YouTube Upgrades Shorts Editor With Timeline View via @sejournal, @MattGSouthern

YouTube is launching an all-new timeline editor for Shorts that puts video clips, overlays, and audio in one unified view.

The update responds to creator requests for more precise controls inside the YouTube app.

What’s New

The editor adds timeline-level controls you can use without switching modes. You can trim and reorder clips via drag-and-drop and use zoom for precise timing and transitions.

YouTube states in a video announcement:

“When you’re creating a short, you’ve told us how critical it is to be able to make all of your edits to a video in one place to be able to really make something you’re proud of. We appreciate the feedback and we’ve been listening. So, we’re launching an all new Shorts timeline editor. Now, everything is visible in one place. all of your video clips, overlays, and audio. You can trim, reorder the clips, simple drag and drop. You can zoom in to make precise edits.”

See it in action in the video below:

Planned Additions

YouTube says it plans to add slip editing, clip splitting, and the ability to add media directly from the timeline. No release window was provided in the video.

YouTube adds:

“And this is just the beginning. We’re making a lot more key improvements, for example, like being able to do slip editing, being able to do splitting, and adding media directly from timeline.”

YouTube also said it will continue expanding Edit with AI, the Gemini-assisted tool that can assemble a first-draft edit with music, transitions, and voiceover.

While integration with the new timeline editor is planned, YouTube didn’t share timing or regional availability.

YouTube said:

“We also plan to continue to expand Edit with AI… to make it even easier for you to craft your vision with Gemini Assistance.”

Why This Matters

A unified timeline brings more of the desktop editing experience into the YouTube app.

The single view reduces mode switching, and the upcoming slip and split controls should improve pacing and transitions without relying on third-party editors.

As YouTube put it:

“The goal is really simple. More creative freedom with less friction. We’re really committed to building easy to use creation tools that help you be your most creative self and make something you’re really proud of.”

Looking Ahead

YouTube didn’t specify when the timeline editor will reach all creators or which regions will get it first.

These updates fit into YouTube’s broader push to enhance Shorts creation tools, following earlier improvements like beat syncing, templates, and AI-generated stickers.


Featured Image: FotoField/Shutterstock

First things first: writing content with the inverted pyramid style

Journalists have been using the inverted pyramid writing style for ages. Using it, you put your most important information upfront. Don’t hedge. Don’t bury your key point halfway down the third paragraph. And don’t hold back; tell the complete story in the first paragraph. Even online, this writing style holds up pretty well for some types of articles. It even comes in handy now that web content is increasingly used to answer every type of question a searcher might have. Find out how!

Table of contents

Key takeaways

  • The inverted pyramid writing style places crucial information at the beginning to engage readers quickly and effectively.
  • Writers should structure articles with core sentences that introduce key concepts to aid comprehension and improve scanning.
  • This style enhances SEO by making content clearer and easier to understand for both human readers and search engines.
  • While effective for many types of articles, the inverted pyramid may not suit creative writing forms like poetry or complex fiction.
  • To implement the inverted pyramid, identify key points, structure your content, and revise for clarity and focus.

What is the inverted pyramid?

Most readers don’t have the time or desire to carefully read an article, so journalists put the critical pieces of a story in the first paragraph to inform and draw in a reader. This paragraph is the meat and potatoes of a story, so to speak. This way, every reader can read the first paragraph, or the lead, and get a complete notion of what the story is about. It gives away the traditional W’s instantly: who, what, when, where, why, and, of course, how.

The introductory paragraph is followed by paragraphs that contain important details. After that, follows general information and whatever background the writers deem supportive of the narrative. This has several advantages:

  • It supports all readers, even those who skim
  • It improves comprehension; everything you need to understand the article is in that first paragraph
  • You need less time to get to the point
  • It gives writers a full paragraph to draw readers in
  • Done well, it encourages readers to scroll and read the rest of the article
  • It gives writers full control over the structure
  • It makes it easier to edit articles

An example

Here’s an example of such an intro. We wrote an article about writing meta descriptions in Yoast SEO that answers exactly that question in an easy-to-understand way. We show what it is and why it’s important immediately, while also triggering people to read the rest of the article. Here’s the intro:

“A strong meta description boosts CTR and signals relevance to search engines. This post shows how to craft descriptions that work, with practical tips and ready-to-use templates. You’ll learn the traits of good meta descriptions, common mistakes, and how Yoast SEO can help you get it right. Using these templates and guidelines can boost CTR, align reader expectations, and improve optimization for both users and Google.”

