He needed a big retailer's data to build the product. No big retailer gives data to a company with no product. Felix Hoffmann is the co-founder and CEO of 7Learnings, a Berlin company whose software forecasts demand for every product at every price, then sets the price that hits a retailer's goal. He spent six years as a pricing consultant at Kearney and two years running price optimization at Zalando, Europe's largest fashion marketplace, where he saw predictive pricing working at scale. From his consulting years he already knew almost nobody else worked that way. The problem was that a forecasting model needs a large retailer's sales history, and nobody hands that to a company with no product. So the first contract 7Learnings signed was not software at all. It was a consulting project: the retailer got help implementing its own pricing approach, 7Learnings got paid, and it kept the right to use the data to build a product of its own. The first paying software customer came through his old consulting network, structured as an A/B test where the algorithm priced half the assortment and the retailer's team priced the rest. The first run was a disaster, far too expensive on high-priced products. They reworked the models, and a later test came back with a 13% profit uplift. Ten customers took 7Learnings to its first $1M ARR, and Felix closed every one himself. Stay for 33:32 where Felix explains why he refuses to put an LLM in a pricing decision, and what he thinks that means for every founder currently building a wrapper. SPONSORED BY HOBBES Hobbes runs personalized product demos on your website 24/7 and books qualified buyers straight onto your team's calendar. Don't book a demo, take one: https://saasclub.io/hobbes š KEY LESSONS šÆ Solve The Data Cold Start By Selling Something Else: 7Learnings could not train a forecasting model without a large retailer's sales history, so it sold a paid consulting project and kept the right to use that dataset. š¤ Shrink A Scary Ask Into A Reversible Test: Retailers would not hand pricing to an algorithm outright, so the first pilot ran as an A/B test on half the assortment while the retailer's own team priced the rest. š Pick An Early Customer Who Can Survive A Failure: The first live pricing run was badly wrong. It survived because the buyer had a big enough problem, no alternative, and understood they were working with a startup. š° Keep Pricing Simple Even When Value Is Provable: Felix charges a monthly fee scaled to the revenue under optimization and refuses success-based fees, because unclear pricing reads as risk and piles pressure onto the pilot. š° Price High Enough To Lose Some Deals: If nobody walks away because you are too expensive, you are too cheap. Especially for a complex product carrying real delivery cost. š Founder-Led Sales Lasts Longer Than You Expect: Felix closed all ten customers behind the first $1M ARR himself, and stayed closely involved through the next forty, because handing off enterprise sales is genuinely hard. ā” Pick The Technology After The Problem: Felix argues founders are all digging in the same technical space. Decisions that need determinism, low cost and explainability should not be handed to an LLM. ā±ļø TIMESTAMPS 00:00 Intro 01:55 What 7Learnings does 03:28 Where the idea came from: Kearney, then Zalando 04:44 The hardest part was finding co-founders 06:56 Why they needed a big retailer's data, and the consulting project that got it 09:08 Why the retailer shared its data 11:30 Finding the first paying customer 13:06 Structuring the first deal as an A/B test 13:49 The first upload was a disaster 16:07 A 13% profit uplift 17:10 How a pricing company prices itself 18:58 Ten customers to $1M ARR, and where the next nine came from 25:05 The price matching objection 33:32 Why LLMs don't belong in the pricing decision 39:20 Is SaaS dead? The wrapper trap 44:19 Lightning round š§ Full Show Notes: https://saasclub.io/494 š Get weekly 5-minute SaaS insights: https://saasclub.io/email #SaaS #FounderLedSales #SaaSPricing #AI