High-load data at global scale
The platform serves users from branded sites around the world. Every one of those interactions is also a data point that has to be captured, not dropped under load.
A global where-to-buy provider needed solution architecture from scratch and a platform that holds up under high load on branded sites worldwide. Bintime built both, and the 23-person team that runs them.
PriceSpider
Visit the client’s sitePriceSpider (now Wayvia) has led the way in where-to-buy solutions and services since 2011. It works with leading global brands, retailers and agencies, helping boost sales across online and offline channels and building trust between brands and buyers by taking care of the customer journey.
In March 2015 the company received a €3.5 million investment from Vortex Capital Partners, which opened the way to double-digit growth and offices in Amsterdam, Bangkok, Kyiv and Taipei. Its integration partners include Samsung, Lenovo, Braun, Canon, Microsoft, P&G, Walmart and Amazon.
The project was considered a challenging one for two reasons: it required managing high-load data, and the solution architecture had to be created from scratch. There was no existing system to extend and no reference load to design against.
The platform serves users from branded sites around the world. Every one of those interactions is also a data point that has to be captured, not dropped under load.
Creating the solution architecture from scratch meant the early technical decisions had to survive a decade of growth rather than the first release.
Delivering it required building an experienced technical team of 23 members — and relocating three engineers to the Netherlands to work closer to the client.
The development team helped PriceSpider handle hundreds of requests per second. That is what made the client’s business goal reachable: serving users from branded sites around the world without the widget becoming the slowest thing on the page.
Delivery went out as an MVP and was extended feature by feature afterwards, so the architecture was proven under real traffic before the analytics and API surface were built on top of it.
Kafka and RabbitMQ carry events; Apache Spark does the heavy processing, so ingestion is never blocked by analysis.
Java, Scala and Spring for the services on the hot path, Python where flexibility is worth more than raw speed.
MongoDB for the document model, Elasticsearch for search and aggregation across gathered statistics.
Docker, Kubernetes and GitLab CI make scaling and deployment routine rather than an event.
Prove the core under load, then add the widget, the dashboards and the API — in that order.
A dedicated team of 23 experienced professionals, three of them relocated to the Netherlands. Over the years the model evolved from outsourcing to outstaffing.
Experienced professionals assembled for the project, with three relocated to the Netherlands to sit closer to the client’s business.
The widget gathers statistics under high load. It runs on brand sites worldwide and still records every interaction.
Extensive reporting gives brands insight into customer behaviour, so where-to-buy stops being a black box between click and purchase.
Partners integrate on their own terms rather than around a fixed contract, which matters when the partner list includes Amazon and Walmart.
The engagement model changed as trust grew: the same people, now working as part of the client’s own organisation.
The Bintime team continues maintaining the solution that brought PriceSpider to where it is now — one of a handful of global where-to-buy providers.
PriceSpider leads the space through its pace of innovation: adding physical retailers to its global network, developing shoppable reviews, and offering tools for digital shelf management.
Ongoing
A service that shows shoppers on a brand’s own site where the product can be bought, online and offline. PriceSpider has led that space since 2011.
Hundreds of requests per second from branded sites around the world, with statistics gathered from each interaction rather than sampled.
Java, Scala and Spring with Python services, RabbitMQ and Kafka, Apache Spark, Splash, MongoDB and Elasticsearch, on Docker, Kubernetes and GitLab CI.
It started as outsourcing and became outstaffing: a dedicated team of 23, three of them relocated to the Netherlands, working as part of PriceSpider.
No. An MVP first, then the analytical widget, reporting dashboards and the customisable API once the core was proven under load.
Where it helps. Three teammates moved to the Netherlands on this project to work directly alongside the client’s business.
Tell us what your traffic and data look like today. We come back with an architecture direction, a team shape and the first ninety days.
We reply within one business day.
We will get back to you within one business day with next steps.