eCommerce Podcast

He runs five businesses with zero employees, and he's never written a line of code in his life.
Summary
Dmytro "Dima" Negodiuk calls himself a fractional AI officer, a title his own AI agent came up with. From New York, he runs five-plus businesses across ecommerce, B2B distribution, retail, education and AI consulting, all orchestrated by a single AI agent, and without a marketing, IT or sales department behind him. In this conversation Matt and Dima get practical about where an ecommerce founder should actually begin with AI, why Claude Code is the on-ramp even for people who've never touched the tech, and how to point it at the biggest problem in your business rather than treating it like a toy.
They dig into the difference between playing with AI and deploying it, the 24/7 use cases that hand small stores an unfair advantage, and why a human still has to stay in the loop. Dima shares his AI call centre running on its 44th version, the meeting "dossier" that knows what golf game you lost twelve years ago, and the overnight autonomous work that means he checks his AI before he brushes his teeth. Matt offers his own mobile-optimisation story, where one podcast transcript turned into a scoring tool and lifted mobile conversion by over 400%.
00:00 — Three takes and a fractional AI officer
04:00 — The AI that works while you sleep
09:24 — Letting AI run your real life
16:00 — Where is all this heading
18:35 — Start Claude Code today
21:00 — When AI gets it wrong
26:48 — Start with your biggest problem
37:46 — Mobile conversion up 400%
40:25 — The dossier and cleaning the mess
43:23 — An AI call centre that never sleeps
47:58 — The best AI tip, save it for last

Why a Fractional AI Officer Runs Five Businesses Alone
(01:00)
Dima describes a model that would have sounded impossible a few years ago. He runs more than five businesses across ecommerce, B2B distribution, retail, education and AI consulting, all sitting on top of one AI agent. As an entrepreneur who has never coded, he's replaced the departments he used to run with AI.
"I never wrote a line of code in my life. Never. So I had like marketing system, marketing department. I have IT department, sales department, everything. Now I'm like sole entrepreneur." — DimaHe points out that the title isn't just his invention. The big AI labs now hire for the same role under a different name, the forward deployed engineer, at starting salaries north of $300,000. The word that matters in both is "deploy".
"Main word deploy. So you need not only to know how to, for example, create it, you need to know how to deploy it and all the system need to work." — DimaWhy Claude Code Is the Place to Start
(22:00)
For founders who feel AI is passing them by, Dima's advice is blunt and encouraging at once. Start today, start with Claude Code, and don't be embarrassed about how little you know.
"When I saw my first video and it says open terminal on your MacBook, I ask ChatGPT like web version, what is terminal? And I'm not shy about it." — DimaA few practical pointers from the conversation:
Search YouTube for a "Claude Code for dummies" video and begin with the basics.
Skip the cheapest plan if you intend to build. Dima suggests starting on the $100 plan rather than the $20 one.
Treat ChatGPT or Codex as a reviewer sitting on top of Claude Code, sense-checking the work. A $20 subscription is enough for that job.

