How we migrated New Balance to FastStore in three days

New Balance's Brazilian store now runs on FastStore. We moved its entire traffic over in three days, in stages, with no big-bang cutover.
New Balance Brasil is one of the Grupo Dass brands we operate, alongside Fila and Umbro. The store already had the traffic. It ran on a VTEX IO storefront with years of customization that had passed through many hands. Its Google mobile PageSpeed score sat at 37, and that capped everything else.
A migration in three days
First we rebuilt the store. Over a few weeks we remade the home, product, and listing pages on the FastStore design system, so the new storefront was ready and tested before any traffic reached it.
Instead of flipping the whole store at once, we moved traffic to FastStore in stages, from 25% to 50% to 75% to 100% over three days (October 5 to 7). Each stage ran on real shoppers before the next one opened, so any problem would surface on a slice of traffic rather than the whole store.
The performance jump showed up the moment the first stage went live. Google's mobile PageSpeed went from 37 to 92, and desktop from 33 to 100.
PageSpeed is a lab test. What matters is what real visitors feel, and Chrome's field data (CrUX) showed the same move.
On VTEX IO
- Mobile LCP 4.1s
- Mobile INP 226ms
- Desktop LCP 3.4s
- Desktop INP 78ms
On FastStore
- Mobile LCP 2.7s
- Mobile INP 157ms
- Desktop LCP 1.5s
- Desktop INP 64ms
The change showed up in how people used the store. New Balance's own team pulled these numbers and took them to the group's board.
“Rolling traffic up in steps takes the drama out of a go-live. Every step tells you, with real shoppers on the store, that the next slice can go in.”
The business result
Over the first 21 days on FastStore, conversion from direct and organic traffic rose 9.3% against the 21 days before the migration.
Total sessions and orders fell the following month. That drop was seasonal, the end of a launch and promotion calendar that had inflated the earlier weeks, not a performance regression. We call it out because comparing two windows is only fair when you say what else changed between them.
Keeping the store fast after go-live
The migration is where the work starts. We ran the same one for Fila and Umbro, the other two Grupo Dass stores, and all three now sit inside one operation where a team of people and agents watches and defends performance every day. A store that is fast on go-live day and then left alone is slow again by the next quarter.





