Danny Florián

StrikeFeed

Self-initiated. Designed, built and launched in seven days

2018

Designed, built and launched a working consumer product in 7 days, in public, for 5,000 designers who had never watched one go all the way out.

Role
Everything. Research, brand, UI, front end, CMS, server, launch.
Team
Me. Plus a wife, family and friends who agreed to be interviewed.
Timeline
2018. Seven days, first commit to public launch.
Owned / influenced
Owned all of it, which was the point. Influenced nothing — there was nobody to influence.
Level of detail

Full case study — context, explorations, tradeoffs and how each number was measured. The short version. Switch to deep dive for context, explorations, every tradeoff and measurement notes.

The business problem

I had mentored 5,000 designers who had never seen the unglamorous 80%.

Over a couple of years I mentored more than five thousand designers across various platforms, and the same request kept arriving: document your process, start to finish. The gap was never taste. It was that almost none of them had seen what happens between a nice-looking screen and a thing on the internet that strangers can use — the CMS schema, the deploy, the decision to use a worse tool because you already know it. Portfolios optimise for the artifact. Nobody publishes the boring 80%.

So I set a constraint: pick a real consumer premise, and take it from interviews to public launch in seven days, documenting every decision including the bad ones.

The premise itself came from interviewing the people around me. When a business drops the ball and will not make it right, the consumer’s only real leverage is public noise — and the review channels that exist are largely controlled by the business being reviewed. Some businesses have no review surface at all; the ones that do can bury a bad review under glowing ones. The hypothesis: a platform that broadcasts bad experiences applies enough public pressure that the business resolves them.

The reframe

The constraint was the curriculum.

A seven-day cap is not a stunt, it is a forcing function that makes every tradeoff visible and therefore teachable. Given a year, the interesting decisions disappear under polish. Given a week, you have to say out loud “I am using CraftCMS and PHP because I already know them and a JavaScript stack costs me two days I don’t have” — and that sentence, spoken by someone with no incentive to look clever, is the thing five thousand designers had never heard anyone say.

The work

Sketch, brand, browser, CMS, server, launch — in that order.

StrikeFeed complaint feed with the top complaint of the day, company logos, and an upvote control on each entry.
The feed, modelled on Product Hunt: complaints grouped by day, ranked by votes, with the day’s loudest at the top. Getting the ranking edge cases right in Craft’s templating took most of day five.
Pencil sketches exploring the StrikeFeed feed layout and card hierarchy.
Day one. Sketching is the cheapest way to find where the important elements live before any of it costs anything.
The StrikeFeed logo being constructed in Sketch: two overlapping X marks forming an inverted heart.
The mark. Three strikes and you’re out, drawn as crossed X’s rather than baseball’s K — and in the overlap, an upside-down heart. Nothing says “I don’t love you any more” quite like that.
StrikeFeed login and sign-up screen, designed directly in the browser.
Designed in the browser, not in Sketch. With SASS organised on atomic-design principles, screens made mostly of existing components never needed a comp.
CraftCMS Twig template code showing the conditional logic behind the complaint feed ordering.
The part portfolios never show. Feed ordering is a stack of conditionals, and publishing them was more useful to mentees than any screenshot on this page.
Animation showing a company logo being fetched automatically from a website URL.
One field, two jobs. Typing a company URL fetches its logo through Clearbit and creates the company record — which is how two confusing forms became one.
Twitter timeline showing automated tweets generated for each new complaint, tagging the company.
Distribution, wired with a Zapier webhook: every new complaint tweets itself, tags the company, and links back. Public pressure was the mechanism, so the mechanism had to be automatic.

Explorations that died

  • 01 / 03

    Two forms: complaint, then company

    Users had to author a company profile before they could complain about it. Died on day four when the Clearbit logo API let a single URL field do both jobs.

  • 02 / 03

    Threaded discussion per complaint

    It turned a pressure mechanism into a forum, and moderation for a forum is not a seven-day feature at any level of ambition.

  • 03 / 03

    Business response and resolution

    The actual point of the product — and there was no honest way to build the business half in the days remaining. I shipped without it and said so in the launch post.

Tradeoffs

What got cut, and why.

  • The entire business-facing side. The hypothesis only tests if the business being complained about can respond, and day seven arrived with only the consumer half built. Correct call for the deadline; also the reason the core loop never closed. Both of those things are true.
  • Moderation. A report button and me. No queue, no policy engine, no appeals. Perfectly fine at zero users, indefensible at any scale, and written into the launch post rather than quietly omitted.
  • The better technology choice. CraftCMS, PHP, MySQL and Laravel Forge over a JavaScript stack, purely on my own familiarity — roughly two days saved. A worse long-term choice and the correct seven-day choice, and the distinction between those two is most of what I was trying to teach.
  • Comping every screen. Login and sign-up were designed directly in the browser. On a project made almost entirely of existing components, a static comp of a form is a day spent drawing something the CSS already knows how to do.

Impact

Interviews to public launch in 168 hours.

  • 7 days

    First commit to a public, working launch.

    Measured: 168 hours elapsed, documented daily — research and sketches, brand, UI, CMS build, server and deploy, feed logic, launch.

  • ~5,000

    Designers reached by the process write-up.

    Measured: mentee audience across mentoring platforms at time of publication. It remains the most-requested thing I have published.

  • 100%

    Of the stack documented publicly, including the regrets.

    Measured: every tool and service listed by name — CraftCMS, Laravel Forge, Ubuntu, Nginx, MySQL, Gulp, SASS, Mailgun, Clearbit, Zapier — with the reasoning, including choices I would now make differently.

Reflection

The product’s hypothesis was never really tested, because public pressure only works if the business can respond and I ran out of week before building that half. As a teaching artifact it did exactly what it was meant to: every question I got afterwards was about deploys and CMS schemas, not about the logo. Next time I would cap the scope at the loop rather than at the calendar.