Phased Migration: How to Launch a Redesign Without Losing Traffic

This article is part of the Flatstudio × Stavka.tv case study. Written for product teams and developers who need to do more than just ship a new design — they need to move a live audience onto it without losing traffic or search rankings. Main article in the series — here.

‍

Three years of work. One domain. The forecast: a 25–50% traffic drop and roughly a month to recover. That's what a major migration usually looks like — first a fall, then a slow climb back. We prepared for the worst and got something completely different.

Why Major Redesigns Drop at Launch

There's a mistake almost every team makes during a big redesign: they evaluate whether the design is ready, but not whether the transition is. They think about how good the new site looks — not about how to move a million people onto it without Google deciding the site has disappeared.

With Stavka.tv, we understood the risks early. The platform runs on organic traffic — SEO isn't a channel for them, it's oxygen. Over 3 million visits a month by the time version 2.1 was ready to launch. If Google "lost" the site for even a week during reindexing, that would be a real hit to both users and ad revenue.

So we didn't just launch. We migrated.

Three Waves Instead of One Day

We broke the launch into three phases.

First wave (September 8, 2022). New.stavka.tv — a separate domain with noindex. Access was given only to "Pros" and "Experts" — the most active and loyal users, roughly 2% of the audience. These people understood the product better than anyone: they noticed small things, gave specific feedback, and didn't leave after the first confusing change. Perfect testers — not because we labeled them that way, but because they'd made themselves that over years of using the platform.

New.stavka.tv stayed invisible to Google. The old version on the main domain kept getting indexed and collecting traffic as usual.

Second wave. Opening the new version to all registered tipsters — 100% of active users. At this stage we also deployed popups and an onboarding flow for the new interface: people could see the changes but weren't left to figure everything out alone.

Third wave (October 19, 2022). The switch. The new version becomes the main stavka.tv domain with indexing enabled. The old site moves to old.stavka.tv and stays there for a few more weeks — in case something went wrong and a rollback was needed.

Why We Moved Faster

The third wave was originally planned for later. But in October 2022, we realized: exactly one month remained before the FIFA World Cup in Qatar.

Stavka.tv lives on football. The World Cup is the biggest event of the year. Launching a new site in the middle of the tournament — or right after — means burning your most valuable traffic window on users getting used to a new interface. Better to ship everything early, give it a week or two to stabilize, and enter the tournament already running on the new version.

October 19 — decision made, transition done.

What the Numbers Showed

Two weeks after the transition, Viktor Titov posted a summary in the team chat.

"We predicted that our traffic would drop 25–50% after the migration and take about a month to recover. But we managed to smooth out almost all the disruption — we barely dipped even in the short term, and within a week we were back to baseline numbers."

More specifically, here's what changed:

On top of that, October 2022 turned out to be a record month: 3 million visits and 900,000 unique users — the best numbers in the platform's history at that point.

These metrics didn't "recover after a drop." There was almost no drop. They improved immediately after the transition and stayed at the new level.

The Number That Surprised Us Most

Bounce rate and time on site are expected metrics to track. But +100% profile page visits is a different conversation.

In version 2.1, we substantially rebuilt the profile system: rankings, prediction statistics, divisions. People started visiting each other's profiles — not because we pushed them there, but because the system became interesting enough to do it voluntarily. Doubling profile visits in the first month meant the social mechanic had worked the way we'd intended back at the concept stage.

If traffic is quantity, then time on site and pages per visit are quality. Profile visits are a signal about whether users experience the platform as a community — or just as a collection of pages.

What We Did Right (and What We'd Do Differently)

Three things that worked:

Noindex on the test domain. While the new version lived on new.stavka.tv, Google didn't see it. Zero risk of duplicate indexing, traffic cannibalization, or confused ranking signals.

The old site stayed live. old.stavka.tv kept accepting traffic for several weeks after the switch. If something had gone wrong, we could have flipped back. That parachute was never used — but knowing it existed gave us the confidence to act decisively.

Documentation of temporary decisions. Every project has places where a developer did something "for now, we'll fix it later." Without documentation, those places live forever. We agreed: every temporary decision in the code gets written down separately, with an explanation of why and what's planned instead. This kept technical debt from accumulating as unknowns.

What we'd do differently: the first wave was planned for September 1st and actually happened on the 8th. A week's delay — because of last-minute fixes that always surface at the end. Next time, we'd build that buffer in officially rather than treating it as "just a few more days."

The Migration Is a Product Too

The most important lesson from this launch: the go-live strategy needs to be designed with the same care as the product itself.

Designers often think their responsibility ends where development begins. Or that a release is the PM's and DevOps's problem. But if the migration fails, the design fails with it. Users don't distinguish between "a poorly designed site" and "a poorly executed launch."

We thought about the risks before writing the first frame in Figma. So on October 19, 2022, when the new version became the main domain, it wasn't a jump into the water — it was a planned crossing over a bridge we'd been building with the client for months.

If your product has reached the stage of a major redesign and you're thinking about how not to break what's already working — that's exactly the moment to talk about Post-MVP Evolution. Not after launch. Before it.

‍

← Back to the main article in the series: "Sports Predictions Platform: An 8-Year Case Study with 3M Monthly Visits"

Other articles in the series:

→ Rebranding a Sports Platform: How We Built a Brand System Across 10+ Touchpoints

→ From 2.0 to 2.1: How We Rewrote Two Years of Work Without Losing the Client

→ Promo for Bookmakers: Why Headers Convert at Zero and Popups Actually Work

→ Design Systems for Complex Products: Why It's an Investment, Not an Expense