Short version: for about two days, every result on Candid Tools showed a little octopus asking you to drink some water and check out Sippy, a water-reminder app I'm thinking of building. Sippy is only a waitlist page for now: the point was to find out whether anyone wants it before I build it. 401 people saw the card. 26 tapped "I drank water". 2 clicked through to the waitlist. Nobody joined. I've taken it down. This is what the numbers said and what I'd do differently.
The idea
Candid Tools gets a steady stream of people doing slow things: removing vocals from a song, transcribing a long recording, turning a PDF into audio. Some of those runs take minutes. Sippy is an idea I want to test: a small Mac menu-bar app with an octopus that reminds you to drink water. Before building it, I put up a waitlist page to see who signs up.
The thought: people have been staring at a screen, they're waiting anyway, and "have a sip of water" is a friendly thing to say. So I swapped the feedback bar under each tool's result for a water card:
- a random "you've been at it a while" line, plus one in your local language or slang (guessed from the browser's time zone, nothing looked up)
- an I drank water button, which made the octopus celebrate and offered Get Sippy for your Mac, a link to the waitlist page
- a Not now, which hid it for the rest of the visit, and a cap of once every 15 minutes
A day later I added a second placement: a small version of the card inside the progress panel during long waits, after 8 seconds of waiting.
The data
October 20 (evening, UTC) to October 22, 2026. My own testing is excluded: visits on localhost, and the browser I use for the admin pages.
| Shown | People | Drank water | Dismissed | Clicked to the waitlist | |
|---|---|---|---|---|---|
| Card under the result | 354 | 231 | 10 | 14 | 2 |
| Hint during long waits | 220 | 190 | 16 | 19 | 0 |
| Total | 574 | 401 | 26 | 33 | 2 |
Waitlist sign-ups from all of this: 0.
Why it didn't work
1. These visitors were never going to answer the question
The question was "do people want a water-reminder app?" To answer it, you need people who already care a little about the problem. The visitors I showed it to came for something else entirely. They were here to get a song's vocals out or a PDF read aloud, and they were in the middle of doing it.
Zero sign-ups here doesn't tell me nobody wants Sippy. It tells me this channel can't tell me either way. A waitlist test only means something when the people seeing it are the kind of people who'd want the thing.
2. A tap isn't intent
"I drank water" got 26 taps, and it was the most fun part. But tapping a cute button about water is not the same as wanting a water app. The tap gave a little hit of delight. It didn't create a need. Of those 26, only 2 went on to the waitlist page, and neither signed up.
People come to Candid Tools to get a job done: get the vocals out of this song, get this PDF read aloud. A water reminder has nothing to do with that job, so the card was a pleasant interruption at best.
3. The waiting placement had no way to convert
The hint during long waits looked the best on paper. The waits are real: Remove Vocals takes over 4 minutes for the slowest 10% of runs. And it got the highest engagement: 8% of people who saw it tapped "I drank water", against 4% under results.
But that version had no link to the waitlist at all. I designed it as a gentle nudge and forgot its job was to collect sign-ups. That's my mistake, not a finding about waiting.
4. It wasn't free
The card took the slot the feedback bar normally uses. In the nine days before the swap, the bar was shown 1,400 times and rated 157 times, about 11%. At that rate, the 574 Sippy showings cost me roughly 60 ratings of my own tools. Those are what tell me which tools are broken. That's a real price for 2 clicks and no sign-ups.
What I'd do differently
- Test where the interested people are. A demand test belongs in front of people who already think about the problem: communities about health habits, desk work or Mac apps, not a site people use to convert files.
- Match the moment. A water app next to "your audio is ready" doesn't fit what the person is doing.
- Don't spend the feedback slot. Ratings are worth more to me than a promo that doesn't convert. If there's ever a promo again, it goes somewhere that doesn't push feedback out.
- Write down the success bar before launch. I only decided "no sign-ups means stop" afterwards. Next time I'll pick the number first.
- Always link to the sign-up. Every placement needs a way to say yes, or it can't be measured.
The octopus was cute, the water lines in local slang were fun to write, and people did drink some water. As a way to test demand it was the wrong channel: the wrong audience, at the wrong moment. Sippy's waitlist is still up. It just needs a better place to be found. The feedback bar is back under every result.