Matt agrees it's a genuine on-ramp, and adds that the choice between tools is a bit like Apple versus Android. Dima uses Claude Code for ecommerce; friends building crypto trading bots reach for systems that fire requests many times a second instead. Stick to what fits your niche.
Start With the Problem You Actually Have
(27:00)
The most useful starting point, both agree, is the biggest problem in front of you. No product? Ask for product ideas. A product that won't sell? Ask for marketing ideas. Then check the output, because AI gets things wrong.
Matt tells the story that makes the case. After Adam Pearce came on the show to talk about mobile optimisation, Matt fed the episode transcript to Claude, had it extract the key ideas, research and score each one, and eventually build a tool that audits a website against current best practice.
"Mobile conversion on our website so far is up almost 500%. It's definitely over 400% increase, um, on mobile conversion. You're talking about hundreds, if not millions of pounds worth of difference." — MattDima frames the same idea around two questions worth asking of any business. Where are the problems AI could help with, and what are the repeated tasks you could automate? A one-off phone call is quicker to make yourself. A task you do every day is worth handing over.
"Now your knowledge and your ideas it's your best benefit because everything else will Claude do for you. But one thing he can't do, it's like, I think, to dream and to generate really brilliant ideas." — DimaThe 24/7 Use Cases That Give Small Stores an Edge
(31:00)
This is where the conversation gets concrete. Dima runs an AI call centre built on 11Labs, now on its 44th iteration, with Claude analysing each call and suggesting improvements. For his granite pavers business, where staff sometimes step away to help with a forklift, it means no missed call ever becomes a lost customer.
"If you will not answer immediately, the client will go to someone else who will answer immediately." — DimaThe point isn't novelty, it's the round-the-clock cover that small companies usually can't offer. A customer with toothache at 2am who can't reach a dental practice will phone the next one. The business that answers wins.
He also describes a meeting-prep skill he calls the "dossier", which spends hours researching a person from open sources and then advises on how to approach them. And for mid-sized businesses drowning in mismatched spreadsheets and systems, he makes the case that you don't need to rip anything out. You put AI on top, let it find the mess, and let it clean the data.
"It's how it works that each day you have some kind of ideas, new ideas, and you wanna make it live... It's like I call it like some kind of AI addiction." — DimaAI Is a Co-Pilot, Not a Pilot
(14:00)
For all the enthusiasm, both are clear that a human stays in charge. AI hallucinates, apologises, and moves on, but it doesn't carry the consequences. Matt mentions a man fined £30,000 by the tax office after letting AI file his taxes incorrectly. The AI says sorry. The person pays the bill.
"This is why I often talk about AI needs to be a copilot. It can't be in charge." — MattDima ends on a characteristically direct note. The pace of change is so fast that even the labs won't forecast more than a couple of months ahead, so the only sensible move is to get in the water and start.
"You just need to sit and you just need to understand that if you will not spend at least 10 hours per day learning AI now, AI will— someone will control you who is now doing this." — DimaToday's Guest
Today's Guest
Today's guest: Dmytro "Dima" Negodiuk
Company: Negodiuk
Website: https://negodiuk.ai
LinkedIn: Connect with Dima on LinkedIn
Episode link: https://www.ecommerce-podcast.com/hes-never-written-a-line-of-code-in-his-life

Two founders, both wildly successful, just gave completely opposite advice about whether ecom operators should build a personal brand — and Matt is running a 90-day experiment to find out which one is right for founders like us.
Episode Summary
Davie Fogarty (the $200M-a-year founder behind The Oodie) reckons founders shouldn't bother with personal brand until they're past $10 million — the opportunity cost of a single YouTube video runs to roughly $3,200 of founder time, and that money tests paid-ads creative far more effectively. Daniel Priestley and Alex Hormozi argue the opposite. AI can now replicate most of what a founder does, but it can't replicate lived experience — and the window to build that moat is closing fast.
Rather than pick a side from the cheap seats, Matt is running a head-to-head test for 90 days. Organic Instagram versus paid Meta ads, with a real commercial outcome on the line — signups to Slingshot, Aurion's new AI-for-ecom membership. The methodology is borrowed from Matt's 22-year-old son Zak (the Nutritionist), who went from zero to 100,000 followers in seven months using a four-step loop — Find, Decode, Replicate, Iterate. The niche is deliberately tight (AI workflows for ecom operators under £5M), the lab is Matt's personal Instagram, and the numbers will be published either way.
Chapter Timestamps
0:00 — Cold open — Davie vs Daniel
5:18 — Welcome
6:03 — Act 1 — The £10M Question
20:44 — Act 2 — Zak's methodology
29:39 — Act 3 — What Matt learned from 15 videos
41:43 — Act 4 — The playbook
56:23 — Act 5 — The 90-day bet
1:06:58 — Wrap + the Lab Kit

