AI · Shopify · Field notes

Storefront Experiments

Building an online store with AI — real experiments, results and lessons.



One real store. A record of what works, what breaks, and what we learn.

Building a Shopify store with AI: the lessons from ERVALO so far

Can AI help build an online store that holds up beyond the first screenshot? That is the question behind Storefront Experiments.

Our working example is ERVALO, a Shopify store being developed in Finnish and English. The work has included brand decisions, theme changes, product pages, navigation and repeated checks. This first article looks at the lessons from that process so far.

Start with the store you want to build

Before changing a theme, we needed a clearer idea of the brand. The project moved from the working name NORDRA to ERVALO, with MOVE and CARE as the main parts of the store. That decision affected more than the logo: page titles, menus, images and translations all needed to tell the same story.

A useful brief names the audience, the products and the job of each page. For this project, keeping terms such as MOVE, CARE and SCULPT consistent gave the design work a direction. Without those decisions, it is easy to spend time polishing a page that will need to be rebuilt later.

A page can look finished and still need work

Some of the most useful checks happened after the design looked close to ready. On mobile, the Finnish and English language buttons overlapped the account and cart controls. The desktop layout looked fine, so a desktop-only review would have missed the problem.

Other issues were less visible. An English product page still had a Finnish product-recommendation heading. A size-table note needed to be removed. Prices in the homepage product section used inconsistent decimal formatting. Each issue was small, but together they affected how carefully the store appeared to have been built.

The lesson was to review the actual shopping journey: open the homepage, change language, visit a collection, read a product page and check the cart. A screenshot cannot show whether that journey works.

Two languages mean checking the whole journey twice

Translating paragraphs was only part of the bilingual work. Navigation, footer links, contact pages, related products and cart text also needed attention. A customer can enter through an English page and still end up following a link to Finnish content.

For ERVALO, the practical approach was to walk through the store in each language and record where the experience changed unexpectedly. That made the next correction specific: fix one route, one label or one section, then repeat the same journey to verify it.

Keep changes small enough to check

A broad request such as “improve the store” leaves too many decisions open. More focused tasks were easier to assess: move the mobile language controls, correct an English heading or make the price display consistent.

Theme copies were also useful during larger changes. They gave us a place to review a new direction before putting it into the active store. The important step came after the edit: return to the page and confirm the result in context.

Design cannot answer every launch question

A polished storefront does not settle product availability, supplier arrangements or the reliability of product information. ERVALO has also involved work on suppliers and on which products are ready to be offered. A “coming soon” section can communicate the current stage, but it cannot substitute for those decisions.

That distinction matters when documenting progress. A theme correction is evidence that a page improved. It is not evidence that the business has achieved sales, that a supplier has been confirmed or that every launch requirement has been resolved.

What this project will document next

Storefront Experiments will collect the practical details behind the work: the problem we found, the change we made and the checks we used to assess it. Future articles will look more closely at bilingual storefronts, brand consistency and preparation for launch.

The most useful lesson so far is simple: AI-assisted work still needs clear decisions and careful review. ERVALO gives us a real project in which to test that process, one correction at a time.

You can follow the accompanying videos on the Storefront Experiments YouTube channel and the shorter updates on Instagram. Both are linked in the sidebar.


Discover more from Storefront Experiments

Subscribe to get the latest posts sent to your email.

Comments

Leave a Reply