Tsuyu's first ad budget: $280, 14 days, one question
TL;DR: Tonight I switched on Tsuyu's first Google Ads campaign: US search only, $20 a day, ending November 9, so about $280 at most. The question it answers is narrow. When a stranger who searched for exactly this kind of tool lands on the site, do they start a trial? Getting here took a product pivot, a real demo video, and conversion tracking I'm willing to defend in the privacy policy. I'll post the numbers when it ends, good or bad.
Why spend money now
My own rule for every product is that a launch only counts once the experiment actually runs: real strangers arrive, and the funnel in front of them is honest. Tsuyu has passed the second part for a while. The first part, reach, has been close to zero. A handful of people found it, a few signed up, and almost nobody started the trial.
With traffic that thin, I can't tell "nobody wants this" from "nobody saw this". Paid search is the cheapest way I know to buy a clean answer: people typing things like "story video maker" or "faceless YouTube video generator" are already looking for something like this. If they won't start a trial, the problem is the product or the page, not reach.
The plan
- Where: Google Search only, US only. No display network, no search partners, no "AI Max" broad matching. On $20 a day with zero conversion history, letting Google improvise mostly buys loosely related clicks.
- What: 4 ad groups covering story videos, faceless channels, Reddit-style stories, and horror and true-crime narration. 76 keywords, 48 negative keywords to keep out job seekers, free-tool hunters and the like, and 4 responsive search ads. Every ad sends people to the page that matches its keywords, not the home page.
- How much: $20 a day, max $1.50 a click, November 9 end date. Worst case about $280.
- Bidding: "Maximize clicks" for now. Smart bidding needs conversion data I don't have yet. Once about 15 trials have come in from ads, I'll switch to bidding for trials directly.
- What counts: two conversions. A trial started, which is the goal Google bids toward and is valued at $15, the price of the cheapest plan. A signup, which is recorded for context but doesn't steer bidding. Optimizing for signups would teach Google to find people who make an account and leave, which is exactly the pattern I already have.
There's no free tier. That's deliberate, and it's the main thing this test puts under pressure. The trial asks for a card up front, and that's where people stall today.
Measuring it without being creepy about it
Conversion tracking means the site now loads Google's tag, which it never did before. So the privacy policy says so in plain words: what's recorded (the ad click ID, the conversion event, and your email hashed in your browser before it's sent), that EU, UK and Swiss visitors start with ad storage switched off, and how to opt out. Existing users will get an email about the change, because the policy promises one.
On my side, the first ad click someone arrived from is stored with their account, so I can follow an ad click all the way to a paid plan. Google's dashboard stops at the conversion, and the paid plan is the number I actually care about.
What changed in Tsuyu to get here
Most of the last two weeks went into making the product worth paying to send people to.
It's a storytelling studio now, not a Reddit-to-video tool. Narrated Reddit posts are a format that's been played out, and the platforms have been tightening on reused content. What still pulls views, even with rough art, is a good story. So Tsuyu now leads with your own script or an AI draft. Links to found stories still work; they're just no longer the point.
A demo you can actually watch. People signed up and then wouldn't enter a card. Fair: nothing showed them it was real. There's now a permanent demo account and an unedited walkthrough recorded on it, from pasting a story to downloading the finished video, on the home, pricing and sign-up pages. No staged screens. I re-recorded only the narration, once, because the narrator pronounced "Tsuyu" with one syllable. It's two.
Series, episodes and characters, in beta. Behind the beta flag, Tsuyu can plan a season from a story bible, write each episode with a cliffhanger, and voice dialogue with each character's own voice. I wrote and rendered three episodes of a courtroom mystery to test it. The newest piece animates a single scene into a short clip with the character's look and voice held steady, at about 30 cents a second in real model cost, which is why it's billed per second. None of this is in the public pitch until it's out of beta.
Pages for people searching for this. New guides on what microdramas are, the one-minute vertical drama format that's growing fast in the US, plus a matching use case and ideas page. They're slow SEO, not ad landing pages, but the format and the product finally match.
One mistake worth admitting
While wiring up the conversion tracking, I shipped an update from an old copy of the code instead of the current one. For a few minutes the site ran weeks-old code. I caught it, redeployed the right version, and confirmed the live pages had the new changes before moving on. The fix to my process: build every deploy from a fresh copy of the latest code, and stop if that step fails rather than carrying on. "The health check passed" only proves the site is up. It doesn't prove it's the code you meant to ship.
What happens next
The ads are in Google's review queue as I write this. I'll check search terms after a couple of days to prune junk queries, review at day 7, and judge the whole thing on November 9 against the same continue/park/pivot rules I use for every product. The follow-up post will have the real numbers: spend, clicks, signups, trials and cost per trial. If the answer is "nobody starts a trial", that's worth knowing too, and it'll cost less than another month of guessing.
Product-by-product detail lives on the roadmap.