Key Takeaways
The £10M Question — Two Smart Founders, Two Opposite Answers (6:03)
Davie Fogarty's argument is built on opportunity cost. By the time a founder has planned, scripted, filmed, edited, and replied to comments, a single YouTube video costs roughly $3,200 of founder time — money better spent testing paid-ads creative until the business is well past $10 million.
Daniel Priestley's counter-argument is built on the moat. AI can write the copy, design the ads, even handle support. The one thing it cannot do is be the founder. Lived experience is becoming the only defensible asset a Digital David has — and Hormozi's 35,000 pieces of content a year shows what going all-in on that thesis looks like.
"Relatable beats impressive. Founders, when we're honest, have a real edge here — because we've been through the mess." — Matt EdmundsonZak's Four-Step Loop — Find, Decode, Replicate, Iterate (20:44)
Matt's son Zak grew from zero to 100,000 followers across Instagram and Facebook in seven months, built a recipe-book business off the back of it, and is about to launch his first app. The methodology is brutally simple — find videos that work in the niche, decode why they work, replicate with a personal point of view, and iterate based on real engagement data.
The lesson hiding inside that loop is the tea towel test. Same script, same person, same camera — but a tea towel on the shoulder lifted engagement. A few degrees of camera tilt lifted it again. Most founders make content based on what they think looks good. The audience always tells a different story, but only if someone is systematically testing.
What Transfers From Creator Advice And What Doesn't (29:39)
After running 15 top videos in the audience-growth space through MAGPIE, Matt found a clean split between gold and creator-economy fluff.
Gold:
Kallaway's "pick one platform for six months" rule — commit fully before even thinking about expanding
Chantal Leonhardt's audience-naming trick — naming the audience directly in the title (e.g. "for your small business") pre-qualifies the click and lifted her performance 15× over baseline

Fluff to ignore:
Vanity-metric opens ("This month I did 32M views") — these actively repel founder audiences
Daily-grind cadence advice — a recipe for burnout when social is the seventh priority on a founder's list, not the first

The 90-Day Bet — Real Numbers, Either Way (56:23)
Instagram-attributed Slingshot signups versus paid-ads benchmark, measured over 90 days, published openly. Phase 2 is taking the proven playbook into Aurion's other brands — Vegetology and Seven Yays — once the founder version has been pressure-tested.
Resources
The Founder's 90-Day Instagram Lab Kit (free) — the Decision Sheet, the Tea Towel Test Worksheet, and the Founder Content Scorecard, all in one pack. Grab it at ecommercepodcast.net/resources
Slingshot — Aurion's AI-for-ecom membership. The full 90-day playbook, plus SAM, MAGPIE, Prism, and Scout — all the AI tooling Matt is running this experiment on. Details at ecommercepodcast.net/slingshot
eCommerce Cohort — free, application-only community where ecom founders solve real problems together. Monthly 90-minute virtual sessions, regional groups (UK, AUS-NZ, US). Apply at ecommercecohort.com

About The eCommerce Podcast
The eCommerce Podcast is a show dedicated to helping you deliver eCommerce WOW with real talk about building online stores. Every Thursday. Join Matt Edmundson and guests, experts and founders who've been in the trenches, built the stores, and learned the hard way — so you don't have to.
Host: Matt Edmundson
Connect with Matt on Instagram: https://www.instagram.com/mattedmundson/
Connect with Matt on LinkedIn: https://www.linkedin.com/in/mattedmundson/
Episode link: https://www.ecommerce-podcast.com/90-days-cracking-instagram-for-ecommerce-with-matt-edmundson

Organic Search for Ecommerce in 2026
What happens to an ecommerce business when paid ads stop working? Matt Edmundson walks through the six organic traffic fronts every founder should be building right now, before paid stops carrying the business on its own.
Summary
Paid ads are getting more expensive, less effective, and harder to justify. Facebook CPMs are up 60–90% over the last few years, paid search clicks have effectively doubled across every vertical, and organic clicks have fallen between 11 and 23 percentage points depending on the industry. Matt opens with the story of switching the ads off in his beauty business out of necessity, and watching sales hold because of an organic foundation that had been quietly compounding underneath. He then walks through the six organic fronts he thinks every Digital David needs to be working on in 2026, with one concrete action per front that can be done this week. Twice in the episode he stops to make it very clear he is not telling anyone to turn paid off — the point is to build organic now while paid is still doing some of the work.
00:00 — Turning the ads off and the foundation that quietly held
03:04 — A confession about sponsorship and the SAM reveal
05:06 — Why paid media is breaking for ecommerce right now
15:48 — Why this isn’t an instruction to switch ads off
23:47 — The six fronts of organic traffic in 2026
27:06 — Front 1, AI discovery
29:48 — How Mrs Smith found the beauty store on day one
38:33 — Front 2, site SEO
42:43 — Front 3, blog and content
46:12 — Front 4, YouTube
51:50 — Front 5, social SEO
55:35 — Front 6, digital PR
1:05:00 — What SAM is and how it shortens the runway
1:11:10 — Why a free customer changes the LTV to CAC maths
1:13:24 — A teaser for the next episode and a platform rebuild

