Hello World

Tasuki was supposed to be an AI consulting agent. It still might be. But the consulting product has a longer runway, and while we wait for Nick's Nashville numbers to tell us whether billboards actually move the needle, we figured we'd run our own experiment.

We had a brand that hadn't launched yet, a name that sounds more like a daily word game than a consulting firm, and thousands of digital screens we could buy programmatically through Dropsignal's tools. So we used them.

The bet

The knock on out-of-home advertising has always been the same three things. You can't measure it. You can't click it. And nobody is going to type a URL they saw on a billboard.

Tasuki is a bet that all three are wrong.

The idea: every morning, a new code appears on digital screens across a city. Type it into your phone, answer a trivia question, and if you get it right, you earn the chance to write tomorrow's question. Your name goes up on screens. The screen itself becomes the product. Not an ad for something. The thing.

Think of it as a public feed on physical screens. A URL that changes daily, published to bus shelters and urban panels the same way you'd publish a post. The screen is the channel. The city is the audience.

To make that work, the content needs to update itself. New slug every morning, new question behind it, no one manually re-uploading files. That's dynamic creative. We built the whole pipeline in about a day: a responsive HTML page that calls an API for the daily slug, rotates automatically, works at any screen size. Packaged it into a ZIP file per the Vistar spec, uploaded all seven screen formats, and started targeting screens.

The experiment

We wanted to test the thesis in the most literal way possible. Put a URL on screens in San Francisco. See if anyone types it.

The creative was just a URL on a warm off-white background. `tasuki.co/hello-world`. Nothing else. No logo, no tagline, no instructions. If you know, you know. And in San Francisco, most people walking past a bus shelter would know. Wouldn't they?

Behind the URL: a trivia question targeted (in this case) at one specific person. A friend who bikes to work through the city every morning. We mapped his commute, found 96 digital screens within a few blocks of his route, and targeted 34 of them. Bus shelters on Market Street. Urban panels near his office.

Then we spent more time than anything else in the entire experiment battling the Vistar UI, convinced something was wrong with our ZIP file, our API call, the upload spec. Screens kept coming back with "no qualifying venues." We debugged everything. Repackaged the ZIP. Re-read the docs. Nothing worked.

Then it clicked. There was nothing wrong with any of it. Clear Channel doesn't accept dynamic HTML on their screens. Only static images. I spent twelve years at Clear Channel Outdoor. I know exactly why. One VP of digital ad ops was dead set against it. Always citing the risks to the business. He only allowed his own proprietary implementation. That policy is still in place.

If I'd thought about it for five minutes before building the pipeline, I would have remembered. Other operators do support dynamic creative. But the screens along this specific route are mostly Clear Channel. We rendered static PNGs and moved on.

Then we submitted the creatives for operator approval.

The reviews

Each screen network reviews creatives before they run. Most of the time operators just glance at the image and approve it. But our creative was nothing but a URL. The only way to review it was to type it in.

So they did.

All of a sudden we started getting site visits. Philadelphia. New York. Indianapolis. Cities where we'd never had any traffic. Ad ops reviewers across the country were typing `tasuki.co` into their browsers, clicking through to Archway in the footer, poking around. We were watching the pings come in real time, getting that dopamine hit. The thesis was working. People see a URL on a screen and they type it.

Then two operators denied us. One came back with `UNACCEPTABLE_CONTENT: does not meet advertising policy requirements`. We got a little nervous. The other spent two minutes and forty-one seconds on a page with seven words, then wrote: "I do not know what this company/brand is, and when I go to the website it is blank with no information." Fair enough on both counts.

We updated the homepage to say "A daily game on city screens" and resubmitted. Eleven of twelve operators approved after that.

The creative approval process had generated more engagement than our actual campaign would. Our most enthusiastic audience turned out to be the people deciding whether to let us on their screens.

Tuesday morning

The campaign ran Tuesday, 8 to 10 AM. Two hours. Thirty-six dollars. Almost three thousand impressions across the SF metro.

The venue targeting we set through the API didn't constrain delivery the way we expected. Instead of staying on the commute route, impressions scattered across the entire metro. Grocery stores in Walnut Creek. Bus shelters in the Sunset. Some of the targeted screens along Market Street did fire, mostly near our friend's office. But the signal was diluted across a hundred and thirty venues instead of thirty-four.

We'd both made predictions before the campaign ran. How many people would visit the site. How many would type the slug. How many would answer the question.

The actual number, across every metric: zero.

Nobody typed the URL. Nobody visited the site. Nobody answered the question.

Our friend texted that morning. He'd been looking for it. Didn't see a thing. Most of the screens are on bus shelters, he said, easy to see from the sidewalk but hard to catch from the bike lane.

The hello-world question is still up there, by the way, if anyone wants to take a crack at it.

What it means

Three thousand impressions. A cold brand with no context. A two-hour window. A URL on a white background. Zero conversions. That's the baseline.

It doesn't prove the thesis wrong. It proves a two-hour flight with diluted targeting in one city isn't enough to prove it right. People did type the URL from screens. They were ad ops reviewers in Philadelphia, not bike commuters in San Francisco. And the impressions that were supposed to land on 34 screens along one route ended up on 130 screens across the metro. Wrong audience, wrong screens.

For thirty-six dollars, that's a reasonable thing to learn.

What's next

Longer flights. More days. Fix the venue targeting so impressions land where we aim them. Build recognition through repetition instead of hoping for a single-morning miracle.

Eight episodes in. Dropsignal has a customer whose downloads went down. Tasuki has a game that nobody played. The Archway homepage still describes products that are searching for proof more than traction.

Is this guy going to figure it out? Good question. So is the one at `tasuki.co/hello-world`.