How Graph offered a Eureka! moment for a UK SMB

If you’ve ever wandered into that bit of the O2 venue owned by Sky, chances are you indulged yourself or the kids in pretending to be a newscaster - a pretty routine job these days, as the scripts are only ever about Brexit, but still and all.

In any case, by doing so you will have interacted with a set of what the trade calls digital experiential tools, whereby your physical presence and movements are added to by video and augmented reality, as well as other content: net result being, a nice little professional-looking clip of you or your loved one channeling a news anchor.

Now, here’s something rather interesting. A small Farringdon-based tech SME called Quander helps not just deliver that experience, its whole business model is based on capturing lots of user interaction data on ‘experiences’ like this and other, producing analytics and reports on what features or content visitors liked most… but it was having a disagreement with its client about what that list actually was. CEO Gavin Williams recalls:

Sky was telling us that it wanted to turn off certain features that we were sure were actually the most popular - but we had no way to convince them beyond just showing them lots of database tables.

That conversation did eventually end up going his way - but only after what Williams described as a Eureka! type insight moment:

We have been looking for the right tools to build the new version of our platform, which was originally built in Postgres. We had a really good structure, but we wanted the flexibility to change things however we wanted in the future, and so were looking at things like using Postgres again only with third party extensions, MongoDB and MySQL. Then, about six months ago, I just randomly Googled ‘graph SQL database’ and found software that I thought looked promising.

That produced Neo4J as a possible solution. Williams says that he then loaded in a large amount of row-table data - 1.5 million rows - which included the Sky user data against a 20-line piece of SQL-equivalent graph data query language. The test produced so something so dramatic, he says, that he shouted loudly for everyone in his 8-person company to come to the meeting room:

There we had a visual representation of how everything fitted together - how it all related together. And what was really interesting that we had graphical proof that we had been right - about what visitors had been spending their time on. I made a screenshot, sent it to the client, and they agreed that we had found a way to describe the experience that was better for them than numbers, and they agreed with our proposed changes.

As a result, says Williams and his colleague and Business Director at the company, Fran Scorer, graph has become a very important part of the Quander story and a key element of that next iteration of its core engine, with Scorer telling us how,

We use graph to analyse customer demographics and engagement to ensure our customers get smarter and more effective with their event marketing over time. We have also been using [graph software] as an education tool to visualise the way that data can be interpreted with customers.

Performance gain

Seems like Quander - whose other named clients include YouTube, Samsung and Snickers - is another convert, then, of the list of companies who say graph, a way to model complex systems as a network using mathematical concepts originally laid down by maths genius Euler in the 18th century, offers them truly new ways of both working with and visualising information at scale, says Williams:

We are now starting to roll out the new core software in a Kubernetes cluster so we can move pretty easily to working with potentially billions of nodes. We also like how we can build a really good API with a graph back-end.

You can do things in a relational database management system you can do with graph, of course you can - people have been doing that for years. The opportunity comes in the performance gain you can get with graph - building a recommendations engine in graph is really easy. Another thing it gives you is that you can build something and then change it when I need to, which I’d have to do with tons of migration software and documentation of the changed data structure if I did it in a more traditional system. And finally, there’s the chance to ask your data really random questions you hadn’t thought about!

Williams says that now he is moving to a new way of exploiting data as a result of his exposure to graph, as the original platform - which was centred on data collection for clients, how many times an event was visited, how many social media messages were opened - is making way for an approach much more about being able to focus on tracking and curating individual experiences, allowing the space to ‘learn’ about the user from the moment they check-in. Concludes Scorer,

For the CTO, this is a really good time to play with this type of database as you can see what you’re doing, not just typing into a console.

My take

Though graph remains without question a minority database industry sport, use cases like this do suggest there is enough under the hype for developers to look at it when they are considering non-standard tools to try and work with complex data applications.