Innometrics is a pioneering start-up, started in 2010 and currently operating in Stockholm and London with now around 40 employees.
A fantastic example of a digital first company that is leveraging the explosion of new diverse sources of data to deliver innovative and exciting solutions to organisations such as 3 Mobile, Tele2, Dansk Supermarked. Helping businesses to leverage data for true business advantage and meet the increasing customer expectations for personalisation, Innometrics is at the forefront of the Big Data revolution.
Innometrics is helping customers create comprehensive customer profiles based on information from all their data sources, such as CRM data, behavioral data from web and apps, financial transaction systems, etc. Taking it a step further they then help them use those profiles to enhance and personalize their customer experience.
For example Dansk Supermarked are using their technology to build customer profiles across its many new technologies, through both digital and offline channels with a focus on customer journeys initially looking at pre-purchase customer behaviour on their main website. Using Profile Cloud their key solution with a limited data set, Dansk Supermarked is now personalising web content based on both demographic and behavioural data, seeing conversions increase by over 300%.
Jens Agerberg, CTO, sat down with us to give us some insight into the exciting activities at Innometrics as well as how Riak KV is integrated into their solution.
To support these often business critical projects we heavily rely on Riak KV and integrated it into the core of our technology stack. We are creating profiles in Riak KV based on data points coming from web and mobile clients. Riak KV is ideal as it simply and easily supports a very high write throughput combined with very low-latency reads. Storing many types of data from session to location, our ability to quickly access and analyze that data is critical in the services that we are supplying our customers.
We are currently running a cluster of 5 memory-intensive nodes (64GB) with 2TB disks. And are storing around 200M objects in Riak KV, growing rapidly, something like 10% a month. While we are actively protecting the database against spikes we can reach thousands of requests per second regularly which Riak KV handles well.
We asked Jens how they came to Riak KV
As we were building an entirely new solution we were able to start with a clear sheet of paper and really define our needs from scratch. Reviewing what was important to us came down to a few key factors including:
- High availability
- Easy to scale
- Low latency both on average and low variance
- Supporting high random write-throughput
- Economical when it comes to machine spec required (particularly in terms of memory consumption)
The biggest part of the evaluation was to gain knowledge about different alternatives and evaluate them on the criteria we had. Some benchmarks were also run in the final stage with a few databases.
The core reason for going with Basho and Riak KV was because of high availability at low cost and the ability to scale infinitely as the business grows.
What is in the future?
So far most of our customers are situated in Europe so MDC has not been deployed; however we are planning to use MDC replication to facilitate some important operations work.
What is the one piece of advice you’d give anyone reading this blog and evaluating their database requirements?
Don’t trust benchmarks 😊 Make an in-depth analysis of each database considered (or get help to do that if knowledge is not in-house) in terms not only of features but understanding the storage backend, the consistency model and so on. If requirements are unclear/flexible or scale is small consider going with more traditional databases such as SQL variants but for the use cases where the tradeoffs make sense and the volumes of data are large and important, Riak is a very good choice.