The inverted pyramid is just one of many techniques for presenting and structuring content. Like us, you can use it to write powerful news articles, press releases, product pages, blog posts, or explanatory articles.

This style of writing, however, is not suited for every piece of content. Maybe you write poetry, or long essays with a complete story arc, or just a piece of complex fiction. Critics are quick to add that the inverted pyramid style cripples their creativity. But, even then, you can learn from the techniques of the inverted pyramid that help you to draw a reader in and figure out a good way to structure a story. And, as we all know, a solid structure is key to getting people and search engines to understand your content. We wrote about that in our article on setting up a clear text structure.

The inverted pyramid

The power of paragraphs

Well-written paragraphs are incredibly powerful. These paragraphs can stand on their own. I always try to write in a modular way. That’s because I’m regularly moving paragraphs around if I think they fit better somewhere else in the article. It makes editing and changing the structure of a story so much easier.

Good writers give every paragraph a stand-out first sentence; these are known as core sentences. These sentences raise one question or concept per paragraph. So, someone who scans the article by reading the first sentence of every paragraph will get the gist of it and can choose to read the rest of the paragraph or not. Of course, the rest of the paragraph is spent answering or supporting that question or concept.

The pyramid, SEO, and AI

Front-loading the main point helps SEO perform in an AI era. Lead with the core result to give readers a fast, clear understanding and to signal relevance to search algorithms. Focusing on that idea makes snippets more likely and improves relevance while making the rest of the piece easier to scan, summarize, and reuse across channels. In practice, the inverted pyramid anchors the article in intent, guiding humans and machines toward the same destination: the core answer.

Answering questions

Something else is going on: a lot of content out there is written specifically to answer questions based on user intent. Today, Google answers a lot of questions and answers right away in the search results. That’s why it makes a lot of sense to structure your questions and answers in such a way that is easy to digest for both readers and search engines. This also supports the inverted pyramid theory. So, if you want to answer a specific question, do that right beneath that question. Don’t obfuscate it. Keep it upfront. You can answer supporting questions or give a more elaborate answer further down the text. If you have data supporting your answer, please present it.

Summaries vs. the pyramid

Front-loading the main point highlights the core idea clearly to both readers and search engines. The inverted pyramid delivers that headline idea first, then adds context and support. A summary condenses the piece into its essential takeaways, handy for meta descriptions, snippets, or quick recaps. Yoast AI Summarize can generate tight summaries from your content, giving you ready-to-use openings and meta descriptions that align with the pyramid and improve SEO performance.

How to write with the inverted pyramid in mind

The inverted pyramid forces you to think about your story: what is it, and which parts are key to understanding everything? Even if you don’t follow the structure to the letter, focusing on the essential parts of your story and deleting the fluff is always a good thing. In his seminal work The Elements of Style, William Strunk famously wrote:

“Vigorous writing is concise. A sentence should contain no unnecessary words, a paragraph no unnecessary sentences, for the same reason that a drawing should have no unnecessary lines and a machine no unnecessary parts. This requires not that the writer make all his sentences short, or that he avoid all detail and treat his subjects only in outline, but that he make every word tell.”

In short, writing works like this:

  • Map it out: What are the most important points you want to make?
  • Filter: Which points are supportive, but not key?
  • Connect: How does everything fit together?
  • Structure: Use sub-headers to build an easy-to-understand structure for your article
  • Write: Start every paragraph with your core sentence and support/prove/disprove/etc in the coming sentences
  • Revise: Are the paragraphs in the correct order? Maybe you should move some around to enhance readability or understanding?
  • Edit: I.e., killing your darlings. Do you edit your own work, or can someone do it for you?
  • Publish: Add the article to WordPress and hit that Publish button

Need more writing tips? Here are 10 tips for writing an awesome and SEO-friendly blog post.

