Selling on Amazon means paying almost double Shopify’s fees. Brands paying them make 3x the dollars. @youderian and I were chatting about how many founders are second-guessing Amazon right now. So we pulled the H1 2026 books of 7 and 8 figure consumer brands to see if the fear holds up. The fees first: Amazon: ~30% of every sale (15% referral, 15% FBA) Shopify: ~16% (3% processing, 13% shipping and fulfillment) Now the contribution margins: Brands on both Shopify and Amazon: 27.2% Shopify-only brands: 28.9% Almost the same rate, on a business about 3.3x bigger. Even with fees that high. Because the Amazon side barely pays for anything else. Ads: ~13% of Amazon sales vs 22-33% on Shopify Discounts: ~2% on Amazon vs 12% on Shopify The extra 14 points in fees mostly come back as lower ad spend and fewer promo codes. And it’s not because Amazon Ads are better. They do a different job. Meta and Google → create demand Amazon Ads → catch people already searching with a card in hand Which means your Meta spend is quietly buying Amazon orders. That’s good news, as long as you judge the whole company and not your Shopify ROAS. Amazon can’t capture demand you never created. And the cannibalization fear? For brands on both, Amazon sales grew 28%. Shopify still grew 10%, about the same as Shopify-only brands. Before you cut Amazon to save on fees, look at the company-level margin, not the channel P&L.

Every financial and analytics MCP claims you can “ask anything.” The data founders have now is unreal. But a toolbox is useless if you can’t swing a hammer. So, a new series: Analyze, Act, Lock in. Three ways founders use the Finaloop MCP, one feature at a time. This week, P&L segments (link in reply): Analyze: rank your vendors, benchmarked against similar brands. Act: draft the negotiation emails from your actual invoices. Lock in: a weekly opex budget with live variance and alerts. Same three moves work on any expense line of the P&L.

“I almost never leave reviews.” This is from a customer that has been with us for over two years. Before Finaloop they’d tried just about everything, all the way up to NetSuite. Their review puts into words what we’ve spent the last year working on: "handles every pain point seamlessly", “true metrics,” “worth multiples of what they charge.” I’ve written before about our unconventional decision to freeze marketing in @FinaloopTeam for a year and work on the product. If you’d asked me then what success would look like, this review is my answer. This made my week.

How many accountants does it take to build a P&L? Not a joke. It’s how our new P&L started. We put a bunch of our accountants, the ones who close the books for 2,500+ consumer brands, many of them 8 and 9 figures, in a room for eight hours a day, for days. One question: how does a brand at that scale actually think about its business? Not the way a P&L shows it. A standard P&L is one column ending in one profit number. It tells you if you made money. It can’t tell you where. At $2M, fine. At $20M or $100M across five channels, “where” is the whole job: Amazon ads or Meta? Is that wholesale order worth taking? Is TikTok Shop growing profitably, or just growing? So we rebuilt it. Split your P&L by sales channel, platform, revenue stream, fulfillment channel and more. Stack two breakdowns. Compare periods. Or connect our MCP and just ask Claude. Not a dashboard next to your books. It is your books, the same reconciled P&L our accountants close every month. P.S. All accountants survived and came out needing a drink, but otherwise in good spirits.

Founder math: Refund a $70 order. Put the $21 product back on the shelf. Still lose $23. $8 in outbound shipping. $9 for a prepaid return label. $4 to open, inspect, and restock. $2 in card fees that aren’t refunded. That’s $23 gone. If the product can't be resold, that loss becomes $44. An expensive round trip for $0 in revenue.

Worst time to hire a CFO: before your books are clean. Shopify payouts unreconciled. Amazon settlements in the wrong month. Ad spend booked when cash leaves. COGS missing freight and duties. You can’t build a reliable forecast on that. So the pricing, cash planning, and fundraising work waits while you pay a CFO salary for bookkeeping cleanup. Clean books first. CFO second.

A $16 BFCM discount can turn a profitable $80 order into a $6 loss. $80 paid by the customer − $24 product, freight, duties − $16 shipping & fees − $30 CAC = $10 profit Apply a $16-off code, you lose $6. At $40 CAC, you lose $16. All before returns. Break-even CAC on the discounted order: $24. Know it before the ads go live. Otherwise, a bigger BFCM just means a bigger loss.

CMO: We hit a 4.1 ROAS on BFCM. CFO: Before or after returns? CMO: Before. Most of them land in December and January. CFO: And that’s blended? Total revenue divided by total ad spend? CMO: Yes. CFO: So it includes customers who were going to buy from us anyway. CMO: It includes everyone who bought. CFO: Did the new customers make money on their first order? After product cost, shipping, fulfillment, and fees? CMO: None of that is in ROAS. CFO: So what did we keep per ad dollar? CMO: I’ll know in February. A 4.1 ROAS and a money-losing BFCM can be the same weekend. And more often than you might think, brands don’t realize it until January closes.

