Build-Measure-Kill
How Israeli startups iterate 3x faster than US peers
The American problem with experiments
American business culture loves the 'lean startup' framing - build, measure, learn. It's a great phrase. In practice, most US founders only nail the first two. The 'learn' step routinely becomes 'rationalize why we should keep going.'
We've watched hundreds of pricing experiments where the founder said, 'Let's give it another month.' Six months later, the experiment is still running. The data was clear at week 4 - the new price didn't move conversion - but sunk-cost reasoning kept it alive.
Israeli founders run experiments differently. Build, measure, KILL. The third word is non-negotiable. If the experiment didn't move the predefined metric in the predefined window, it dies on day 31. No emotional debate. No 'what if we just...' No 'one more month.' Dead.
Why kill criteria must be set in advance
The single most important step in build-measure-kill is the part most founders skip: defining what would make you kill the experiment, before you launch.
Before you launch a new pricing tier: 'We need 5 paid signups in 30 days. If we don't hit that, we kill the tier and refund anyone who signed up.' Before you launch a new marketing channel: 'We need $1 CAC less than our current channel in 60 days. If we don't, we shut it off.' Before you make a new hire on the team: 'We need 3 specific deliverables shipped by month 3. If we don't, we exit the hire.'
Most US founders set 'success criteria' for experiments. Almost none set 'kill criteria.' The difference matters because kill criteria force you to be precise about what failure looks like - which is what protects you from rationalization.
Examples from Israeli companies
Monday.com tested 7 pricing models in 18 months. Six of them died on day 31. The seventh became the per-seat freemium model that took them to IPO. They didn't shoulder all seven simultaneously - they tested one at a time, killed fast, and only kept what worked.
Lemonade killed 60% of their product bets in the first three years. Renters insurance was the original thesis; they killed it for a homeowners-first wedge. They killed the pure-AI claims process when it didn't scale and brought humans back in. The discipline wasn't 'be right'; it was 'be willing to be wrong fast.'
Wix iterated through 8 pricing pages in their first 5 years. Each one tested for 30 days. The current 4-tier structure looks obvious in hindsight - but it was experiment 8, not experiment 1.
How to apply it this quarter
Pick one thing in your business you've been 'meaning to try' for the last six months. New pricing. A new service tier. A different positioning. A new ad channel. Anything.
Define the success metric in advance. One number. Not 'see how it goes' - a specific KPI with a target.
Define the kill metric in advance. Below what number do you kill it on day 31?
Launch on Monday. Measure weekly. On day 31, check the metric. If it hit, scale it up. If it didn't, kill it that day. Don't extend. Don't rationalize. Don't 'give it more time.'
Israeli operators run two to three build-measure-kill experiments simultaneously. US operators run one experiment that lasts six months. The Israelis ship more learning in a quarter than the Americans ship in a year. That's the gap.
Key takeaways
- →Define kill criteria, not just success criteria, BEFORE launching experiments
- →30-day cycles. Kill ruthlessly on day 31 if metrics don't hit
- →Israeli companies (Monday, Lemonade, Wix) iterated through 7-8 versions before finding the winner
- →Run 2-3 experiments simultaneously, not sequentially
- →The discipline that wins is willingness to be wrong fast