Our work up close.

Four in Europe, two in Bulgaria and one in Australia.

A view of the system: Automotive parts distribution

Automotive parts distribution

Problem

The client ships car parts from five warehouses across Germany, Poland and the Netherlands. Each warehouse kept its own stock list, and the online shop was updated by a file import once a night. By morning the numbers were already wrong: customers were ordering parts that had been sold the previous afternoon, and someone had to call them back to apologise.

Solution

We connected all five warehouses to VertaStream. Stock now moves as events. A part leaves a shelf in Hamburg and the shop knows about it in under half a second. Even in the slowest five per cent of cases it takes under 1.4 seconds. Reservation rules hold a small buffer for each shop and marketplace so two customers can never claim the same part, and every order is routed to the warehouse that can actually ship it.

Results

241k
Items under sync
1.4s
Propagation, 95th percentile
96%
Fewer cancellations
VertaStreamSAP Business OneEvent streamingPostgreSQL
A view of the system: Home goods retail

Home goods retail

Problem

One warehouse, a shop and two marketplaces, and no agreement between them. On busy weekends the same lamp was sold on Shopify, Amazon and bol.com within the same hour. The team spent Mondays cancelling orders and writing apologies, and the cancellation rate on both marketplace accounts was climbing towards the two per cent ceiling bol.com allows before it restricts a seller.

Solution

VertaStream now owns the stock and lends it to the shop and the marketplaces rather than the other way round. Each of them gets an allocation with a small buffer held back during peak hours, and the moment something sells anywhere, VertaStream stops offering it everywhere else immediately. The shop follows within a second, bol.com usually within fifteen minutes, Amazon as soon as its feed queue allows. The Monday morning cancellation ritual stopped completely.

Results

0
Oversells in 14 months
3
Places kept in sync
0.4%
Cancellations under the 2% bol.com limit
VertaStreamShopifyAmazon SP-APIbol.com API
A view of the system: Pharmaceutical distribution

Pharmaceutical distribution

Problem

Batches and expiry dates lived in spreadsheets. Every inspection meant three days of someone reconstructing who moved what, and where it went. In this industry one mistake with an expiry date is not a bad review. It is a licence problem.

Solution

We put batch and expiry tracking into the same system that moves the stock. Every movement is signed and kept for ten years. EU good distribution practice sets a five-year minimum, so that is twice as long. Anything approaching its date is quarantined automatically before a picker can reach it, and an inspector now gets a full history exported in a few minutes instead of three days.

Results

12 470
Batches tracked
100%
Audit coverage
0
Expired units shipped
VertaStreamBatch trackingSigned audit logGDP compliance
A view of the system: Fashion e-commerce

Fashion e-commerce

Problem

A catalogue of more than 100,000 products fed by 38 suppliers, each sending data in its own format and its own idea of what a product name is. The old platform needed hours to import a single feed, duplicates piled up, and adding a supplier meant writing another one-off script that only one person understood.

Solution

We rebuilt the whole thing on a platform where this size is normal. Supplier feeds are normalised and deduplicated on arrival, attributes are mapped automatically, and the whole catalogue rebuilds inside VertaStream in six minutes. Each shop and marketplace then receives the changes as fast as its own API allows. Adding a new supplier is now configuration, not a development project.

Results

104 720
Products in catalogue
6 min
Full catalogue refresh
0
Records lost in migration
VertaStreamFeed normalisationDeduplicationData migration
A view of the system: Retail & fulfilment

Retail & fulfilment

Problem

Forty-one fulfilment sites across Australia and an order router that decided by distance alone. Orders went to the nearest store regardless of whether it had the goods, so items were shipped twice, split badly, or missed the carrier cut-off and sat overnight.

Solution

We joined the client's engineering team as a partner on the routing layer. Decisions now weigh live stock, distance and carrier cut-off times together, split shipments are calculated rather than guessed, and the whole decision takes under a second at nine thousand orders a day.

Results

9 400+
Orders routed daily
41
Fulfilment sites
0.8s
Routing decision
Distributed systemsOrder routingAWSEngineering partner
A view of the system: Manufacturing

Manufacturing

Problem

The production report that the management team reads on Monday morning was assembled by hand from three places: the ERP, the machine data and two spreadsheets kept by shift supervisors. Two people spent most of Friday and part of Monday on it, and the numbers were argued about at the meeting rather than acted on.

Solution

The three sources are pulled on a schedule, reconciled against each other and published before the first shift starts. Where the sources disagree, the report says so instead of silently picking one. Nobody assembles anything, and the argument at the meeting is now about the production rather than about the numbers.

Results

41 min
Instead of two days
52 / 52
Runs without a person
3
Sources reconciled
Scheduled pipelineERP exportPostgreSQLReconciliation rules
A view of the system: Insurance

Insurance

Problem

Every policy that needed a second opinion travelled by email. Who had it, how long they had had it and what they decided lived in individual inboxes. Cases sat for days because nobody knew they were waiting, and an audit meant asking people to search their mail.

Solution

Cases are checked for completeness on arrival, routed to whoever should see them and timestamped at every step. The decision stays with the underwriter. What changed is that the queue is visible, the waiting is measured, and the trail is written as the work happens rather than reconstructed afterwards.

Results

4 h 10 m
Median time to decision
100%
Cases with a full trail
0
Files lost in inboxes
Workflow engineAudit trailRole routingEmail intake

More projects

A view of the system: Real Estate

Real Estate

Property Expert

We built a full-scale property platform with smart search, interactive maps for every Dubai district, and an automated enquiry pipeline that routes leads directly to the right agent. The team manages everything through a custom CMS. No developer needed for day-to-day operations.

Next.jsTypeScriptCMSGoogle Maps API

The first step is a short conversation.

We reply within 24 hours. A detailed brief is not needed beforehand.