Try the inverted pyramid

Like we said, not every type of content will benefit from the inverted pyramid. But the inverted pyramid has surely made its mark over the past century or more. Even now, as we mostly write content for the web, this type of thinking about a story or article makes us focus on the most important parts, and how we tell about those parts. It forces you to separate facts from fiction and fluff from real nuggets of content gold. So, try it out, and your next article might turn out to be the best yet.

Read more: SEO copywriting: the ultimate guide »

Google Tests “Analytics Advisor” Inside GA4, According To Reports via @sejournal, @MattGSouthern

Some GA4 users are reporting that a new conversational interface called Analytics Advisor is appearing inside Google Analytics.

The sighting comes from analytics consultant Himanshu Sharma, who posted screenshots on X:

What We Know So Far

A welcome card in the screenshot lists three functions:

  1. Answering data questions with single-number answers and visualizations.
  2. Pointing you to the most appropriate report.
  3. Providing educational guidance based on relevant documentation.

The interface appears as a chat entry point inside a GA4 property.

Google hasn’t issued a help-center article or “What’s New” note for Analytics Advisor, and there is no formal announcement detailing availability, eligibility, or timelines.

Background

At Google Marketing Live, Google described agentic AI tools coming to Ads and Analytics, including a “data expert” for Analytics that surfaces insights and helps troubleshoot issues.

Google said:

“In Google Analytics, the data expert will proactively show insights and trends, plus enable easy data exploration with simple visuals for improved decision-making and performance. These new capabilities will also help marketers troubleshoot campaign issues.”

The company also previewed a separate Marketing Advisor agent for marketers. Analytics Advisor in GA4 would be consistent with that direction, though Google hasn’t connected these names publicly.

Why It Matters

If this interface rolls out broadly, it could lower the effort required to get quick answers from GA4, especially for teams still learning the product’s structure.

The transparent, step-by-step answers shown in the screenshot may also help you check how an answer was derived before acting on it.

What We Don’t Know Yet

Key details remain unclear, including:

  • Rollout scope, property eligibility, and regional availability
  • Whether this is an experiment, staged rollout, or permanent feature
  • Data handling, controls, and any opt-out settings

We will update this story if Google publishes documentation or confirms details.

Looking Ahead

Check your GA4 properties for a new chat icon or welcome card labeled Analytics Advisor.

If it appears, test on non-sensitive questions first and compare responses to your existing reports before relying on it for decisions.


Featured Image: Mamun_Sheikh/Shutterstock

Meta Ads Pixel Tracking, Explained

The Meta Pixel is a snippet of JavaScript with powerful implications for advertising performance.

Ecommerce operators are typically experts in their products or industry, but not in advertising. I’ve seen conversion campaigns, for instance, with poor performance because the Meta Pixel was firing on the wrong event.

One merchant went from a 0.3% conversion rate to 27.0% with the correct setup. The difference was not the ad creative, audience targeting, or page layout. It was pixel event tracking.

Pixel Mechanics

The Meta Pixel connects a website and the Meta Ads platform. The process goes like this:

  • A Facebook or Instagram user clicks on an advertisement.
  • The pixel associates the ad click with the website visit.
  • When the visitor completes an action (e.g., a purchase, abandoned cart, page view), the pixel reports back to Meta Ads.
  • Meta learns from that data who is most likely to buy.
  • Meta Ads targets those prospects.

Put another way, the pixel tracks the first purchase, and Meta Ads uses that data to optimize who then sees the ads and the headlines, descriptions, and images.

With each cycle of action and learning, Meta Ads gets better at finding customers for the store.

Better Than Billboards

Imagine a billboard on a busy highway. Thousands of cars pass by each day. But an advertiser on that billboard can only guess or estimate how many folks see or think about the message.

The business does not know:

  • The number of drivers or passengers who looked at it,
  • The billboard’s impact on sales,
  • The most effective part of the billboard (image or text).

Certainly a unique message on the billboard can facilitate tracking, but not with the clarity of a pixel.

Shopify

Shopify offers a plugin for Meta Ads. It also works with Facebook and Instagram Shops.

Screenshot of Shopify's page promoting the Facebook and Instagram integrations

The Facebook & Instagram app for Shopify tightly integrates with Meta Ads.

