With over 15 years of experience in creating, delivering, and personalising content for top UK telecom companies, Ryan McDonnell has helped modernise the industry’s approach to content excellence.
From his early work in digital development at Three to his current role as Senior Personalisation Manager at Virgin Media O2, McDonnell has played a pivotal role in driving change and leading data-driven strategies to win over customers in a competitive and highly regulated industry.
Could you tell us a little about your background?
RM: I started out with content creation at Three—I was there for about 12 years. Then I progressed into digital delivery because I was always interested in the technical side of things, and I went on to lead the digital development function.
During this time, Three was moving across from a legacy stack onto the Adobe suite of products. And as is always the case with every transformation, it was significantly delayed. I think the move was planned to be a year and a half long, and six years into it I realised it wasn’t going to plan—it was awful.
Even though Three had been paying for the Adobe products, it wasn’t using them at all. So I was asked if I could try and get some money’s worth out of them. We did things like switch the tag manager over, which was fairly simple, and we were able to move elements of the CMS onto Adobe's instead, but not all of it.
At this time, we weren't doing any sort of personalisation at all, so we brought in Adobe Target. We played with it and had some success, so we decided that we should spin up a personalisation team. And so I moved out of digital delivery into personalisation and that’s where I've been since! I started working for Virgin around two years ago.
How confident is the telecoms industry with personalisation right now?
RM: There are quite a number of industries who are only at the very early stages of personalisation, and I would say the telecommunications industry is probably one of those as well.
While working at Three, I assumed that we were way behind the curve. Because we were reading about personalisation for years, but we weren't doing anything at all about it. And then I assumed, coming to Virgin Media and O2, that it’d be a much more technologically advanced organisation. Yet, I found myself in exactly the same situation again.
At the minute, we’re still taking tentative baby steps with personalisation [at Virgin]. But now, things are coming together more cohesively and we're starting to see the benefits—it’s like a factory line.
What do you think is the business case for personalisation?
RM: Personalisation is only ever as good as the data you feed it with. The idea behind personalisation is to take customer data and congregate it into one manageable area, then use that data to create journeys and experiences for the customer. You want to put the right offer in front of the right customer at the right time to increase your conversion rate. But equally, you may be just trying to help a customer out.
Personalisation is only ever as good as the data you feed it with
If you recognize ahead of time that a customer has a fault in their area—which can be resolved in a certain timeframe—you’re going to make their experience so much better if you put that information in front of them as soon as they land on the site. It'll improve their trust in you, and it means that they won’t get frustrated having to go searching for everything.
So if you’re a customer, for example, who’s struggling with internet latency because you’ve chosen a slower speed and have multiple high capacity devices attached to it at the same time, you might not be technical enough to realise this. So you call up thinking that you have an issue with your router or line.
If however, we intercept you when you land on the website to look for our contact details, recognise that you have the above issue due to the data that we hold, we can personalise a message to you in the Next Best Action model that tells you to upgrade your package to a suitable speed to solve your issue. In doing so we’ve just achieved three things: We’ve intercepted and deferred a contact centre call and increased digital retention. A call costs on average £6 per person. We’ve proactively given the customer a solution to their problem, increasing NPS potentially. Increasing the brand trust/loyalty – which leads to a longer length of service and thus greater ARPU. And we’ve initiated an upsell, giving us more revenue and more ARPU again.
In the past, absolutely everything was done in person, and this is still probably preferable in terms of trying to solve problems. It’s a two-way street because you can articulate more and you can get more information across. When you’re online, however, you’re trying to anticipate what someone wants and put it in front of them.
You're trying to reach as many people as possible… you're bringing content down to a conversational level again
Without personalisation, you’re offering a very broad stroke for everyone. But everyone is an individual and their problems and needs are unique. You're trying to reach as many people as possible with personalisation—you're providing a unique, tailored experience, and bringing content down to a conversational level again.
How much effort does personalisation take from your content production process?
RM: It all depends on the type of content you're talking about. For us at Virgin, we mostly focus on offer-based content.
We don’t sell a very big range of products. Our job is to get the right products, or the right solutions, in front of the right customers at the right point in their journey.
We have a template that looks largely the same each time, just with different elements of it swapped out. Then we create lots of different versions of the template that can be matched with lots of different cohorts of people.
While it takes a little bit of extra time to create extra versions of these templates, they’re versions of the same thing, so it doesn’t take that much time. And with the template already built, we can just swap content in and out using our content management system (CMS) as well.
What tools do you use to personalise your content?
RM: Our data is collected from a variety of sources and it flows into our
Google Cloud Platform (GCP). This is the single repository for all
Virgin Media and O2 data. From there, we lift the data that we need and we flow it into the customer data platform (CDP). The CDP can then serve up that data as audiences or segments, or however you want. It feeds into any platform from then onwards.
From there, we can use it to flow into the CMS, so you could push it into whatever you're using to create your content. We can also connect directly from there into platforms for paid search, paid media, and so on. We can even connect it from there as a feed directly into a data clean room, to sell data to other companies.
The good thing with the CDP and CMS is that you’re coming from the same repository. If I build an audience inside the CDP and then I send that audience to the site, we know that it’ll be the same people. Because everything is held centrally, the same audience will be sent to the social platforms, to display, to paid search, and so on. This means there's no room for human error; no room for failure. It's always going to be the same no matter what, across all of the channels.
How do you segment your content?
RM: Every customer interaction creates data, which is stored centrally in one repository where it’s processed and made available to other systems and channels.
If an agent calls at your door and you indicate that you’re an interested prospect, but the agent isn’t able to close the sale for any reason, that data will be mapped to your address and given an ID. If you then navigate to the website, for example, and search for coverage at your address, we can map that session to your ID and any previous data that’s held. So the website now knows that you’ve been spoken to previously and that you were interested in a specific product or service.
