Business and Sales · Tabnews · 🇧🇷 Brazil
Discovering the problem wasn't the model, it was the question asked to customers
An entrepreneur improved a demand forecasting model, but only made progress in sales after changing the question asked to customers.
- Profession
- Entrepreneur, Data Scientist
- Tools
- Machine learning, Python
- Source
- tabnews.com.br ↗
- Country
- 🇧🇷 Brazil
- Report date
- September 9, 2026
- Verified on
- October 24, 2026

What they did
He built a demand forecasting model and went out prospecting for customers.
How it went
The error (WAPE) dropped from 44% to 31%, but 160 direct messages resulted in zero customers. The turning point came when he asked about the customer's pain, not about the model.
Good to know
Validating the problem before the technology saves months.
The prompt
I'm going to interview owners of [type of business]. Write 8 open-ended questions about how they decide on [inventory/purchasing] today, without mentioning my product, and 3 signs that the pain is real.
Replace what is in [brackets]. Prompt written by us from the account, not a literal copy of the source. Do not paste other people’s personal data (privacy laws such as LGPD and GDPR apply).
Summary in our own words. Results as stated in the original account; we have not tested or audited them. Read the source ↗ Found a mistake?
Similar cases
01Separating business expenses from personal card statements
The author had Claude talk through the first few statements to learn her patterns before classifying the rest.
02Giving an agent a server and the mission of building a business
A dev let Claude Code run every 30 minutes on a VPS, with memory kept in files, asking for help via Telegram.
03Building an app at 66 without knowing how to code
An Argentine numerologist in Belo Horizonte used AI to create Aurameets, which connects patients with holistic therapists.
04Using several AIs as critics to validate an idea
A developer created a "tribunal" of 25 AI personas to judge ideas without the flattery typical of chatbots.
05Build prospecting lists from Google Maps
A developer built a local lead extractor with Python and Playwright, speeding up development with AI.
06Let an AI send prospecting emails on your behalf
The founder of BloodLink authorized Claude to write and send emails to blood centers and hospitals.
07Pay-by-Weight Restaurant Lets AI Draft the Menu for the Owner to Approve
The owner of Cantinho das Delícias, in Diadema, uses a ChatGPT assistant that proposes the menus.
08Disputing a contractor's prices on a building project
A project owner gave the AI every item, the purchase order and the emails; it found charges he hadn't noticed.