Standing Up a Demand Forecast in a Week
Research
Demand ForecastExperimentStartupUrbanHarvest

Standing Up a Demand Forecast in a Week

A practical guide from our UrbanHarvest experiment on how to decide with confidence when you have zero historical data.

Published Sep 05, 2025 • 5 min read

One of the biggest hurdles for any new venture is the unknown. For 'UrbanHarvest,' a D2C grocery startup, the question was existential: which neighborhood should they expand into next? With five potential options but zero historical sales data, making the wrong choice could mean wasting tens of thousands of dollars on a failed launch. This is where rapid, intelligent experimentation comes in.

The Challenge: Deciding Without Data

Traditional market research is slow and expensive. Surveys can be misleading, and demographic data alone doesn't capture intent. We needed to find a way to measure actual demand from real potential customers in each target neighborhood, and we needed to do it in under two weeks.

A team collaborating around a whiteboard with charts and graphs. Data-driven decisions start with a clear hypothesis.

The goal wasn't to be 100% certain. The goal was to be 75% confident, which is more than enough to de-risk a major business decision.

Our Experiment: A Micro-Funnel Approach

We designed a simple, seven-day experiment. We built a micro-funnel combining hyper-local search ads on platforms like Google and Facebook with a simple waitlist landing page. The ads were targeted geographically to each of the five neighborhoods, with messaging tailored to local interests.

  1. Define Success Metrics: What would a 'successful' neighborhood look like?
  2. Build Micro-Landing Pages: Simple, focused pages for each neighborhood, capturing interest via a waitlist.
  3. Hyper-Targeted Ad Campaigns: Utilizing social media and search engines to reach specific geographic areas.
  4. Monitor & Iterate: Daily tracking of ad performance, sign-ups, and user feedback.
  5. Analyze & Rank: Comparing results across neighborhoods to identify the strongest contenders.

Tools We Used

  • Google Ads for hyper-local search intent
  • Facebook Ads for broader demographic targeting
  • Mailchimp for waitlist management
  • Google Analytics for tracking page performance
  • Custom Python Script for data aggregation and initial demand modeling

The Outcome: Data-Backed Confidence

The result was a ranked list of all five neighborhoods, complete with projected CAC and a 75% confidence score. The startup moved forward with the top-ranked area, saving an estimated $50,000 that would have been spent on a launch in the second-worst performing neighborhood. The process became a reusable playbook for future expansion decisions.

Back to Research