Young & Rubicam was losing control of its IT estate - this is how it recovered

When you have offices across six continents and employ more than 180,000 people, working across physical boundaries becomes a real challenge. This was the case at marketing firm VMLY&R (a merger of VML and Young & Rubicam, by parent company WPP in 2018), with each office running different instances of the same Atlassian development tools - making it difficult for them to work together and increasing costs. That had to change.

Uhub is an internal platform that acts as an enabler for the wider business. It developed organically as the brainchild of CTO Craig Castle-Mead and Global EVP Opher Lichter. Castle-Mead - then a business systems manager - began a consolidation programme when he first recognised the issue in 2015.

"When I started [at Y&R] Australia-New Zealand, Y&R was 20 companies, 20 offices and 2,000 people, give or take. One of the companies in Melbourne had a Confluence and a Stash [instance]; then the same company, but in Sydney, had a Confluence and a Jira, or a Stash; and then some other organisation had Jira and Bitbucket, or Jira and Confluence. So there was already probably about 10 or 15 different environments that were just isolated to each of the different teams…

"We started consolidating all of these different Confluences, Stashes and Jiras into one environment and allowing teams to work independently the majority of the time, but start working together, kind of cross-borders, where that makes sense. That was where the thing that became Uhub really started: consolidating existing usage of the tools, making it more streamlined in terms of upgrades and management and servers, because we were running it like it was before the cloud products even existed. So instead of updating five different Confluences, we could do it once with Jira, Confluence, everything."

For the last five years, Uhub has been working to optimise the development operations of the various digital teams across WPP, although initially there was some friction in - of all things - the branding. Lichter explains:

"We started to realise that what we grew…was applicable and could be made available to other WPP agencies, and even before it was that there was an interest from the agency that we served to not brand our platform with one or the other of the agencies that used it.

"So if we're talking about this, for instance, Young & Rubicam and VML and Wunderman being on the platform, none of them would like it if the others' brand was used. So what we did is come up with a brand of our own, and it's a non-brand brand, called Uhub."

One of Uhub's first cross-borders projects was to simplify the HR onboarding process. Previously, when an agency in the Y&R Group wanted to bring on a new hire, the request would need to go through the central Finance and HR departments, and that person would need to be licensed across five or six different Atlassian environments. Uhub's analysis of the economies of scale showed that the pros of moving to a single platform "massively" outweighed the cons.

Choosing the right tool

The Atlassian suite consists of many different solutions, and it is not only the technical staff who use them. Only about half of Uhub's ‘customers' are in the digital community - who, Lichter says, see Atlassian as "a leading product." The other customers are advertising people "of various sorts," and one of their most common pieces of feedback is that the system is too complicated.

The Y&R Group's structure is complex, and - in keeping with Conway's Law ('Organisations that design systems are constrained to produce designs that are copies of the communication structures of those organisations') - its Atlassian deployment mimics that. "It's not really meant for simplicity, and the way in which we configured it is not simple," says Lichter.

"We regularly acknowledge that it may not work for everybody. We will do the best we can, but we're really only nine full-time people. [We often hear] 'Just make it easier' and 'We want to collaborate', but getting that balance between ease and collaboration, and then ensuring that there's security and segmentation and only the right people are seeing the right things, becomes quite complicated."

The complexity stems from the scope of Y&R's Atlassian deployment; the company uses Confluence as a knowledge base; Jira and Workflow for task and project management; Bitbucket for code management; and Bamboo for CI/CD, as well as multiple add-ons from the Atlassian Marketplace to enhance and add to the standard functionality. Castle-Mead says, "The products out of the box might do 80 per cent of the work, or 80 per cent of the usage requirements, but that last 20 per cent can make a massive difference." He also points out, though, that add-on licensing can be expensive:

"The big part really, with the Atlassian ecosystem, is if you want to get an add-on that 200 people want to use, but you have 10,000 licences, then you need to license the add-on for the full user base."

Complicating matters further is the fact that not all teams will use all of the Atlassian tools; they might pick and choose the ones that they want to use, such as Confluence and Jira but not Bitbucket, and Jenkins instead of Bamboo.

Simplifying on a grand scale

In spite of the complaints from advertising staff, the use of Atlassian has removed complexity from the development process. Since the early rollout and platform consolidation, Uhub has continued to smooth the use of Atlassian tools throughout the Y&R Group. Lichter says that much of the work has focused on advanced areas like continuous integration and continuous development, such as using Bamboo to deploy a client deliverable to production servers for immediate feedback.

According to Castle-Mead, the Uhub team enabled one of their agency teams to performs around seven deploys a day: taking content checked into Bitbucket and deploying it to an AWS S3 bucket. This takes 5-10 minutes to do manually, but "Bamboo does it in five seconds."

"So when you're talking about, on average, say 30 deploys a week: if you're saving five minutes, every single deploy, that translates into about two and a half hours, and at an average rate of, say, £150 an hour as a charge-out rate for a developer in a location like London…that starts turning into a pretty significant saving."

That's a relatively straightforward use-case, but he describes a more complex one that crosses continents:

"[The London team] are working with another team in Los Angeles; they check code into Bitbucket, and then Bamboo takes the code; runs a bunch of unit and linting tests to look at the quality of the code; takes that code, build some containers and ships the container images to AWS Elastic Container Registry (ECSECR). It then spins up some load balancers and security groups and takes the containers out of ECS ECR and...injects them into Elastic Container Register, ECR, then pulls them out of ECR into ECS - AWS' Elastic Container Service.

"Again, this whole process happens in, I think, seven minutes. And if you had someone manually doing that, that's probably an hour - if you even know how to do it."

It hasn't all been smooth sailing, and there are a few areas where new customers should be aware of some of the quirks of the Atlassian platform. One, which Lichter has fed back and the company is working on, is that all that flexibility has a trade-off.

"It's like sitting at the controls of an aeroplane. There's about 250 switches, and unless you really know what switch number 173 does, you're sort of hesitant to turn it off. Then even if you do know, sometimes the combination of those switches can come back and bite you. So what I think that Atlassian is doing more and more - and I think for organisations of less complexity it makes a lot more sense to do - is use their cloud product, where a bunch of that granular control was taken away from them… That presents another challenge, but at least it's easier."

He adds, "[B]ecause of the level of control or flexibility that large organisations typically need, you're probably going to wind up wanting to host your own [Atlassian instance], integrate it with your systems and configure it. And so…you need a team of people who are going to be experts in it. Having to hire for that…it is a challenge to meet people who are going to wind up being subject matter experts - not just in Atlassian, specifically, but in workflows and collaboration and business analysis and taking requirements and then turning them into reality. So that's a pretty unique skill set."

For Castle-Mead, the balance between governance and control can be a sticking point. "Hitting that balance between…the way that the business needs to work, and the way that the staff may want to work…is quite difficult."

Despite those issues, both Uhub founders are very happy with their Atlassian deployment. Lichter says, "The power and the flexibility within the various products has allowed us to support this model of different people at different stages of the journey wanting to do different things."

Castle-Mead sums it up rather more pithily: "With ultimate power comes ultimate responsibility."

This case study originally appeared on Delta, Computing 's market intelligence service for IT leaders.