Your BFCM forecast may be counting some revenue twice. A repeat customer who orders every 90 days at $65 was due mid-December. She sees your Black Friday offer and buys Nov 28 at 30% off. Dashboard: a returning-customer conversion. Books: $65 of December revenue moved to November, minus $19.50. Multiply that across thousands of repeat buyers and part of your "record BFCM" is next month's revenue, bought at a discount. Check when your repeat buyers normally reorder. If it's late Nov or Dec, the promo isn't creating that order. It's discounting it.

AI was supposed to kill SaaS. Instead, brands increased spend by 42% and added 2.5 more tools. We compared H1 2026 with H1 2025 software trends across our cohort of seven- and eight-figure brands: 1. Brands canceled an average of 3.8 recurring tools and added 6.3. Net: 2.5 more tools. 2. Brands using AI building tools like Cursor, Lovable, Bolt or Windsurf added 11% more subscriptions and increased software spend 13%. Brands not using these tools reduced their overall subscription spend by 2% on average. 3. Claude grew from one in eight brands to two in three, reaching 2.4x ChatGPT’s spend. Most brands pay for both. 4. Creative tooling moved in-house. Figma, CapCut and ElevenLabs gained brands while template libraries and social schedulers lost them. 5. There’s no sign of smaller teams yet. Payroll accounts and Slack both grew. The software spend didn't leave, it moved one step upstream. The real test comes at renewal, when something they built finally replaces something they bought.

The most important TikTok Shop halo effect doesn’t show up in your margins. It shows up in how far you can scale before CAC breaks. At last week’s TikTok Shop Operators Summit, @jordantwestecom shared our data showing that brands doing more than $1M in TikTok Shop GMV averaged over 3x the revenue, contribution profit, and EBITDA of the other 7–8 figure brands in our portfolio. That doesn’t prove TikTok Shop caused the growth, but it’s pretty convincing. @_RafiGanz went one layer deeper into the data. Turns out the TikTok Shop brands didn’t have better margins. Their margins were roughly the same, and actually a little lower. So where did the halo go? They spent it. When the TikTok Shop brands increased their budgets by 10%, CAC increased by only 1.9%. For everyone else, it increased by 4.2%. Massive difference 🤯. Great operators don’t leave an arbitrage sitting in their margin. They scale into it until the efficiency advantage gets absorbed by volume. By the time you look at the P&L, the margins may look similar. But one business has become much bigger. The differentiator is the TikTok Shop flywheel. Its affiliate network becomes a creative engine, constantly producing new content so acquisition can keep scaling without creative becoming the constraint. Bottom line: TikTok Shop can be a great funnel for brands looking to scale. Not because it gives you better margins, but because it gives you the infrastructure to keep going. 🎧 Full recording in the replies. And thank you to Social Commerce Club for putting on an awesome event.

@jackybh 💪 That's actually what AI has opened up for our customers. We always had the best reconciled business and marketing dataset for consumer businesses. Not by choice. We do accounting and taxes and produce audited financials, so we have no option but to account for every penny. That one order in Shopify that also went out as a Stripe invoice and got paid offline still has to land once, with revenue and cash each counted where they belong. With full ledger and audit trail. What you could do with all of it was capped by whatever we built into the UI, and keeping up with customers feature requests was never easy. Connect it to AI tools and the cap is gone. Customers build practically anything now, with no garbage in, garbage out. Reporting, and actual apps. We've seen custom daily reporting into Slack. Teams talking to their data through Claude or Slack. CFO-grade dashboards. Stockout and overstocking alerts. 13-week cash flow models. One customer built a light order management system. Every result is deterministic, not statistical, because the data is the source of truth. The AI just harnesses it and shapes it per our customers' needs. We're turning these into skills and distributing them across customers. Every finance team can build anything now, without touching data cleaning, reconciliation across many overlapping sources, data engineering, integrity tests, scale, or maintenance. We do all of it for you.

I think within a year, 90% of Finaloop's users will be agents, not humans. Here's why. Four weeks of data on how brands use our Finaloop MCP. Two things stood out: 1. Most usage isn't chat. Two-thirds of requests come from Claude Code, Codex and Cursor. Only 30% from a chat window. Brands are building their own tools on top of their books: → An LTV cohort prediction which updates for actuals every week → A Slack notification for repeating influencer purchase → A Slack alert when inventory level per SKU drops below a threshhold A fifth of users run workflows like these. They generate half of all usage. We didn't build this. Customers did. 2. It doesn't stop on weekends. 26% of requests happen Saturday and Sunday. Agents running overnight, and founders checking numbers before Monday. Nobody opens an accounting system on a Sunday for fun. Accounting and finance became a builder's sport. The 2026 DTC founder: ships code on Saturday, checks numbers on Sunday, and doesn't take the agency's ncROAS claim at face value without checking it against their own data. Haven't felt energy like this in ecom since 2020.