The plugin uses two methods to connect a Shopify store to Meta Ads. (It also works with Facebook and Instagram Shops.) First is the Meta Pixel. Second is Meta’s Conversion API. Both share similar events from Shopify to Meta, including:

  • PageView,
  • ViewContent,
  • Search,
  • AddToCart,
  • InitiateCheckout,
  • AddPaymentInfo,
  • Purchase.

The pixel sends the data collected in visitors’ web browsers. The API captures events from Shopify directly. Together, the pair answers questions that a billboard could not, such as:

  • What is working (measurement)? For example, 1,000 consumers saw an ad, 50 clicked it, and five made a purchase, resulting in a specific return on ad spend.
  • Who to target next (optimization)? Meta’s algorithm analyzes the five folks who purchased and then automatically shows the ad to millions of others with similar characteristics and behaviors, dramatically increasing ad efficiency.
  • How to improve (insight)? Meta Ads constantly refines targeting and creative combinations based on actual sales, not intuition.

Privacy

A pixel’s potency stems from tracking consumers. Meta (Facebook) has long been a source of privacy concerns, since the company collects loads of personally identifiable information across devices.

When it adds a pixel to its web store, an advertiser essentially invites Meta to track visitors’ on-site behavior — the ads they click, the pages they view, the products they add to the cart, and what they buy.

All of that data goes to Meta’s advertising network, which combines it with information from other sites, apps, and internal portals such as Facebook and Instagram.

The result is remarkably better ad performance, coupled with the need for heightened awareness by advertisers.

Understanding both ad performance and privacy implications allows merchants to make informed decisions, using data wisely without losing the trust of the customers who make businesses possible.

This test could reveal the health of your immune system

Attentive readers might have noticed my absence over the last couple of weeks. I’ve been trying to recover from a bout of illness.

It got me thinking about the immune system, and how little I know about my own immune health. The vast array of cells, proteins, and biomolecules that works to defend us from disease is mind-bogglingly complicated. Immunologists are still getting to grips with how it all works.

Those of us who aren’t immunologists are even more in the dark. I had my flu jab last week and have no idea how my immune system responded. Will it protect me from the flu virus this winter? Is it “stressed” from whatever other bugs it has encountered in the last few months? And since my husband had his shot at the same time, I can’t help wondering how our responses will compare. 

So I was intrigued to hear about a new test that is being developed to measure immune health. One that even gives you a score.

Writer David Ewing Duncan hoped that the test would reveal more about his health than any other he’d ever taken. He described the experience in a piece published jointly by MIT Technology Review and Aventine.

The test David took was developed by John Tsang at Yale University and his colleagues. The team wanted to work out a way of measuring how healthy a person’s immune system might be.

It’s a difficult thing to do, for several reasons. First, there’s the definition of “healthy.” I find it’s a loose concept that becomes more complicated the more you think about it. Yes, we all have a general sense of what it means to be in good health. But is it just the absence of disease? Is it about resilience? Does it have something to do with withstanding the impact of aging?

Tsang and his colleagues wanted to measure “deviation from health.” They looked at blood samples from 228 people who had immune diseases that were caused by single-gene mutations, as well as 42 other people who were free from disease. All those individuals could be considered along a health spectrum.

Another major challenge lies in trying to capture the complexity of the immune system, which involves hundreds of proteins and cells interacting in various ways. (Side note: Last year, MIT Technology Review recognized Ang Cui at Harvard University as one of our Innovators under 35 for her attempts to make sense of it all using machine learning. She created the Immune Dictionary to describe how hundreds of proteins affect immune cells—something she likens to a “periodic table” for the immune system.)

Tsang and his colleagues tackled this by running a series of tests on those blood samples. The vast scope of these tests is what sets them apart from the blood tests you might get during a visit to the doctor. The team looked at how genes were expressed by cells in the blood. They measured a range of immune cells and more than 1,300 proteins.

The team members used machine learning to find correlations between these measurements and health, allowing them to create an immune health score for each of the volunteers. They call it the immune health metric, or IHM.

When they used this approach to find the immune scores of people who had already volunteered in other studies, they found that the IHM seemed to align with other measures of health, such as how people respond to diseases, treatments, and vaccines. The study was published in the journal Nature Medicine last year.