Why Paid Media Is Breaking and What to Do About It
[05:06]
Matt walks through the shape of the problem most founders are quietly carrying. Customer acquisition costs climbing, return on ad spend falling, and a treadmill that moves you from Google to Facebook back to Google to TikTok to whatever the marketing industry invents next. Around 8 out of 10 of the founders he speaks to right now have meaningful issues with paid media, and the businesses most exposed are the ones depending on first-time customer acquisition with low repeat rates.
“The marketing industry will keep inventing new places for you and me to spend our advertising budgets.” — Matt EdmundsonHe stops twice to make this clear, once at 15:48 and again at 16:42 — he is not telling anyone to turn paid off. If first-time customer acquisition is the lifeblood of the business and ads stop today, the founder will feel it by the end of the day if not the end of the hour. The point is that paid is still working at some level, and that makes now the right window to build the organic foundation underneath it.
The Six Fronts of Organic Traffic in 2026
[23:47]
Matt argues that thinking of organic as just SEO is a 2018 mindset. In 2026 it sits across six channels, and most founders are only dabbling in one or two of them. Before he runs through them, he flags a warning — the same agency-promises-the-world problem that plagued SEO in 2018 is now happening with AI search, and most of the agencies charging premium fees for AI discovery don’t really know what they’re doing yet either.
AI discovery — getting recommended by ChatGPT, Claude, Perplexity, and Google’s AI Overviews. 78% of consumers have used large language models for product research in the last year and 64% plan to use AI chatbots for shopping decisions this year. Three concrete actions, add an llms.txt file to the root of the site, complete full product schema (including GTINs, return policy, aggregate rating), and keep content fresh — ChatGPT pulls about three-quarters of its most cited pages from content updated in the last 30 days.
Site SEO — the technical foundation that made it possible to turn the beauty business’s ads off without dying. Pick the top 10 product pages and audit each one for four things: a 50–60 character title tag using keyword plus differentiator plus brand, a meta description written for humans, og:type set to product rather than the default website, and image alt text that describes the image rather than just naming the file.
Blog and content — the trust layer AI now uses to decide what a brand actually knows. Brands that blog consistently see roughly 13 times more positive return on investment than sporadic publishers. The action is simple — ask the customer service team for the top three questions they get every week, and write one proper 1,200-word answer per month. That’s 12 articles a quarter, 50 a year, each one a piece of content AI can cite.
YouTube — the second largest search engine on the planet, with 35 billion hours of shopping-related content watched last year (a 250% year-on-year increase). 61% of 14–25 year olds say YouTube introduced them to brands they hadn’t heard of. One product demonstration video a month, with a proper title, three-paragraph description, chapters in the description, and product cards on the video.
Social SEO — Gen Z search inside Instagram (67%) and TikTok (62%) before they touch Google (61%). Captions, hashtags, bios, and on-screen Reels text are all searchable assets now. Treat captions like meta descriptions and bios like homepages.
Digital PR — press mentions, podcast appearances, founder authority, Reddit presence. Brand search volume correlates with AI recommendations roughly three times more strongly than backlinks do.

