Engineering backup that helped Yepoda grow from 1 to 8 stores in five years

Each time the business grew, the store had to catch up

When we started working with Yepoda in 2021, the initial request was straightforward: keep the website running. For a small team of five people using a purchased Shopify theme, that was exactly the support they needed.

However, soon it became obvious that the standard Shopify functionality wouldn't be enough. Yepoda had specific ideas about how customers should shop. The brand wanted shoppers to create their own product bundles, apply discounts, or receive gifts. Shopify didn't support most of these features out of the box, and existing apps could only cover parts of what the team had in mind. The rest required custom development.

As Yepoda grew, new challenges followed. The business expanded from one store to eight, and the backend originally built for a single market became harder to manage across six. Customer reviews accumulated in different languages, making them less useful to shoppers who couldn't understand them. The Shopify theme that had served the business for two years was also reaching its limits, just two months before Black Friday.

What started as simple retainer support gradually became a much larger partnership. Over the next five years, our work covered everything from multi-store architecture and automated review translations to a complete storefront rebuild.

Each stage of Yepoda’s growth brought new challenges and required solutions that could keep up.

Custom development built around Yepoda's sales model

Every requirement pointed to the same goal: a storefront built around how the brand sells. Shopify provided the commerce foundation and clear ways to extend it as the brand grew. Our goal was to help Yepoda go beyond Shopify's standard capabilities, improve the customer experience, and build a reliable foundation for business growth.

In practice, that meant a set of custom features that connect Yepoda’s sales rules to what shoppers see in the store. Here’s how they took shape, one stage of growth at a time.

2021 → The selling mechanics Shopify didn’t come with

Yepoda wanted to boost sales by encouraging new customers to try more than one product and converting more campaign traffic into orders. To do that, the brand needed custom shopping features: build-your-own sets and campaign links that applied discounts to every price on the page before customers reached the cart. Shopify didn't support either out of the box .

Set Creator. A customer picks between three and eight products, with the range configurable in metafields, and the price recalculates as items go in and out. The admin side runs without us: a collection carrying a bundle template, products with a Bundle option and a variant per bundle size. On the storefront, collection data is written out as JSON, prices are recalculated dynamically in the browser, and placeholder cards swap to real images as the customer builds the set.

The discount modal. A Lambda service takes the code from the URL, queries Shopify through the official API, checks whether it is live and what type it is ( fixed amount, percentage, or Buy X Get Y), then returns the value and the full product list, expanding collections into their products. The storefront walks the product cards, recalculates each price, writes a compare-at price where none existed, and updates the discount badge. Modal copy is editable in the theme customizer through ${codeName} and ${discountValue} placeholders, so campaigns go live without a developer.

Nothing on the app market did the bundle builder in 2021, and nothing does it this way now.

2022–2023 → One backend, built for one store, now serving six

The discount service had one store’s name and credentials hardcoded because, when it was built, there was only one store. France, Italy, and Germany then got their own, each with their own credentials. The alternative to a rewrite was eight copies of the same application, with every rule change made eight times.

One deployment that knows which store is asking. We rewrote the service as a multi-store application, with the store resolved per request and credentials looked up rather than compiled in. A visitor lands on the Italian storefront, and the backend fetches Italian data.

That took the engineering work out of a market launch. The rest stayed with us: configuring the existing bundle and discount features for each new store, setting up the products, writing instructions for the team, and working through the launch on calls. Poland went live that way, then the UK and the US.

A language modal, alongside it. Shopify’s browsing_context_suggestions endpoint identifies the visitor’s location. If a dedicated store exists for their country, we offer it, and if not, we point them to the EU store. The choice is stored in a cookie for 14 days, and it suppresses the discount modal for two minutes so the two don’t fire at once.

2023 → Preventing a costly technical mistake

Headless was trendy back then, and Yepoda were interested in using Instant Commerce (lower costs, better performance, a site the team could assemble themselves). The risk was that nobody had checked the platform against what their storefront actually did.

A technical review instead of a migration. We assessed Instant Commerce against the existing functionality and found it couldn’t reproduce roughly half of it. Not through customisation, not through a workaround, the platform itself was limited. Total cost of ownership would have gone up rather than down.

We didn’t refuse the work or overrule the client. We showed them specific features they already had that the platform couldn't support, and what maintaining the result would cost. They stayed on Shopify and migrated to 2.0 instead.

Autumn 2023 → A redesign and re-platform with two months deadline

The Shopify 1.0 theme had reached its limits, so we recommended moving to 2.0. Yepoda went further and suggested pairing the migration with a full visual redesign and launching it for Black Friday. Final designs arrived in August, leaving two months to rebuild the storefront, migrate years of content, and re-implement every custom feature on a 30-hour retainer.

