We started Cadence with a 7-day free trial. Founders loved it. Engineers absorbed the cost, and quality dropped. After 6 weeks of testing 48 hours instead, we shipped 48 as the standard. Trial-to-paid conversion went from 41% to 67%. Founder NPS rose +18 points. Engineer participation in the free tier grew 3.4x, including senior and lead-tier engineers who would never accept a 7-day free arrangement.
This is the post about why 48 hours is the right number, what we learned about trial economics, and why the structure that founders ask for isn't always the structure they want.
Our first version of Cadence had a 7-day free trial. The reasoning was straightforward: founders need time to evaluate fit; engineers need time to ramp; a week is plausible for both.
The math broke:
The structural problem was adverse selection in slow motion. Every week of free work that didn't convert to paid pushed the better engineers off the platform. The free tier became a negative signal for engineer quality. By month 4, our matched candidates were noticeably weaker than the broader pool of engineers who'd signed up but weren't taking trials.
We tested shortening the trial to 48 hours. The results were immediate.
The headline numbers from the test:
| Metric | 7-day trial | 48-hour trial | Delta |
|---|---|---|---|
| Trial-to-paid conversion | 41% | 67% | +26 pts |
| Engineer participation in free tier | baseline | 3.4x | +240% |
| Founder NPS | baseline | +18 pts | +18 pts |
| Senior + lead tier participation | <15% | 41% | +26 pts |
| Average time to first commit | 38 hr | 27 hr | -11 hr |
The change in senior and lead tier engineer participation was the leading indicator. Engineers at $1,500 and $2,000 per week have outside options. They won't trade 30 hours of free work for an unproven booking. They will trade 8 hours.
That single shift, opening the platform to engineers who wouldn't touch a 7-day free, changed the quality distribution of the matched pool. Founders started seeing meaningfully stronger candidates. The conversion improvement followed.
Why not 24 hours? Why not 72? We tested both.
24 hours. Too short for the engineer to ship anything meaningful. The trial felt like a meeting with code; founders didn't get the work-evidence signal they needed. Conversion at 24 hours was 51%, between the 7-day and 48-hour versions.
72 hours. Engineer participation was similar to 48 hours (the marginal 24 hours doesn't change the financial calculus much). Founder conversion was the same 67%. The extra day didn't help; it just delayed the decision. Some founders pushed the trial through to day 3 just because they could, not because they needed to. We dropped back to 48.
48 hours. Roughly the time required to ship something small. The engineer onboards Monday morning. They ship a feature by Wednesday morning. Both parties have evidence to decide. Most founders make the keep-or-release call in the morning of day 3.
The 48-hour window is short enough to be cheap for the engineer (8 hours of work, not 30) and long enough to be evidentiary for the founder (one shippable thing, plus communication patterns).
The thing we learned that surprised us: founders don't actually want a 7-day free option. They want certainty in 48 hours. We discovered this by interviewing 30 founders post-trial. None of them said "I wanted more free time to decide." All of them said "I wanted to know faster."
The 7-day trial was answering a question founders hadn't asked. They thought they wanted runway; what they actually wanted was a tighter loop. The trial wasn't a sampling exercise; it was a decision deadline. Pushing the deadline out made the decision worse, not better.
This generalizes. Most "give me more time" requests in software hiring are reframed: founders want better evidence, not more time. A 48-hour trial with a real shipped artifact gives better evidence than a 7-day trial with vague code reviews.
The 7-day trial was a bad deal for engineers; the 48-hour version flipped the math.
Cost of the trial. 7 days × ~4 hours/day = 28 hours unpaid. At $50/hour (a US senior contractor's blended rate), that's $1,400 of opportunity cost. For a senior engineer, that's a hard sell unless the booking has a high probability of converting.
48-hour version. ~8 hours unpaid. At $50/hour, $400 opportunity cost. Acceptable for a probabilistic conversion to a $1,500/week booking. The math is straightforward.