“The question isn’t how do I rank anymore. It’s how many people are searching for your brand name.” — Matt EdmundsonHow Mrs Smith Found the Beauty Store on Day One
[29:48]
Inside the AI discovery section Matt tells a story from 2006, when he launched his beauty business. They set the site live, planned to come back the next day to keep testing, and overnight Google indexed them and put them in front of a real customer called Mrs Smith — who placed the first order before the team had even told anyone the shop was open.
“I appreciate that doesn’t really happen anymore.” — Matt EdmundsonThe point of telling it now is that AI feels like that early-Google moment all over again. There’s no rulebook everyone is following, the ones who get in early are the ones who get discovered, and a handful of his own businesses are already getting leads who say “ChatGPT told me about you.”
SAM, the Slingshot AI Mentor
[03:04] [1:05:00]
For the first time on EP, Matt openly opens the episode with a sponsor — his own company. SAM stands for Slingshot AI Mentor, the e-commerce layer Aurion has been running on Claude internally and is now opening up to the public. It’s Claude plus a focused e-commerce knowledge layer plus the context of the founder’s own business. Inside SAM lives Scout, the skill that runs the same six-front audit covered in this episode automatically across a site and hands back a playbook of specific changes. SAM membership is currently $2,000 (about £1,600) on a waitlist with personal onboarding from Matt. Anyone who joins the waitlist by Thursday 22 May 2026 gets a 25% discount code plus extra credit for Scout. Available at ecommercepodcast.net/slingshot.
Why a Free Customer Changes the LTV to CAC Maths
[1:11:10]
The point of building organic isn’t to abandon paid. It’s to change the underlying maths of the business. A customer who finds the brand organically has, in effect, a CAC of zero. Organic isn’t actually free — it costs time and staff — but it costs differently, and it compounds.
“Organic isn’t actually free. It just costs differently.” — Matt EdmundsonPaid media CAC across ecommerce is currently running about 2.5 times higher than blended CAC. Drop ten organic customers into the LTV to CAC ratio, then a hundred over a year, and the shape of the business changes — not because more customers are being acquired, but because the average cost of acquiring them falls.
Free Resource Mentioned in This Episode
Matt has put everything in this episode into a free printable PDF — the Organic Audit Kit. It includes the full checklist for each of the six fronts, 18 AI prompts (one to audit, one to fix, one to verify per area) that paste straight into Claude or ChatGPT, a scoring radar to see strengths and weaknesses at a glance, a 20-minute quick win for each front, and a glossary for the technical terms. Grab it at ecommercepodcast.net under Resources.
About Today's Episode
Host: Matt Edmundson
Company: Aurion
Website: aurioncompany.com
LinkedIn: Connect with Matt on LinkedIn
Episode link:

Up to 40% of Amazon US sales now touch Rufus, Amazon's AI shopping assistant — and most sellers haven't optimised a single product for it. That's the gap David Balan, founder of Selluna, walked through with Matt this week. The customer who used to type "blue football" into a search bar is now asking a chatbot what kind of football would suit a six-year-old, and the products that show up first are the ones answering the right 25 questions.
David has spent the last few years helping thousands of brands across the US, UK, Europe and Asia-Pacific sell more on Amazon. In this conversation he laid out a three-pillar framework — traffic, clicks, conversion — and made the case that anyone selling on Amazon without a Rufus strategy right now is leaving 12-24 months of compounding advantage on the table.
Key Point Timestamps
[04:29] Why the Amazon algorithm is simpler than most sellers think
[05:28] The three pillars of Amazon — traffic, clicks, conversion
[07:00] Rufus, the top 25 questions, and how to answer them in priority order
[12:03] Traditional Amazon SEO — 80+ keywords, frequency, and the "stuffing" myth
[19:30] Why Rufus optimisation now is the new 2016 SEO play
[24:57] Why not A/B testing the main image is "criminal" in 2026
[33:08] The conversion stack — seven images, A+ Content, brand story, premium branding
[39:32] Text speaks to the algorithm, images speak to the customer

The Algorithm Is Simpler Than Sellers Want To Admit [04:29]
Most sellers treat Amazon's algorithm like an opaque, mystical force. David disagrees. "The algorithm is more simple than we like to make it," he told Matt. "The algorithm is trying to make Amazon as much money as they can get out of it. So once we understand their goal, if we align our goals with theirs, then everyone's going to be happy and you're going to be selling more at the end of the day."
That's the philosophical shift. Stop trying to outsmart Amazon. Start asking what Amazon wants and give it more of that. Every metric David teaches — traffic, clicks, conversion — is a proxy for the same goal, putting the right product in front of the right customer at the right time.
The Top 25 Questions That Make Rufus Recommend Your Product [07:00]
Rufus is Amazon's ChatGPT-style assistant, embedded in the marketplace. Instead of typing "football" into a search bar, customers now ask questions — what colour, what size, what age group, premium or budget. Rufus interrogates their needs and recommends specific products. Some data suggests up to 40% of Amazon US sales now involve Rufus at some point in the journey. David thinks the real number is closer to 20%, but it's growing fast and rolling out across Europe right now.
The play is simple. Rufus, like any AI, leans toward whatever it has the most data on. To make a product the answer, sellers need to identify the top 25 questions customers ask about their category and then answer those questions across every part of the listing in priority order.
The most important questions go in the title
Next most important in the bullet points
Then the product images
Then the back-end attributes
Then the reviews and Q&A

