SAP integration
A two-way update model, covering what used to live in a manual process.
- Pricing system and discount groups
- Categories and products
- Cart price rules
- Customers, orders and stock
Case · Lindt Brazil
45 stores became one e‑commerce in nine days.
March 2020. Malls closing, fourteen days to Easter, and an entire retail chain with no online sales operation in Brazil. We took over the brand’s digital front and put the store live, with every physical unit acting as stock and as the starting point of delivery.
Retail · Premium chocolate
The starting point
Fourteen days to Easter,
and no online store.
Lindt had a substantial physical chain in Brazil and no e-commerce operation in place. This was not migrating an existing store: it was creating the whole operation, from the catalogue to the last mile.
The pandemic compressed the deadline to the limit. Malls were closing, Easter is the strongest date in a chocolate brand’s calendar, and it was fourteen days away. Either the chain sold online, or it did not sell.
The decision was to turn each of the 45 physical stores into an online store, with a fully digital operation and presence on the delivery apps as well. The first sale went through on 4 April 2020.
-
Phase 1 · 26 to 28 March
Environment and catalogue
Installing and releasing the staging and production environments, and loading the catalogues and products of all 45 units.
-
Phases 2 and 3 · 29 to 31 March
Store and checkout
Adapting the layout and the backoffice and checkout processes to the distributed model, blocking cross-selling between units.
-
Phase 4 · 1 to 3 April
Team and launch
Training the store teams on the backoffice process and preparing the launch of the online store.
-
Go-live · 4 April
Live
Store open, running with all 45 units as stock and as the origin of delivery.
What had to be solved
Building a site was never the problem.
Making 45 stores act as one was.
Every unit had its own stock, its own point of sale, and a delivery radius that overlapped the store next door. The architecture had to solve that before the first order came in.
Architecture of the 45 units: a central catalogue of product and price feeds a delivery router, which takes the customer postcode and decides which unit serves it. Each unit keeps its own stock and point of sale, and selling between units is blocked.
And the delivery radius
Chocolate does not travel well. Each unit’s reach is a quality limit before it is a logistics one, and it is what decides which store serves each address.
Delivery radius schematic: three stores, each with its own maximum reach. Where two reaches cross there is an overlap zone, in which more than one store could serve the order and the router picks by real distance. An address outside every reach gets an immediate answer instead of becoming an order.
- Maximum distance per origin Each unit has its own limit, adjustable per city. Chocolate has a temperature constraint, so the radius is a quality decision before it is a logistics one.
- Overlap zone Where two radii cross, more than one store could serve the order. The router picks by real distance, not by registration order.
- Out of range A postcode no unit covers gets an answer immediately, instead of becoming an order that never ships. And it lands on an unmet-demand panel.
-
45 distributed, independent units
Each store is at once an online shop window and a distribution centre. An order has to be fulfilled by the unit that best serves that address, not by a central warehouse.
-
Delivery radius and overlap zones
Two nearby stores compete for the same postcode. We had to design a delivery radius limit and a convergence rule that decides which unit serves it, with no conflict and no coverage gap.
-
Stock behind an independent POS per store
Stock did not live in a single system: each unit had its own point of sale. Online availability had to reflect the real shelf of each store, all of the time.
-
Payments approved without manual review
With high volume and a short window, there was no room for a review queue. The gateway had to approve automatically, with anti-fraud on, and nobody in the middle.
-
Physical stores on the same platform
Store staff started working inside the online store’s platform. One system, two channels, and no reconciliation spreadsheet at the end of the day.
-
No cannibalising the brand in search
The 45 online stores could not compete with each other in search engines. Each unit was kept out of the index, and the main domain took over both the corporate site and the online store.
How an order moved
Robots did everything.
Except taking it off the shelf.
From the click to the delivery close-out, the order moved on its own. There was exactly one stop, and it was physical: someone taking the chocolate, packing it and saying it was ready. That proportion is what let the operation absorb peaks without hiring a technology team.
The order’s path
7 steps. 1 goes through a person.
- 6 automated
- 1 human
- 01
Purchase and approval
The customer places the order and the gateway returns the approval. Nothing waits for a check.
approved - 02
Invoicing and queue
The order is invoiced and joins the queue of the store that will serve it, chosen by distance.
invoiced - 03
Picking and packing
Someone takes the product off the shelf, packs it, issues the invoice and marks the parcel ready. It is the one step you cannot automate with physical stock across 45 stores.
in preparation - 04
Shipment
Agents generate the shipment and dispatch the carrier or courier, with nobody opening a system.
on the way - 05
Collection
Transport arrives within minutes and collects at the store. At peak, it became a queue of couriers waiting for parcels.
- 06
Delivery
It leaves from the nearest store, so the final leg is short by construction.