The researchers behind it hope that a test like this could one day help identify people who are at risk of cancer and other diseases, or explain why some people respond differently to treatments or immunizations.

But the test isn’t ready for clinical use. If, like me, you’re finding yourself curious to know your own IHM, you’ll just have to wait.

This article first appeared in The Checkup, MIT Technology Review’s weekly biotech newsletter. To receive it in your inbox every Thursday, and read articles like this first, sign up here.

How do our bodies remember?

MIT Technology Review Explains: Let our writers untangle the complex, messy world of technology to help you understand what’s coming next. You can read more from the series here.

“Like riding a bike” is shorthand for the remarkable way that our bodies remember how to move. Most of the time when we talk about muscle memory, we’re not talking about the muscles themselves but about the memory of a coordinated movement pattern that lives in the motor neurons, which control our muscles. 

Yet in recent years, scientists have discovered that our muscles themselves have a memory for movement and exercise.

When we move a muscle, the movement may appear to begin and end, but all these little changes are actually continuing to happen inside our muscle cells. And the more we move, as with riding a bike or other kinds of exercise, the more those cells begin to make a memory of that exercise.

When we move a muscle, the movement may appear to begin and end, but all these little changes are actually continuing to happen inside our muscle cells.

We all know from experience that a muscle gets bigger and stronger with repeated work. As the pioneering muscle scientist Adam Sharples—a professor at the Norwegian School of Sport Sciences in Oslo and a former professional rugby player in the UK—explained to me, skeletal muscle cells are unique in the human body: They’re long and skinny, like fibers, and have multiple nuclei. The fibers grow larger not by dividing but by recruiting muscle satellite cells—stem cells specific to muscle that are dormant until activated in response to stress or injury—to contribute their own nuclei and support muscle growth and regeneration. Those nuclei often stick around for a while in the muscle fibers, even after periods of inactivity, and there is evidence that they may help accelerate the return to growth once you start training again. 

Sharples’s research focuses on what’s called epigenetic muscle memory.Epigenetic” refers to changes in gene expression that are caused by behavior and environment—the genes themselves don’t change, but the way they work does. In general, exercise switches on genes that help make muscles grow more easily. When you lift weights, for example, small molecules called methyl groups detach from the outside of certain genes, making them more likely to turn on and produce proteins that affect muscle growth (also known as hypertrophy). Those changes persist; if you start lifting weights again, you’ll add muscle mass more quickly than before.

In 2018, Sharples’s muscle lab was the first to show that human skeletal muscle has an epigenetic memory of muscle growth after exercise: Muscle cells are primed to respond more rapidly to exercise in the future, even after a monthslong (and maybe even yearslong) pause. In other words: Your muscles remember how to do it.

Subsequent studies from Sharples and others have replicated similar findings in mice and older humans, offering further supporting evidence of epigenetic muscle memory across species and into later life. Even aging muscles have the capacity to remember when you work out.

At the same time, Sharples points to intriguing new evidence that muscles also remember periods of atrophy—and that young and old muscles remember this differently. While young human muscle seems to have what he calls a “positive” memory of wasting—“in that it recovers well after a first period of atrophy and doesn’t experience greater loss in a repeated atrophy period,” he explains—aged muscle in rats seems to have a more pronounced “negative” memory of atrophy, in which it appears “more susceptible to greater loss and a more exaggerated molecular response when muscle wasting is repeated.” Basically, young muscle tends to bounce back from periods of muscle loss—“ignoring” it, in a sense—while older muscle is more sensitive to it and might be more susceptible to further loss in the future. 

Illness can also lead to this kind of “negative” muscle memory; in a study of breast cancer survivors more than a decade after diagnosis and treatment, participants showed an epigenetic muscle profile of people much older than their chronological age. But get this: After five months of aerobic exercise training, participants were able to reset the epigenetic profile of their muscle back toward that of muscle seen in an age-matched control group of healthy women.  

What this shows is that “positive” muscle memories can help counteract “negative” ones. The takeaway? Your muscles have their own kind of intelligence. The more you use them, the more they can harness it to become a lasting beneficial resource for your body in the future. 

