Use Atorie's D2M Playbook to Cut Middlemen and Capture Margin: A Founder's Checklist

Why D2M is a margin play, not a discount race

Atorie's public numbers make the D2M question concrete: it aims to sell products made with materials such as Italian leather and cashmere at roughly 10% to 20% of typical retail prices, and it generated approximately $5 million in sales in 2025 while expecting an annualized revenue run rate to exceed $55 million in 2026. The Los Angeles-based fashion-tech company, founded in 2024 by Redouane Ramdani and Luis Angulo, operates a marketplace that connects manufacturers directly with consumers and works with more than 40 factories. The capital story is not the playbook: Atorie raised a $9.5 million seed round announced on August 27, with participation from a16z Speedrun, Night Capital, and Jeremy Liew of Lightspeed Ventures. The revenue jump is a reason to test whether the operating model can absorb scale, not a proof that it can.

A defensible D2M pilot rule is: make only two to four weeks of forecast, require a 30-day sell-through bar, and scale only if gross margin beats your best wholesale channel by at least 10 points. Direct-to-manufacturer sounds simple: remove the middlemen, lower the price, keep the difference. The markdown risk is why the pilot needs a cap and a sell-through test; Atorie's public AI use includes overproduction control, which is relevant to that discipline. A concrete test case is a $120 leather tote, a 500-unit first batch, a 70% 30-day sell-through bar, and a 55% gross-margin bar.

For a founder with $1M to $20M in revenue, the lesson is not to copy Atorie's category. It is to borrow the operating discipline: one SKU, two factories, AI-assisted demand and material signals, and a hard cap on first production.

The pilot checklist

1. Choose one high-markup, low-loyalty SKU. Pick an item where customers buy on price, fit, or material, not brand attachment. A $120 retail product with a $45 wholesale cost is a better test than a $400 designer item with a cult following. The goal is to prove that a lower consumer price can produce a better gross margin than your current wholesale or D2C channel.
2. Source two factories with 2- to 4-week lead times. One factory creates dependency; two create a negotiation and a backup. You do not need a global network for a pilot. You need enough capacity to make a small batch, enough speed to test demand, and enough cost transparency to model margin before the first purchase order.
3. Set AI price bands and a stop trigger. AI pricing should not be a single number. It should be a range: a floor that protects margin, a ceiling that respects demand, and a trigger that changes production. Atorie uses AI for trend analysis, color testing, demand forecasting, material-shortage anticipation, and overproduction control. For a pilot, use those same signals to decide when to raise price, when to hold, and when to stop production. The stop trigger is explicit: if 30-day sell-through is below 70% or gross margin is below 55%, stop the next batch, reprice, or change the SKU.
4. Cap first production at two to four weeks of forecast. If your forecast is 500 units per week, make 1,000 to 2,000 units, not 10,000. The cap is the point. A smaller batch that clears the stop trigger is more valuable than a large batch that does not, because the first batch teaches you what the second batch should be.
5. Track sell-through, gross margin, and overproduction. Do not scale on revenue alone. The scale test is explicit: D2M gross margin minus your best wholesale gross margin must be at least 10 points, and inventory turns must improve. If the pilot produces a lower price but a worse margin, the model is a discount factory, not a D2M business.

What to watch before scaling

• Sell-through is the demand test. The 30-day window is where the price band, product, and demand signal get checked. If the stop trigger fires, the fix is not more inventory; it is a smaller next batch and a sharper test.
• Gross margin is the margin test. D2M only earns its complexity if it improves margin after COGS, freight, returns, and payment costs. If the lower consumer price does not leave more money, the model is not working.
• Inventory turns are the risk test. If turns fall as you scale, you are converting supply-chain speed into working-capital drag. A D2M pilot should make inventory easier to manage, not harder. If you need a larger warehouse, longer markdowns, or more discounting to move product, stop and reprice.
• Channel mix is the distribution test. Atorie says it is seeing increasing purchases from generative AI platforms such as ChatGPT and Claude and is positioning around agentic commerce. That does not mean every founder should build for AI agents today. It does mean that, at least for Atorie, AI-assisted discovery is part of the purchase path it is planning around. If your product is searchable, comparable, and price-sensitive, test whether AI-assisted discovery changes conversion before you invest in a new channel.

The practical takeaway is narrow. Do not try to become a direct-to-manufacturer company across your whole catalog. Pick one SKU, run a capped pilot, and apply the checklist: cap production, test the stop trigger, and clear the scale test. If the pilot clears all three, scale. If it only improves price, you have built a discount factory with extra steps.