Skip to content

Sample memo

Plausible-but-fictional company. The structure is exactly what Atlas generates for any thesis you bring it. Generate your own memo →

Investment memo

ReturnLoop

AI-triaged returns decisioning for mid-market apparel DTC brands.

GlobalPre-seedB2B SaaSRetail ops

Thesis

Mid-market apparel DTC brands (~50K–500K orders/year, $15M–$120M GMV) lose 4–8 points of contribution margin on returns. The status quo splits the problem badly: incumbent tools treat returns as a portal UXproblem (the customer's side), while the real margin loss lives in ops decisioning(the warehouse's side) — wrong reshelve calls, mis-graded items, slow refund cycles, fraud miss. ReturnLoop sits between the portal and the 3PL: an agent that ingests photos + customer reason + order history and outputs a triage decision (restock / outlet / donate / dispute / hold) before the box reaches the warehouse.

Comparable companies (as of training data — verify currency before any investor conversation)

  • Loop Returns — dominant self-service exchange portal for Shopify mid-market. Customer-facing, not ops-facing. Likely a future partner or acquirer; not a head-to-head competitor on the decisioning wedge.
  • Happy Returns (PayPal) — physical drop-off network plus portal. Adjacent. Their value prop is consumer convenience, not back-of-house decisioning.
  • Optoro — enterprise reverse logistics. Sells to Walmart, Best Buy. Builds for Fortune 500 scale; not a credible offer for a $30M GMV brand.
  • Returnly (Affirm) — instant refunds via credit. A financial primitive, orthogonal to triage.

Wedge

None of the above own the five-minute window between “customer presses Return” and “3PL decides what to do with the item.”That window is currently a spreadsheet, and it's the most expensive spreadsheet in the company.

What must be true (30–60 day validation)

Three mid-market apparel brands ($20M–$80M GMV, 100–400 SKUs, single 3PL) sign a paid pilot at $2,500/month within 45 days based on a no-code prototype that triages a sample of their last 1,000 returns and projects a margin lift of ≥2 points.

Fail signal

The founders you talk to call the output “interesting”but won't put a credit card on it. That's a no, not a maybe — don't talk yourself into a slower validation loop.

Risks worth naming

  1. The 3PL is the gatekeeper.If your top three 3PLs don't expose API hooks for triage decisions, you're selling brands a workflow they can't actually execute. Validate API access before validating willingness to pay.
  2. Loop or Happy Returns ship this themselves. Their data position is better than yours; the only durable moat is being in-market with 10+ design-partner brands before they notice.
  3. Photo quality variance is brutal. Customer-uploaded images of a $40 dress vs. a $400 coat. Triage accuracy will look great in demos and bad on a real Tuesday afternoon. Build the eval harness before the first pilot, not after.
  4. The economic buyer is not the user.CFO authorizes the spend; ops manager runs the tool. If the ops manager doesn't trust the decisions, the renewal dies quietly in month four.

Next steps

  1. Within 7 days — cold-email 30 heads of ops at mid-market apparel DTC brands. Ask for 20 minutes and one month of anonymized return data. Target: 8 conversations.
  2. Within 14 days — build the no-code triage prototype (Claude API + a spreadsheet upload) against the first dataset you receive. Show it back within 72 hours.
  3. Within 30 days — sign one paid pilot. The pilot revenue is irrelevant; the signal is that the buyer believed enough to expense it.

This is what Atlas generates for any thesis you bring it.

Generate your own memo →