Selluna's free tool at app.selluna.ai/audit will surface the top 25 questions for any listing. Sellers can also run a title and bullets through ChatGPT, or write the questions out manually based on what real customers ask in reviews and support emails.
Traditional Amazon SEO Still Drives Most Sales [12:03]
The SEO pillar still drives most sales on Amazon. David's team targets over 80 keywords on a single listing, with each one appearing on average three times across the title, bullet points, back-end and Premium A+ Content. Top keywords appear 35 to 40 times.
And before anyone panics about keyword stuffing — David's continuous testing shows the conversion penalty is negligible (one or two percent) compared to the traffic gain (ten to twenty percent or more). "Keyword stuffing" is a fear from the Google 2012 era. On Amazon in 2026, frequency is still a seller's friend.
Why Now, Not Next Year [19:30]
David is urgent about Rufus optimisation, and there's a specific reason. Amazon SEO used to be like this. In 2016 to 2018, sellers were building empires on a single product image and a short title because almost no one was doing proper SEO. The brands that put in the work then are still ranking today, a decade later, on the authority they built when the field was empty.
Rufus is in that 2016 moment. Adoption is on a hockey-stick curve. The questions a seller teaches Rufus to associate with their product now will compound into 12-24 months of defended position. The brands that wait will be fighting for second place, the same way most SEO sellers are today.
Why Not A/B Testing Your Main Image Is "Criminal" [24:57]
Once a product is showing up — in Rufus answers and on search pages — people still need to click. And 80% of click-through rate is driven by one thing. The main image.
David did not soften this. "Not doing at least two or three A/B tests for a product that's doing over $1,000 per month in sales is criminal in 2026."
The maths is brutal in a good way. A 10% lift in clicks is not a 10% lift in sales — it's much bigger over 12-24 months, because more clicks tells the Amazon algorithm a product is performing better than the competition, so Amazon shows it more. The compounding can be the difference between $40,000 and $50,000 in monthly revenue from the same listing.
A few practical rules from David —
Make the product as large as possible in the frame
Add contrast, especially heavy shadows behind white products that disappear into a white background
Test the image at the size of two thumbs on a phone, because that's the actual viewing surface for most shoppers
Add a label on or attached to the product showing the main keyword (a "blue football" label on a blue football). Even a 5% subconscious lift in click-through is significant
Don't break the label free from the product — Amazon's AI flags floating elements

Conversion — Where Premium Branding Earns The Price Tag [33:08]
The conversion pillar is where most sellers leave the most on the table, because it's the largest surface area to work with. Seven product images. Premium A+ Content. The brand story, a section above or below the product description that almost everyone skips.
David's principle for product images is the same as for Rufus listings — answer the top 5 to 10 questions in priority order, but visually. If a seller is selling a £60 ($75) premium product against £12 ($15) competitors, the first image after the main one needs to be a "us versus them" answer. As David put it, "a confused buyer is not a buyer."
One of David's clients, Doodlebrush, sells at £60 ($75) against competitors at £12 ($15) and holds 20% market share. The reason is branding. Premium pricing demands premium branding — the same logic that lets Nike and Apple charge what they do. If the branding doesn't match the price tag, the price tag loses every time.
Text For The Algorithm, Images For The Human [39:32]
David's closing line is the one Matt said he'd be quoting. "The text speaks to the algorithm. The images speak to the customer."
The titles, bullet points and back-end fields aren't there for the customer to read (almost no one does). They're there for Rufus and the search algorithm to chew through and decide where to rank a product. So write that text for the machine — keyword-rich, question-answering, dense.
The product images are where the human meets the product. They need to be visual, scannable, and built around the questions a real buyer is asking — not paragraphs of text the algorithm could already read elsewhere.
Where To Start This Week
Audit the top 25 questions — through Selluna's free tool, ChatGPT, or manual research from reviews
Rebuild the title and bullets to answer those questions in priority order
A/B test the main image on any product doing $1,000+ per month with Brand Registry. Run for at least four weeks
Audit the conversion stack — seven images, Premium A+ Content, brand story. Each one needs a job
Match the branding to the price tag — premium pricing demands premium branding before customers will accept it

