem, and only you know it was not.

If you do not have the new layout yet

Nothing here requires waiting. Advanced mode has existed for years, groups have existed for longer, and every report in the table above can be built in the old interface — you will just configure it each time instead of saving it, which is an argument for keeping a note of the configurations rather than for skipping them.

The one thing worth doing in advance is settling on your metric set. When saved reports arrive, the creators who benefit are the ones who already know which six questions they ask every month. The ones who do not will fill their slots with whatever the interface put in front of them.

It is also worth knowing that the mobile side is moving in the same direction. At Made on YouTube in September 2026, YouTube described new Insights and Research destinations in the Studio app, including a research feed for spotting what is working on the platform and outlier videos worth reacting to — the phone version of the same split between reading performance and finding ideas. If outliers are the part you care about, the outlier multiplier explainer covers how far that number can be trusted.

A routine that fits the new shape

The redesign rewards a smaller, more repetitive habit than most creators have. Something like this, once the reports exist:

• Two minutes, the day after publishing. Open the launch report. Impressions, CTR, engaged views. Write the three numbers down somewhere permanent. Do not act on them.
• Ten minutes, weekly. Traffic source report, then the AI summary if you have it, in that order. Data before narrative, so the narrative has something to check against.
• Forty minutes, monthly. The length, loyalty and subscription-status reports, plus your groups. This is where format decisions belong — not in the week after one soft upload.
• Quarterly. Device and geography, and a look at whether your saved reports are still the questions you care about. Delete the ones you have stopped opening.

The discipline of writing launch numbers down and not acting on them is the part most likely to be skipped and the part that pays. Weekly variance on almost every channel is larger than the effect of anything you will change in response to a single video, and a pattern across four or five uploads in the same format is the smallest thing worth calling a finding.

Frequently asked questions

Where did the Analytics tab go?

It was renamed Insights as part of the Studio update that began rolling out in July 2026. The data is the same and Advanced mode is still there, reachable by clicking into any chart or metric.

Is the underlying data different now?

No. The redesign changes presentation, grouping and navigation. The one recent change to a metric definition is separate: since 24 August 2026 a public view is recorded as soon as playback starts, while earnings and Partner Programme eligibility continue to run on engaged views.

Why can I not see the AI insight cards?

They require Ask Studio access, and the redesign has been testing with a limited group of channels. Both gates have to open before the cards appear, which is why some creators have the new layout without them.

Should I trust what the cards tell me?

Trust them to describe. A card reading a real pattern in your data is doing something useful and fast. Do not treat "videos like this perform better" as a causal claim about a sample of a few dozen uploads, and click through to the breakdown before you change your format on one.

Do saved reports replace exporting to a spreadsheet?

For most channels, yes — a saved report reopens the exact view you wanted without a download. Exporting still earns its place for anything you want to keep as a record outside Studio, or for combining channel data with numbers YouTube does not hold, such as your own publishing calendar.

Did the Research tab change too?

The test moves the Trends surface into Research, separating idea-finding from performance-reading. It is the same direction as the mobile research feed announced in September: one place to read what happened, another to decide what to make.

The useful conclusion

Studio is becoming easier to glance at and more powerful to interrogate, in the same release. Those two things pull in opposite directions, and which one you get depends entirely on habit. A creator who reads the summary and closes the tab will get a slightly better-designed version of the shallow relationship they already had with their analytics. A creator who builds six saved reports and reads them on a schedule will, within a quarter, know things about their channel that no summary would have surfaced.

The gap between those two outcomes is perhaps twenty minutes of setup. Name your questions, build the reports that answer them, put your launch numbers somewhere permanent, and let the AI cards do what they are genuinely good at — telling you quickly that something moved, so you know which breakdown to open.

And when the breakdown points at packaging, which on a browse-dependent channel it frequently will, the fix is upstream of any report. Studio can tell you that a thumbnail did not earn its impressions; it cannot draw you a better one. Thumblore generates variants from a prompt or a frame in about a minute, which is what makes a real test — two genuinely different approaches rather than the same image twice — cheap enough to run on every upload. The reading is what tells you which one to keep.