Food and fine goods. Stores built for repeat, occasion and freshness.

The difference sits in the second purchase. Fashion gets bought, coffee gets bought again. A store that only optimises the first order leaves the larger share of its contribution margin on the table. So it is decided early whether subscription, reordering and the customer account are designed together or retrofitted years later.

On top of that, the goods age. Batches, roast dates and best-before dates are not a footnote; they shape what you are allowed to sell, what you want to ship first and what is better bundled. That information lives in the ERP, not in the store; the store only has to read it reliably.

And revenue arrives in waves. Christmas, Easter, Mother's Day, corporate gifting: occasion business creates peaks that hit a grown setup harder than everyday trading. At BlumeIdeal, exactly that instability during seasonal peaks was one of the triggers for moving from Shopware 5 to Shopify Plus.

A subscription is not a discount model, it is a product model. The first decision is which product can be subscribed to at all, and at what rhythm. With coffee the rhythm follows consumption, not the calendar. Ship 250 grams and one kilogram on the same cadence and you produce gaps and cancellations.

The second decision concerns control inside the customer account. Move the date, swap the variety, change the quantity, pause: every one of those actions a customer can handle alone is one cancellation less and one service email less. At BlumeIdeal the subscription logic runs on Recharge, connected to the existing Shopify checkout.

The third decision is the handover into the ERP. For logistics a subscription order is an order like any other and has to arrive with batch, weight and delivery date attached. Build the subscription as an island next to the merchandise system and you have only moved the work into accounting.

For coffee and tea, sensory detail is the actual sales advice. It therefore belongs in the product data model as structured fields, not as prose inside a description. Only in structured form can it be filtered, pulled into comparisons, used in recommendations and, later on, read reliably by machines.

At Blankroast Manufaktur those very attributes came out of the weclapp ERP: aroma, origin, roast level and acidity were mapped as Shopify metafields. The team keeps maintaining the data in one place, and the store can make it filterable and surface it directly on the product page.

The effort rarely sits in the technology, it sits in agreeing on values. A roast level with seven free spellings filters nothing. Before a migration it pays to check every property for whether it filters, whether it informs, or whether it merely sits in the catalogue for historical reasons.

A requested delivery date is only as good as the rule behind it. The calendar in the ordering flow may only offer dates that are genuinely possible in logistics. For that it needs three inputs: blocked days, a cut-off time per shipping method, and current availability from the merchandise system.

At BlumeIdeal we built exactly that logic: weekends and defined public holidays are blocked automatically, the cut-off time shifts the earliest possible date, and availability reaches the store through a JTL worker cloud. In the end the customer only sees days that can genuinely be selected.

Freshness also affects stock. If batches and best-before dates are held in the ERP, the store does not have to calculate anything itself. It only has to learn in good time when a batch should no longer be sold. That handover belongs in the integration design, not in a plugin added afterwards.

Gift buyers are different customers. They often do not know the product, they buy once, under time pressure and for somebody else. So the store has to take part of the choice off their hands: tasting sets instead of variety selection, bundles instead of single items, a concrete delivery date instead of a shipping time in working days.

At BlumeIdeal a configurator solves this: pick a base product, add extras such as a vase or chocolates, write a personal greeting card, and take matching upsells in the same step. The higher average order value in that project comes precisely from those configurator upsells, not from a price increase.

At Sirocco the same mechanic appeared in a different form: variants, bundles and gift sets had to fit into a product structure that stays maintainable. That is where gifting becomes technically expensive, namely when every set is built as a special case rather than as a reusable pattern.

Stock, prices and batches belong in the ERP. The store consumes that data, it is not a second source for it. As soon as both sides are allowed to maintain the same fields, you get discrepancies nobody resolves cleanly, usually in the week before the largest peak of the year.

In this industry B2B almost always comes along too: hospitality, offices, resellers, sales representatives. At Blankroast the structure for that was laid out from the start without diluting the D2C experience. The store is also prepared to use Shopify POS in the company's own roastery later on.

And then there is weight. Coffee, preserves and drinks are heavy, so shipping costs are a real block of cost rather than a side calculation. Tiers by weight, zone and shipping method belong modelled properly, not as a flat rate that eventually eats the margin of the whole subscription business.