The trial isn't free for the engineer; it's a marketing cost. The right structure is one where the engineer's cost is small enough that they'll take the trial and convert at the higher rate, not large enough to filter out the strong engineers.
Tighter feedback loops produce better engineering work. This is not a Cadence-specific insight; it's a general one. We see it everywhere we look:
The 48-hour trial is the smallest practical free-trial unit that produces real shipped evidence. The next-smaller unit (24 hours) produces interview-quality evidence. The next-larger unit (7 days) produces the same shipped evidence but at meaningfully higher engineer cost.
Cadence's whole model is built around picking the smallest practical unit at every layer. Booking instead of hiring. Weekly instead of monthly. Daily ratings instead of quarterly reviews. 48-hour trial instead of 7-day. Each shrinkage of the unit improves the feedback signal.
Practical guidance for founders running their first 48-hour trial:
Day 1 (Monday). Engineer onboards. You set up access (repo, Slack, Linear, deployment). You agree on one shippable feature for the trial. Daily standup at 5pm or async equivalent. Review the engineer's first commit by end of day.
Day 2 (Tuesday). Engineer ships. You review the work. You give a thumbs up or thumbs down on the daily rating. If the engineer is clearly off-fit, you can release them mid-trial; no charge.
Day 3 (Wednesday morning). Decision: keep or release. Most decisions are obvious by this point; the trial isn't a sampling exercise, it's a decision deadline.
If you keep them, weekly billing kicks in. Junior $500/week, mid $1,000, senior $1,500, lead $2,000 per the locked pricing tiers. Cancel any week.
If you release them, no charge. We shortlist 4 new engineers for the next intro calls.
Try the 48-hour trial yourself. Book on Cadence, describe what you need, take 30-minute intro calls today, pick one for the trial. Worst case, you've burned 30 minutes. Best case, you have a working engineer by Wednesday.
Two things we got wrong before we landed on 48 hours, worth flagging:
We assumed founders wanted maximum optionality. The 7-day version maximized optionality. Founders kept asking for more time during the test phase; we extended for some founders to 14 days as a courtesy. The 14-day cohort converted at 38%, worse than the 7-day baseline. Optionality doesn't help; it dilutes the decision.
We assumed engineers would absorb the cost. Senior engineers explicitly told us they wouldn't. Mid-tier engineers absorbed the cost but selected against the platform on subsequent bookings. The cost to the platform of a 7-day free trial wasn't paid by Cadence; it was paid by the engineer pool, and the engineer pool degraded as a result.
We tested 24, 48, and 72 hours. 24 was too short for shippable evidence. 72 produced no improvement over 48 and just delayed decisions. 48 hours is roughly the time required to ship one small feature, with both parties having evidence to decide.
No. The 48-hour cap is structural; it's what makes the trial economically work for engineers. If you're not sure at hour 48, the answer is usually "release and try a different candidate." Founders who try to extend almost always end up making the same release decision they would have made at hour 48, just later.
No. The trial is unpaid by design; that's what makes it a real evaluation rather than a paid first week. Engineers absorb the cost (~8 hours of work) as a marketing cost. The 67% conversion rate makes this economically rational for engineers; the math wouldn't work at 7 days.
Most founders who say this discover by hour 60 that the engineer is actually fine and the founder was overthinking. The trial deadline forces a decision; the decision is usually clearer than the founder feels at hour 48. If the engineer is shipping and the founder is on the fence, the right move is "keep them for week 1 and re-evaluate at week 1 boundary." Weekly billing makes that low-cost.
Yes for technical fit (shipped artifact, code quality, tooling fluency). Yes for communication style (Slack messages, written status, response patterns). Less so for long-term cultural fit, which takes 4-6 weeks regardless of trial length. The trial is for go/no-go, not for committed-marriage.
Yes. After intro calls, you can pick more than one engineer for parallel trials. Each enters their own 48-hour window. After the trials, keep any combination. Each engineer is on a separate weekly subscription so you can cancel them independently.