- 07
Close-out and confirmation
The delivery closes itself out and the customer gets the confirmation.
delivered
At peak, the bottleneck stopped being the system and became the shelf: couriers queueing for parcels, and the shopkeeper simply handing them over. That is the kind of bottleneck you fix with more people on the floor, not with a rushed technology project.
What was delivered
Nine days to go live.
And the list that fitted inside them.
The operation went live with what it needed to actually sell, not with the minimum required to say it was up. Every stream below landed before or right after the first order.
-
Operations
- Logistics integration
- Stock integration
- Advanced order grid
- Payment gateway moved to Adyen
-
Delivery
- Postcode mapping and delivery page
- Postcode without dashes, validated automatically
- Origin calculated by real distance
-
Acquisition
- SEO implementation
- Feeds for Google Shopping and Facebook
- Landing page with lead capture
- Campaigns and storefront updates
-
Storefront
- Gift bags and gift cards
- Gift message on the order
- New browsing and conversion features
What was measured
A result is a number, not an adjective.
-
9 days
from kick-off to go-live
From 26 March to 4 April 2020, with catalogue, checkout, backoffice and a trained team.
-
99.995%
availability across five years
The quality of the first quarter became the annual standard of the operation, held year after year since 2020.
-
1.1M
users on the platform
82.8% new and 17.2% returning, across campaign reach and organic search in the period.
-
45
stores acting as e-commerce stock
Every unit selling, picking and dispatching. The chain is now past 50.
And in the three months that followed
- Logistics integration and stock integration
- SEO implementation and feeds for Google Shopping and Facebook
- Postcode mapping and delivery page, with automatic validation
- Payment gateway switched to Adyen
- Gift bags and new experience features
- Landing page with lead capture and an advanced order grid
Project Reset · eShop 2.0
The store that got through the emergency
was rebuilt for scale.
Five years later, the operation born to get through an emergency needed something built for scale. Projeto Reset rebuilt the e-commerce from scratch, as a bespoke solution for Lindt’s business: unified stock, logistics consolidated at national level, and the store network sustaining digital rather than competing with it.
-
Stock
one store, one stock pool
unified national stock
-
Delivery
orders split across shipments
delivery from the nearest store
-
Back office
spreadsheets between systems
two-way SAP integration
-
Omnichannel
physical stores competing
the store network sustaining digital
Stock, delivery and omnichannel
The physical chain stops competing with digital and starts sustaining it.
- Integration with CISS and the in-store pinpads
- Distributed stock for availability and delivery origin
- Automatic and manual geolocation
- Split delivery with a distinct SLA per origin
- Click and collect with a configurable preparation time
- Carrier integration, with tracking included
Store and conversion
Features that shorten the path between wanting to buy and a paid order.
- Smart search with AI that learns from queries and clicks
- Related products by category, with the same learning loop
- Product personalisation with a preview of ribbons, tins and packaging
- Bundles, kits, and kits assembled by the customer
- Passwordless login, with a code sent by email
- Smart repeat-purchase checkout and a cart shareable by link
- Shipping cost and instalments shown on the product page
- Moderated reviews with a score and photos
Security and compliance
Controls go into the design, not into the eve of the audit.
- Privacy centre for data protection, with consent management and a channel to the DPO
- Triple-factor authentication on the backoffice admin
- Top security score on the public certificates, verifiable on Qualys SSL Labs
- Clearsale anti-fraud and a purchase returns module
Two haapit Labs products running inside the operation
Software we conceived and operate, wired into the store like any other component of the architecture.
-
Rota302
Routing and administration of delivery zones per unit: radius limits, activating and deactivating stock points, and real-time visibility of the addresses being attempted at checkout.
-
Mag Shipping
Carrier integration, with configurable priority per partner and automatic fallback when one of them does not respond.
The model
We are the support, development,
operations and security team.
Lindt did not build a technology department in order to run a digital operation at international standard. The capability is contracted and the standard is maintained: no headcount to recruit, no on-call rota to staff, no redundancy to fund, no training to replace at every departure, and no knowledge concentrated in a key person.
-
Own headcount and turnover
A complete team for support, development, operations and security, with replacement that stays invisible to the client.
-
On-call cover and 24x7 redundancy
Continuous cover already sized, across three tiers, with no on-call to hire and no rota to administer.
-
Continuous training and re-skilling
Expertise kept current across platform, integration, cloud and security, with no learning-curve cost.
-
Key-person risk
Knowledge documented in runbooks and spread across the team, so no departure interrupts the operation.
State of the art is not a presentation adjective. Day to day it means stock unified at national level, routing by real distance, a single human step out of seven, the highest security score on the public certificates, and compliance designed alongside the product rather than added on the eve of the audit.
Back to all casesHave a physical chain that still does not sell as one?
It is the same problem with a different product on the shelf. A technical conversation at no cost to look at your inventory, your point of sale and your last mile.