Rebuilding the Future While Running the Present

Approximately 2020 — what we were replacing

I remember the work beginning in approximately 2020 as a rebuild that could not wait for the present to stop. The existing storefront carried the business while we prepared its replacement.

We were moving from SiteGenesis to SFRA. Nearly the entire storefront was being reconsidered around a mobile-first experience while customers and the seasonal calendar kept moving through the existing one.

I understood the old foundation and customer experience as no longer fitting what we were trying to operate and design for.

The change did not begin with a clean break. It began with overlap: one experience carrying the day while another accumulated plans, decisions, dependencies, and uncertainty behind it.

2020 through summer 2021 — building while running

Through 2020 and into summer 2021, the hardest part was simultaneity. We planned the replacement while operating the storefront in front of customers. Multiple web experiences had to be maintained at once. Seasonal work, product changes, and everyday decisions did not pause for the future state.

My part was bounded. I made digital merchandising and third-party hosting decisions inside my remit, shaped what content, recommendations, personalization, and measurement would have to do on the new storefront, and drew mobile-first wireframes for the homepage, listing and product pages, and the surfaces around them. Partner teams owned development, customer service, email, and the specialist analytics, channel, and asset work.

Some of that work did not stay finished. Homepage work tied to a date on the calendar was built and ready to run, and then the launch date moved, and it had to be built again. Nothing about it had been wrong. It had been made for a version of the future that no longer existed.

That was the shape of the whole period. Work was finished against a date, the date moved, and finished work became work again.

Summer 2021 — the planned launch slips

The original target was the beginning of the third quarter of 2021, as I remember it. The site was not ready, and the planned launch slipped.

There was no single dramatic decision. Plans became provisional, work was revisited, and the team kept deciding what the storefront still needed before it could carry the business. The work had already absorbed months of planning and compromise.

As summer narrowed, the holiday calendar stopped being background and became the pressure around every choice. September was effectively the last safe window; after it, Black Friday and the rest of the holiday schedule made a platform change too risky to carry.

Missing that window would not have meant waiting a few weeks. It would have meant running the storefront we were replacing through the heaviest operating period of the year, and holding the replacement until the calendar opened again.

Early September 2021 onward — launch and stabilization

I place the production launch in early September 2021, inside the window that had been closing all summer. The dates we had planned against stayed planned dates; this was the one the business actually moved on.

Launch did not complete the story. The team moved into stabilization, working through customer-facing defects while the business ran on the replacement storefront. Things a customer could see still needed fixing after the storefront was live.

Data and analytics discrepancies made interpretation harder. We were working through visible defects while reporting behavior was changing underneath them, and it was sometimes hard to separate an actual customer-behavior issue from a reporting or analytics discrepancy. I cannot collapse those conditions into one causal account.

August 6, 2026 — what I would not repeat

Years later, when I ask interview candidates what they were good at but would never want to do again, I think of this replatforming.

I could do this work well. I could hold the future state and the live business together, decide inside a shifting calendar, and keep working when a defect and a reporting discrepancy were hard to separate.

What I would not choose again is the duration of that split attention. Maintaining the present while rebuilding nearly everything behind it meant no decision belonged only to the future. A changed date reached backward into finished work. Launch opened stabilization rather than closing the effort. The calendar approached whether the replacement was ready or not.

That is what remains for me after the platform names and the planned dates fall away: this was work I could do well, and I would not choose to repeat it.