Today's Guest
Today's guest: David Balan
Company: Selluna
Website: https://www.selluna.ai/
LinkedIn: Connect with Selluna on LinkedIn
Episode link: Up to 40% of Amazon US sales now touch Rufus, Amazon's AI shopping assistant — and most sellers haven't optimised a single product for it. That's the gap David Balan, founder of Selluna, walked through with Matt this week. The customer who used to type "blue football" into a search bar is now asking a chatbot what kind of football would suit a six-year-old, and the products that show up first are the ones answering the right 25...

Rachel Hanretty pays £500 a month for a tool that fact-checks her marketing agency — and she has opinions about why every seasonal ecommerce founder should do the same.
In this episode, Matt Edmundson sits down with Rachel Hanretty, founder of Mademoiselle Macaron, the Scottish brand shipping up to 25,000 macarons a week UK-wide. Rachel built the business from a St Andrews student flat in 2013 after training in Paris, and thirteen years on she has strong views on attribution, agency accountability, the seasonality trap, and the gulf between what Instagram wants and what authenticity actually looks like. The conversation covers Triple Whale, Klaviyo's tendency to overstate its own impact, a Pinterest wedding ads experiment that burned thousands, a Charlotte Tilbury collaboration that produced lipstick-matched macarons at H Beauty, and why Rachel feels "gaslit by her own business" every April to September.
It is half ecommerce masterclass, half therapy session — and Rachel is refreshingly honest about both.
In This Episode
(01:30) From a Paris kitchen to 25,000 macarons a week
(07:15) The seasonality trap: "gaslit by my own business"
(14:00) Attribution wars — why Klaviyo and agencies overclaim
(17:00) The £500/month tool that fact-checks her agency
(22:00) Everyone needs an Alan: fractional CFOs and complementary hires
(27:00) ROMS — return on macarons spent, and experiential retail
(34:00) How do you do authentic Instagram when your niche is Paris?

Attribution Is Broken, And Every Platform Is Overclaiming
(Around 14:00)
Rachel's biggest challenge is not product or operations — it is knowing which marketing pound is actually doing the work. Klaviyo, she says, is "way too generous" with the revenue it claims. Agencies do the same. Until she started using Triple Whale, she had no independent way to push back.
"When you're being charged thousands of pounds for a service, you're like, no, I'm really sorry, but that number doesn't match up with the number in my bank." — Rachel HanrettyMatt picks up the thread with a story from Neil Hoyne, Google's Chief Measurement Strategist and author of Converted, who once traced a single pair of shoes back through 236 separate customer touchpoints. If every platform claims that sale, the maths does not work. Matt's recommendation (crediting previous EP guest Matt Putra): stop fighting about platform-level ROAS and track blended ROAS — the one number no agency can spin.
The £500 Tool That Coaches Her Through The Data
(Around 17:00)
Rachel pays £500 a month for Triple Whale, but the attribution modelling is not actually what she values most. The real prize is the platform's AI chat — she calls it "MobiChat" — which she uses as a coaching tool.
"I just say, can you explain what's going on with the Meta ROAS? Should I cut the budget? And it throws up the halo effect model of Meta on Google. I'm like, thank you so much for explaining it — because you're not paying by the hour." — Rachel HanrettyOver six to nine months she has used the tool as much to build her own knowledge as to interrogate the data. When an agency sends a report, she pastes it into Triple Whale and asks: is this right? That single habit has changed the power dynamic in every agency conversation she has.
Shopify-only tool (a limitation worth flagging for non-Shopify founders)
Value comes from the coaching layer, not just the attribution model
Used to sanity-check agency claims and reduce reliance on fortnightly calls