Bonnie Tsui is the author of On Muscle: The Stuff That Moves Us and Why It Matters (Algonquin Books, 2025).

The Download: our bodies’ memories, and Traton’s electric trucks

This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.

How do our bodies remember?

“Like riding a bike” is shorthand for the remarkable way that our bodies remember how to move. Most of the time when we talk about muscle memory, we’re not talking about the muscles themselves but about the memory of a coordinated movement pattern that lives in the motor neurons, which control our muscles.

Yet in recent years, scientists have discovered that our muscles themselves have a memory for movement and exercise. And the more we move, as with riding a bike or other kinds of exercise, the more those cells begin to make a memory of that exercise. Read the full story.

—Bonnie Tsui

This piece is part of MIT Technology Review Explains: our series untangling the complex, messy world of technology to help you understand what’s coming next. You can read more from the series here.

This story is also from our forthcoming print issue, which is all about the body. If you haven’t already, subscribe now to receive future issues once they land. Plus, you’ll also receive a free digital report on nuclear power.

2025 climate tech companies to watch: Traton and its electric trucks

Every day, trucks carry many millions of tons of cargo down roads and highways around the world. Nearly all run on diesel and make up one of the largest commercial sources of carbon emissions.

Traton, a subsidiary of Volkswagen, is producing zero-emission trucks that could help clean up this sector, while also investing in a Europe-wide advanced charging network so other manufacturers can more easily follow suit. Read the full story.

—Amy Nordrum

Traton is one of our 10 climate tech companies to watch—our annual list of some of the most promising climate tech firms on the planet. Check out the rest of the list here.

This test could reveal the health of your immune system

We know surprisingly little about our immune health. The vast array of cells, proteins, and biomolecules that works to defend us from disease is mind-bogglingly complicated. Immunologists are still getting to grips with how it all works.

Now, a new test is being developed to measure immune health, one that even gives you a score. But that’s a difficult thing to do, for several reasons. Read the full story.

—Jessica Hamzelou

This article first appeared in The Checkup, MIT Technology Review’s weekly biotech newsletter. To receive it in your inbox every Thursday, sign up here.

The must-reads

I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.

1 China is cracking down on imports of Nvidia’s AI chips 
Customs officers are combing shipments looking for the company’s China-specific chips. (FT $)
+ US officials are investigating a firm that’s suspected of helping China sidestep export restrictions. (NYT $)

2 Tesla’s ‘full self-driving’ feature is under investigation
After multiple reports of vehicles using it ran red lights. (WP $)
+ The company is slashing its prices to compete with Chinese giant BYD. (Rest of World)
+ Elon Musk will still receive billions, even if he fails to achieve his ambitions goals. (Reuters)

3 A data hoarder has created a searchable database of Epstein files
Making it simple to find mentions of specific people and locations. (404 Media)

4 OpenAI says GPT-5 is its least-biased model yet
Even when proceeding with “challenging, emotionally charged prompts.” (Axios)

5 The developers behind ICE-tracking apps aren’t giving up
They’re fighting Apple’s decision to remove their creations from its app store. (Wired $)
+ Another effort to track ICE raids was just taken offline. (MIT Technology Review)

6 The world’s biodiversity crisis is worsening
More than half of all bird species are in decline. (The Guardian)
+ The short, strange history of gene de-extinction. (MIT Technology Review)

7 YouTube is extending an olive branch to banned creators
It’s overturned a lifetime ban policy to give the people behind previously-banned channels a second chance. (CNBC)
+ But users kicked off for copyright infringement or extremism aren’t eligible. (Bloomberg $)

8 This startup wants to bring self-flying planes to our skies  
Starting with military cargo flights. (WSJ $)

9 Your plumber might be using ChatGPT
They’re increasingly using the chatbot to troubleshoot on the ground. (CNN)

10 Do robots really need hands?
Maybe not, but that’s not standing in the way of researchers trying to recreate them. (Fast Company $)
+ Will we ever trust robots? (MIT Technology Review)

Quote of the day

“Social media is a complete dumpster.”

—Hany Farid, a professor of computer science at the University of California, Berkeley, describes the proliferation of AI slop videos infiltrating digital platforms to the New York Times.

One more thing

Who gets to decide who receives experimental medical treatments?