More capacity, at the same rate. We expanded the team on the retainer. The hourly rate stayed the same, and the hours went up.

One decision that protected the schedule. The piece we worried about most was gift-with-purchase. We'd built it up over two years on the old theme, and carrying it across meant taking it apart first. The team reviewed it and concluded a clean rebuild on the new theme would be faster and safer than the migration.

The site launched on time, on Shopify 2.0. For the brand, the immediate difference was that marketing could build and rearrange pages in the admin instead of filing a ticket for every layout change.

2024 → Reviews in languages the shopper speaks

In beauty, reviews can close sales. Stamped's cross-store feed put reviews from all eight stores on every product, but in whatever language they were written in. For instance, a German review on the Italian page, unreadable to the shopper looking at it. The problem was also that the app couldn't provide a proper translation.

A translation pipeline on AWS Lambda. When a review is created or updated, Stamped fires a webhook. A handler identifies which store it came from, pulls the full review from the Stamped API, and queues it on SQS. A processor detects the original language through Google Cloud Translation and translates the title, body, merchant reply, location, and questionnaire answers into the six store languages, writing each version to its own Elasticsearch index. The original is kept separately, a language that already exists isn't translated again, and an updated review only refreshes what changed.

The storefront reads from our API rather than Stamped’s, with filtering by product, text search, pagination, and sorting by date, rating, helpfulness or featured status.

A shopper in Milan now sees reviews from every Yepoda store, in Italian.

2024-2025 → Cart and checkout following different rules

Marketing ran campaigns constantly, each with its own rule (gifts at a threshold, gifts tied to a code, gifts live for two dates in November, tiered percentages, per-item Multibalm pricing). The cart could handle them. Checkout belongs to Shopify and theme code doesn't work there, so an expired gift or a sold-out variant could reach the payment step unchecked.

The rules moved into Shopify Functions and checkout extensions, all configured through metafields by the Yepoda team:

• Specific discounts – a code tied to a named product list
• Tiered discounts – four cart-value bands with their own percentages, plus an exception list for products that shouldn’t discount
• Progress gifts – a free gift unlocked at a cart threshold, calculated after other discounts
• BOGO and cart drawer gifts – tied to a trigger product, with an expiry date on the offer
• Checkout validation – extensions that re-check gift eligibility, apply codes that should apply, drop expired BOGO gifts, and remove out-of-stock variants with a message rather than letting the customer hit a failed order

Now, the offer a customer sees in the cart is the one they pay for.

2026 → Making Yepoda AI-search-ready

Shoppers looking for skincare products increasingly turn to AI assistants like ChatGPT for recommendations. If Yepoda isn't mentioned, it risks losing sales to competitors. Worse, Shopify and Google Analytics often classify AI-driven traffic as referral or direct, leaving the brand unsure how much revenue comes from AI.

To address this, we built AiVisible, our Shopify app available on the App Store since September 21, 2026. It's currently running on Yepoda's French and German stores.

The app helps Yepoda track AI-driven sales and recommendations:

• Sales and traffic analytics – sessions, orders, and revenue from ChatGPT, Claude, Perplexity, and Gemini, with order and landing-page reports.
• Recommendation tracking – weekly checks of real shopper queries to see whether and where AI assistants recommend Yepoda.

It also helps improve Yepoda's visibility in AI search:

• Product audits – checks descriptions, shipping, returns, FAQs, and Schema.org markup, with a list of recommended fixes.
• llms.txt – a structured catalogue map updated daily to help AI assistants understand the product range.
• Bot access monitoring – nightly checks to detect blocked AI crawlers that could prevent Yepoda from appearing in recommendations.

Other Shopify brands can install AiVisible to see how often AI assistants recommend their products, what they might be missing, and how much revenue AI search brings in.

€65M+ in revenue, 1M+ customers, on storefronts we've maintained since 2021

Yepoda's growth is Yepoda's as they built the brand, raised the money, and won the retail deals. What we can say is that for five years, our team built and maintained every storefront growth ran through, and none of it slowed them down.

Here are some more details we're proud of:

Record-breaking 2-month full store rebuild

Full visual redesign and Shopify 1.0 → 2.0 migration, delivered on time.

1 → 8 storefronts

From a single DE/EU store in two languages to eight markets, on shared infrastructure rather than eight copies of the same application.

2 → 6 languages

English and German at the start; English, German, Italian, French, Spanish, and Polish today, with every review in the fleet available in each of them.

€65M+ revenue in 2024

120%+ YoY profitable growth and 1M+ customers, running on storefronts we build and maintain.