Last month, 508 people used Proxyceptor for the first time.
We're a small team, so that number means a lot to us. But the numbers that taught us the most weren't the flattering ones. This post walks through our first 30 days of data from Microsoft Clarity, what surprised us, and what we're fixing next.
The setup
Clarity is free, privacy-friendly, and takes about five minutes to install. It gives you session recordings, heatmaps, and behavior signals like rage clicks and dead clicks without sampling your data.
For an early-stage product with no analytics budget, it was an easy choice.
[Add 1-2 lines: what Proxyceptor does and what stack you built it on.]
The headline numbers
| Metric | Value |
|---|---|
| Unique users | 508 |
| Total sessions | 623 |
| Sessions from new users | 534 (85.7%) |
| Sessions from returning users | 89 (14.3%) |
| Most engaged user | 38 sessions |
Two things stood out:
- Most traffic is new. At 85.7% new-user sessions, people are discovering us, but we have a retention story to build.
- Some people came back. Returning sessions are only 14.3% of the total, but one user visited 38 times. We're watching those recordings closely to learn what kept them coming back.
Where users come from
India made up 82.15% of sessions (511), the US 11.58% (72), and the rest 6.27% (39). Our top cities were Mumbai, Meerut, Surat, Ludhiana, and Chandigarh.
One caution: a few of the US sessions come from Ashburn, Virginia, which is a major data center region. If you see that in your own data, some of it may be crawlers or automated traffic rather than people. It's worth segmenting before you celebrate.
Performance: an 80/100 with one clear problem
Clarity scored us 80/100 across 302 pageviews:
| Core Web Vital | Our result | Status |
|---|---|---|
| LCP (Largest Contentful Paint) | 2.4s | Good |
| CLS (Cumulative Layout Shift) | 0.001 | Good |
| INP (Interaction to Next Paint) | 320ms | Needs improvement |
Roughly 56% of pageviews were rated good, 40% needed improvement, and 4% were poor.
LCP and CLS are fine. INP is the one to fix. At 320ms, interactions feel slower than they should. Google's "good" threshold is 200ms or less, so we're in the needs-improvement range. INP is also the metric that replaced FID, so it's worth watching if your product is interaction-heavy.
Our plan for INP:
- Profile long tasks in the browser Performance panel
- Break up heavy JavaScript work on interaction handlers
- Defer anything non-critical until after first paint
The uncomfortable numbers
Clarity also surfaced these:
- Dead clicks: 36.17% of sessions (225 sessions)
- Rage clicks: 7.23% of sessions (45 sessions)
- Quick backs: 12.70% of sessions (79 sessions)
- Excessive scrolling: 0%
A dead click rate that high means people are clicking things that don't respond. Sometimes that's a bug, and sometimes it's a design problem, like an element that looks clickable but isn't. Either way, it's a to-do list sent to us by our own users.
The quick backs (people who land and leave almost immediately) tell us something about first impressions too.
What we're doing next
- Watch recordings of sessions with rage clicks and dead clicks, and fix the top offenders first
- Reduce INP below 200ms
- Filter out probable bot traffic so our numbers are cleaner
- Dig into what the returning users have in common
Lessons for other small teams
- Instrument early. Free tools like Clarity make it painless to see real behavior from day one.
- Read the recordings, not just the dashboard. The percentages tell you something is wrong, but the recordings tell you what.
- Don't hide the ugly metrics. A 36% dead click rate hurts, but it's also the most useful number we have.
- Treat early numbers as a baseline. 508 users isn't a finish line, it's the starting point we'll measure against.
Our next goal is 5,000 users, and I'll share a follow-up on what moved the numbers.
If you've used Clarity or fixed a high INP, I'd love to hear what worked for you. Drop it in the comments. 👇
Top comments (2)
the ashburn note is the most important paragraph here. datacenter ranges are the cheapest filter you can apply, and it's worth doing before you quote the 508 anywhere — your 5,000-user follow-up needs a clean baseline to mean anything.
that one user with 38 sessions is worth more than the whole new-vs-returning split. if you find what they have in common, that's your retention story.
36% dead clicks hurts, but you're right that it's where the action is.
Official Platform Update
Security protocols have been updated for all developer accounts.