There has been a trend toward lowering the bar for new medicines, and it is becoming easier for people to access treatments that might not help them—and could even harm them. Anecdotes appear to be overpowering evidence in decisions on drug approval. As a result, we’re ending up with some drugs that don’t work.

We urgently need to question how these decisions are made. Who should have access to experimental therapies? And who should get to decide? Such questions are especially pressing considering how quickly biotechnology is advancing. We’re not just improving on existing classes of treatments—we’re creating entirely new ones. Read the full story.

—Jessica Hamzelou

We can still have nice things

A place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet ’em at me.)

+ I love this crowd-sourced compendium of every known Wilhelm scream in all sorts of media.
+ Happy birthday to pocket rocket Bruno Mars, who turned 40 this week.
+ Here’s how to visit an interstellar interloper.
+ Bumi the penguin is having the absolute time of their life with this bubble machine 🐧

Subscription Pro on Growth, Churn, LTV

For 12 years Andrei Rebrov managed infrastructure and operations at Scentbird, a perfume subscription company he co-founded in 2013. He learned the importance of acquiring the right subscribers, those who stay and generate lifetime value for the business.

The key, he says, was accurate, timely analytics to assess channels, creative, and promos. Finsi, his new company, provides those metrics, enabling merchants to predict a prospect’s value over the long term.

In our recent conversation, I asked Andrei to share acquisition tactics, churn avoidance, product selection, and more.

Our entire audio dialog is embedded below. The transcript is edited for length and clarity.

Eric Bandholz: Tell our listeners who you are and what you do.

Andrei Rebrov: I’m the co-founder of Finsi, an analytics platform for subscription-based businesses, launched in 2024. We help companies acquire and retain profitable subscribers.

Before that, I spent 12 years building and scaling Scentbird, a perfume subscription service, where I served as CTO.

I handled much of the engineering, including coding the website, building back-office systems, and managing online payments and warehouse infrastructure. We launched in August 2014 and surpassed 1 million subscribers by the end of 2024. I left the company in March 2025.

We started Scentbird alongside subscription pioneers such as cosmetic brands Ipsy and Birchbox, and apparel provider Fabletics. We were inspired by Warby Parker’s “try before you buy” model, and we applied the concept to fragrances. We built our own platform, which gave us flexibility and scalability over the years.

We began with fragrances from other brands. Some were hesitant, but over time, Scentbird became a mutually beneficial partner, giving brands access to younger audiences, online shoppers, and consumers who wanted to try before committing to a full bottle. Our website’s motto became: “Date your fragrance before you marry it.”

Customers could select their monthly fragrances or receive a default “fragrance of the month.” Thousands chose the default, enabling the rapid collection of reviews and insights that brands could use to refine formulas, marketing copy, and strategies.

Subscription businesses require nonstop acquisition to stay in place. The challenge isn’t just reducing churn; it’s acquiring customers who will stay. Most SaaS companies separate acquisition and retention teams, which can create disconnects. Success comes from collaboration — aligning acquisition, retention, and operations — so the entire company functions as one system. Increasing customer acquisition spend is usually worthwhile if it improves lifetime value.

We were vertically integrated, which meant that fulfillment, logistics, and marketing had to move together. If one team outpaced the others, something would break quickly.

Bandholz: What drives profitable acquisition?

Rebrov: Accurate analytics. It’s one of the hardest parts of running a subscription business, and it’s a big reason I started Finsi. At Scentbird, we invested early in analytics because every acquisition channel behaves differently. Each has its own lifetime value, payback period, and acquisition cost, so analyzing them separately was essential.

We needed to understand what customers purchased through each channel and how these purchases affected retention. Traditional LTV calculations rely on historical data, which is typically dated. That delay makes it impossible to know if current strategies are working. To solve this, we built predictive LTV models that provided early insight — often within a month — so we could gauge the impact of new creatives and A/B tests faster.

For example, we tested a two-product-per-month plan. It initially lowered conversion rates, but predictive data revealed much stronger long-term value. That insight helped justify warehouse adjustments for the new fulfillment process.

We explored various customer acquisition channels. TikTok Shops became a top performer. Since it integrated only through Shopify, we built a faceless Shopify store connected to TikTok, routed orders through it, and shipped sample bundles to introduce users to the Scentbird experience before converting them into subscribers.

