Context
Segmentic is a marketing-automation platform built for the Iranian market: a segment builder, a campaign sender, and a visual journey canvas, backed by a real-time event pipeline built for the scale of companies like Digikala, Snapp, Tapsi, and Cafe Bazaar. I'm a designer and frontend contributor on a small team; the backend, the event pipeline, and most of the delivery engineering are other people's work, not mine. What I own is the surface a customer actually opens: the dashboard, the bilingual Persian/English panel and landing page, and the UI for the campaign and segment builders.
The reframe
Partway through, a human review of the landing page turned up something the code alone couldn't catch: the marketing copy and the actual product had drifted apart. The panel called something an 'event' everywhere; the landing page called the same thing something else in one spot. A report card promised 'churn prediction', when the product deliberately doesn't show a predicted score, for reasons written elsewhere in the same codebase. Small mismatches, but each one meant a customer reading the site and then opening the dashboard was looking at two different products. I went through the landing page line by line and matched every claim to the exact word the panel uses for that feature, not the word an earlier draft of the roadmap used.
What I built
My work is the surface a customer actually uses. Dashboard charts and KPI cards, redesigned for readability. A fully bilingual build: Persian and English panel and landing page, with copy checked against each other rather than translated once and left to drift. Brand presentation across error pages and the wordmark, so a broken page still looks like the product and not like a server complaining. And detail work on shared components: the channel picker's segmented control now slides its selection between options instead of swapping backgrounds, because a marketer comparing channel previews side by side needs to see where the selection came from, not re-read the whole row to find it. The slide position is measured from the DOM rather than an index, since Persian labels change width the moment the webfont loads, and a ResizeObserver keeps it correct if the drawer resizes or a font swaps in mid-session.
What I killed
I dropped the appearance settings page. It held three controls, light mode, palette, and paradigm, and all three already lived in the header drawer, one click from anywhere on the site. The settings page was a second place to change the same three values, and it was the only reason a couple of supporting components existed at all. Removing it took out the whole route, its sidebar entry, and the components that had no other caller. The one real cost: a reset-to-defaults action that only that page exposed. The underlying function is still there; nothing in the UI calls it now, and that's a small debt I'm aware of, not one I've paid off yet.
Outcome
Segmentic is still being built, so there's no launch number to report. What I can say is the state of it: automated tests in the thousands, passing clean with the race detector on, and a measured throughput in the low thousands of events per second on a single laptop running the full stack at once, which isn't the same as production load but is a real, repeatable number rather than a claim. The rest of the specifics, customers, revenue, exact scale, belong to the company, not to me.
Where my role stopped
I'm a designer and frontend contributor on a small team, not the product owner and not the engineer who built the backend. The event pipeline, the messaging infrastructure across eight channels, the analytics engine, and most of the platform's actual engineering are other people's work. What's mine is the interface: the dashboard, the bilingual panel and landing page, and the shared UI components I built or revised.
What I know now
Working inside someone else's architecture is a different discipline than owning a product end to end. I can't decide what the backend supports; I can only make sure the surface tells the truth about it, which is most of what the landing-copy pass and the churn-prediction wording fix actually were. The lesson I'm carrying forward is a plain one: a product with more than one person writing customer-facing copy needs an actual pass that checks the marketing words against the product's own words, on a schedule, not just when someone happens to notice the drift.
A note on numbers
Some figures here are given as direction rather than absolutes, and some are omitted entirely, to respect what is confidential to Segmentic. Everything stated is my own and does not necessarily reflect the views of Segmentic.