"Gaslit By My Own Business" — The Seasonality Trap
(Around 07:15)
October to March is the heroic half of Rachel's year: Halloween, Christmas, Valentine's Day, Mother's Day, Easter. April to September is a different business entirely — different audience, different messaging, different paid media strategy. The same product, but the customer and the context shift underneath her.
"Sometimes I feel like I'm being gaslit by my own business." — Rachel HanrettyHer Pinterest wedding-ads experiment is a case in point. The team put thousands into Pinterest last summer aiming at brides, only to realise that people already getting married that summer had booked their suppliers months earlier. Wedding customers need to be reached a year ahead — not in the season itself.
Rachel's lesson: cost per acquisition, not ROAS, is the number that matters when you are used to seeing 5-8x returns and suddenly cannot.
Every Company Needs An Alan
(Around 22:00)
Rachel started the business at 23 with very little reference point for managing people. Two leadership coaches and one fractional CFO later, she has a working philosophy: hire for the gaps in your own skill set, and do not be afraid when those people outshine you in their area.
Her fractional CFO is called Alan. He builds the dashboards that link a website order to grams of sugar, the cost of a 20% off promo, and the capacity implications of a given campaign. He is, as Rachel puts it, "the adult in the room."
"Every company needs an Alan." — Rachel HanrettyMatt echoes the point with Andrew Carnegie's epitaph: "Here lies a man who knew how to surround himself with people cleverer than himself." The discipline is not just hiring well — it is managing your ego when someone else shines.
ROMS, Charlotte Tilbury, And The Rise Of Experiential Retail
(Around 27:00)
Rachel's framework for the business breaks down into three "moments of luxury": moments of celebration (weddings, birthdays), moments to impress (corporate gifting and brand activations), and moments of thoughtfulness (gifting). The middle category is where the most interesting growth is sitting.
Last year Mademoiselle Macaron worked with Charlotte Tilbury across H Beauty stores, producing thousands of macaron boxes colour-matched to individual lipsticks. The end customer may or may not eat the macarons — they are often photographed and posted on social media alongside the product. Rachel coined a term for measuring this kind of spend: ROMS — Return on Macarons Spent.
"Sometimes I don't even know if the macarons get eaten, frankly. I'm okay with it as long as they were paid for." — Rachel HanrettyThe Freezer Confession (And A Note From Pierre Hermé)
Rachel shared an operational secret that many founders in the fresh-food space will recognise: Mademoiselle Macaron freezes its macarons. Far from being a downgrade, the French pastry chef Pierre Hermé has publicly said that "anyone who says they don't freeze their macarons is a liar — because it improves the texture." Freezing makes the texture more marshmallow-like, and it gives Rachel the operational flexibility to take pre-orders from October and ship fresh-feeling product right up to the last Christmas posting day.
Authenticity vs The Algorithm
(Around 34:00)
Rachel's "question for Matt" was the most honest moment of the conversation: her niche is Paris, which is aesthetically beautiful, so every other bakery brand on Instagram is polished to within an inch of its life. She went to Paris in January, paid for an Airbnb-experienced photographer to shadow her for 36 hours, and hated every minute of it. Her Scottish phrase for how she feels about polished-female-bakery-Instagram: "gie's me the boak" — it makes her feel ill.
Matt's advice: stop trying to be polished. The videos doing one to two million views are shot on iPhones. The accounts winning are the raw, authentic ones. He points to his own son, Zak the Nutritionist, who built 70,000 Instagram followers in six months doing 60-second IBS-help clips filmed in Matt's back garden — and has now turned that audience into a successful recipe book.
"Consistency wins every single day and authenticity wins every single day. It does not have to be polished. People buy you." — Matt EdmundsonKey Takeaways
Fact-check your agencies. Platform-level ROAS from Klaviyo, Meta, and Google are all overstated. Use blended ROAS and an independent tool to sanity-check claims.
Cost per acquisition > ROAS when you move between seasons or audiences.
Long-tail channels need long-tail thinking. Pinterest for weddings works — but the year before, not during the season.
Hire your Alans. A fractional CFO (or anyone filling your skills gap) is worth more than another marketing hire when you cannot read your own numbers.
Experiential retail is a real revenue channel. Brand activation work pays per unit and builds social proof.
Stop trying to be polished. The algorithm rewards authenticity, not aesthetic perfection — even in aesthetically-driven niches.

Resources Mentioned
Mademoiselle Macaron —