We grandfathered long-term subscribers to reward loyalty. Some stayed seven or eight years, though many churned within 12 months. Early, accurate analytics made it possible to balance growth and retention effectively.

Bandholz: What size company benefits from Finsi’s analytics?

Rebrov: It’s less about size and more about the growth stage. Each stage faces different challenges. One of the biggest is cash flow. Every physical SKU has its own lead time, so if inventory takes three months, companies must accurately forecast demand, churn, and cash flow. For early-stage brands, those with annual revenue under $10 million, we help stabilize operations and predict cash needs.

At $10 to $50 million, segmentation becomes crucial: identifying lapsed customers for personalized win-backs and recognizing high-value customers early to offer premium experiences.

At $50 to $150 million, the focus shifts to eliminating surprises and aligning systems, ensuring promotions run correctly and teams understand how one decision impacts another. Larger brands often expand into new product lines and face the same scaling issues again. Across all stages, success depends on accurate, unified data to guide smarter decisions.

Effective retention depends on understanding why customers cancel.

Bandholz: How do you do that?

Rebrov: We usually start with surveys to gather both structured and unstructured feedback. Multiple-choice questions provide quantifiable insights, but the real value lies in open-ended responses, where customers share their personal stories. Surveys let you reach thousands of people efficiently, but phone conversations are invaluable. Talking directly with customers often reveals unique motivations and use cases that spark creativity and guide product development.

Spending even 30 minutes on the phone with a few customers, especially loyal ones, can uncover more insights than analytics ever could.

Certain products naturally fit subscriptions. Examples are consumable supplements, protein powders, and snacks. But infrequently purchased goods are better suited for one-off sales.

Companies must decide early because it shapes their marketing strategy. For traditional ecommerce, profitability often depends on the first sale. You aim to cover acquisition, cost of goods, and shipping upfront, often by selling bundles.

For subscriptions, the focus shifts to lifetime value. Sellers can afford to lose money initially if they know the customer will stay long enough to become profitable. Predictive LTV helps qualify customers early and informs how much you can spend to acquire them.

Simplicity wins. Don’t confuse prospects with multiple purchase paths or offers. Ensure a “subscribe and save” offer is consistent and easy to understand.

The beauty of subscriptions lies in predictable cash flow. Yet rising acquisition costs make retention even more vital.

Bandholz: Where can people follow you, reach out to you, or hire your services?

Rebrov: Our site is Finsi.ai. I’m on LinkedIn.

Google Quietly Signals NotebookLM Ignores Robots.txt via @sejournal, @martinibuster

Google has quietly updated its list of user-triggered fetchers with new documentation for Google NotebookLM. The importance of this seemingly minor change is that it’s clear that Google NotebookLM will not obey robots.txt.

Google NotebookLM

NotebookLM is an AI research and writing tool that enables users to add a web page URL, which will process the content and then enable them to ask a range of questions and generate summaries based on the content.

Google’s tool can automatically create an interactive mind map that organizes topics from a website and extracts takeaways from it.

User-Triggered Fetchers Ignore Robots.txt

Google User-Triggered Fetchers are web agents that are triggered by users and by default ignore the robots.txt protocol.

According to Google’s User-Triggered Fetchers documentation:

“Because the fetch was requested by a user, these fetchers generally ignore robots.txt rules.”

Google-NotebookLM Ignores Robots.txt

The purpose of robots.txt is to give publishers control over bots that index web pages. But agents like the Google-NotebookLM fetcher aren’t indexing web content, they’re acting on behalf of users who are interacting with the website content through Google’s NotebookLM.

How To Block NotebookLM

Google uses the Google-NotebookLM user agent when extracting website content. So, it’s possible for publishers wishing to block users from accessing their content could create rules that automatically block that user agent. For example, a simple solution for WordPress publishers is to use Wordfence to create a custom rule to block all website visitors that are using the Google-NotebookLM user agent.

Another way to do it is with .htaccess using the following rule:


RewriteEngine On
RewriteCond %{HTTP_USER_AGENT} Google-NotebookLM [NC]
RewriteRule .* - [F,L]