So your journey can be tailored to reflect that. This means that the conversation you had in person with the agent isn’t dead, it’s picked up and continued virtually instead.
Let’s take a fictional example—the recent iPhone 16 launch—and say that our commercial team wants to run a campaign offering preferential pricing to customers within the last 3 months of their current contract.
So we’d build an audience of active O2 customers who have IOS listed as their operating system, a contract about to expire within the next 90 days, who are not in arrears, and who have a higher propensity to upsell (0.6 and above on a scale of 0-1). This audience would then get automatically fed into our outbound communication generating system, so Customer Comms can contact the right people via SMS/Whatsapp/Email, and so on.
The same audience would also get centrally sent to the social and paid platforms. This means we can create similar advertisements that match the emails they’ve been sent. The onsite CRO team can also ingest the exact same audience and make sure there are banners available onsite that mirror the ads.
So now you have a scenario where a customer receives an email with a specific deal to upgrade to the latest Iphone. They may or may not open that email. However, they’ll go onto social media and see the same targeted ad reminding them to upgrade, with a preferential discount. If they try to search for alternative offers via Google, they’re likely to see the same ad again. And when they do finally navigate to the site—either because they want to upgrade, they want to cancel or just organically for any other reason—they’ll once again be greeted with the same offer.
So, in this sense, we have a cohesive, omni-channel approach that tackles customer retention and XSUS, all managed from the central CDP. This creates multiple touchpoints across the internet for eligible customers to see and interact with, making it very difficult to miss altogether.
In an ideal world, your data science department would have a multi-touch attribution model that links all of these together and the associated weight that each point had in getting that customer to avail of the offer.
Do you have an example of a personalisation campaign that surprised you?
RM: We do a lot of campaigns via email—for customers who are coming up to the end of their contract, customers who are out of their contract, and so on.
In an attempt to get them to recontract, we send them an offer and depending on how close they are to the end of their contract, that offer will deepen. Three months out, you'll get one deal. Two months out, it'll be a bit better, one month, and so on until you're out of contract.
These offers are sent via an email into your inbox with a link to check out straight away. This was all done solely by email previously and almost like a mop up exercise, because it was very simple to execute. We started putting banners on the website so that if you came to the site organically, for any reason, you would see the same experience that you're seeing in the email now.
What we didn't quite anticipate was the level of emails that aren't opened. The open rate for emails is through the floor—it's just not what it used to be.
The open rate for emails is through the floor—it's just not what it used to be
People don't often open emails, and even those who do often ignore them or aren’t in the mood for the offer right there and then. So the traffic that comes from email drops right the way down.
The initial hypothesis was that only a small proportion of our emails would be clicked through and result in conversions, and that it would then be a small mop up exercise to grab those leads.
But it turned out that a huge chunk of traffic that came to the site hadn't seen the emails at all! They were coming to the site for other reasons, and still seeing the banners. In the end, the recontracting and conversion rate were huge! It shocked everybody.
What do you think the lesson is from this story?
RM: What I like about this example is that it shows how you always go big with your hypothesis when you start out, and quite often you're wrong. But every once in a while, something small surprises you.
Every once in a while, something small surprises you
When you start out, you're always looking for big hitters. It's like dropping pebbles into a jar. You put the big hitters in first and it fills it up, but then you can put all the smaller ones in the spaces alongside it as well and along, like, you often overlook the small ones in favour of the big ones, but the small ones can surprise you.
There's always value in trying a hypothesis, especially when it's quick and it's easy and it doesn't require a lot of resources
There's always value in trying a hypothesis, especially when it's quick and it's easy and it doesn't require a lot of resources.
What challenges do you face with personalisation?
RM: For us, it's a constant battle with the data protection officer (DPO) and Legal to deal with data because the repercussions are so heavy if we mismanage it in any way. Anything that we do has to go through the DPO. There's a whole series of hoops that have to be jumped through protection and privacy and then through legal and everything as well.
The financial implications [of non-compliance] are so huge that we won't take any risks with it at all
It's a lengthy process and it can really delay things. In the past, I would say the longest process was always technical constraints. Now I would say that the technical stuff runs in the background while you're trying to get through all of these legal hurdles. You do get over them eventually, but it's just because the financial implications are so huge that we won't take any risks with it at all.
Have you found a way to make the legal side of things more efficient over time?
RM: We want to sell data, which sounds really bad to the layperson because it sounds like you've got Big Brother following you around. Whereas, in actual fact, you're going to see adverts no matter where you go on the Internet. Except with data monetisation, these are adverts you’re actually interested in—if you consent to it.
But every single time we sell data as analysis or whatever else, we have to ask the DPO for consent. The first time we did it, we needed to make a new request each time. But now the method has become templated, which makes it a whole lot more straightforward. It's a quick check and it’s done. It doesn't take the weeks and sometimes months that it would have taken initially.
How do you measure the success of your campaigns?
RM: Before any hypothesis goes to the sword, you have to agree in advance what the measure of success is. This could be something as simple as an A/B test, or “we want to improve the conversion rate”, or “we think this tagline on some content works better than this.”
Before any hypothesis goes to the sword, you have to agree in advance what the measure of success is
We show two different audience cohorts exactly the same content—except for the message—and then we measure the response. We look at things like attribution models for multi-touch across different platforms, like seeing ads on Meta and then on Google, and then maybe in other places as well. And then we look at how many of those touch points are hit before someone converts at the back end.
“If you don't know what success looks like, then you can't measure it”
So it really is specific to the individual campaign, but that always comes up at the start of the process. You have to agree that in advance, because if you don't know what success looks like, then you can't measure it.
.png)
