Programmatic buying: Get with the programme

Morag Cuddeford-Jones asks whether brands are deploying data effectively enough to realise the full potential of their online ad buying.

Virgin Media
Virgin Media splits programmatic spend into retargeting ads and prospecting ads

If you believe the hype from advertising technology companies, the new silver bullet of online marketing is ‘programmatic’ buying.

As a method of buying online display ads, it claims to deliver significantly greater sales compared to conventional online ad spend. Using highly segmented data, brands can bid for ad impressions that target specific consumers at the most appropriate time, place and point in the journey towards a purchase – as well as across a variety of media platforms.

But research provided to Marketing Week by Audience Science, an enterprise technology company, suggests that anything between 50 and 80 per cent of budgets spent on programmatic buying is currently being wasted. The study surveyed major international brand advertisers at the end of 2012, finding that the majority were paying for ad impressions that missed their target audiences and were suffering poor management of the frequency with which ads are shown to the same consumers.

Josiah Amartey, senior online marketing manager at Virgin Media, believes the risk of wastage should be viewed with pragmatism: “The key is not to get caught up in the hype or the acronyms while trying to gain a solid but simple understanding of what really makes a difference to your business. In our case, we have set caps on frequency to avoid over-exposure on all types of ad spend, which is typically informed by ad server data. We also make sure we split out programmatic spend into retargeting ads [tailored for people who have used or researched a service] and prospecting ads [aimed at new consumers] to help manage our campaigns.”

But proponents of programmatic buying point out that it allows marketers to make use of their customers’ wide and varied online activity, instead of restricting ad spend to a narrow set of websites considered a fit for the brand. Programmatic buying’s benefit is to place ads where the customer is, instead of where the brand presumes them to be.

Scott Voeller, senior vice-president of brand strategy and advertising at MGM Resorts International, notes that programmatic buying is able to unlock potential in customer segments that weren’t previously interacting with the brand. “We can bump up against a database that is dormant, that hadn’t been responding to marketing elsewhere. On a different website and with an opportunity to connect through a different channel we see huge conversion rates.”

Having good audience data is key to programmatic buying. These data sets determine where ads are placed and without the necessary rigour they will show up in less relevant advertising inventory. The cost per acquisition (CPA) of a customer will increase and the return on investment dwindle.

For MGM Resorts, Voeller has identified two distinct consumer ‘buckets’ – casino and hotel customers. “We have a lot more data on the casino players because they were rewarded for loyalty through our M Life loyalty programme. We’re much more programmatic with our casino marketing efforts because we have the data, but we have begun integrating our hotel customers into the loyalty programme and we can start to apply the same principles to a broader audience.”

Sky’s real-time bidding campaign uses the company’s full product and customer databases

Supermarket Waitrose discovered the importance of having data that accurately describes a brand’s customers during an online marketing campaign it carried out with Infectious Media in the run-up to Christmas 2012. Peter Burns, manager of online marketing at, says the brand wanted to raise web conversions, average order value and customer order frequency, while also reducing CPA.

Waitrose targeted customers with product offers through October and November 2012 to encourage them to place Christmas orders early. It then used creative and messages tailored to be relevant to particular consumers, in order to maximise the resulting click-throughs.

After Christmas, infrequent shoppers were targeted with a unique offer through January and February 2013 to encourage another purchase.

“By using data to find the right audience both pre- and post-shop we improved online performance significantly during the period. Web conversions and average order value were up year on year and there was a very large reduction in online CPA,” Burns claims.

National Rail Enquiries (NRE) uses technology from PubMatic to ensure that ads against the 68 million monthly page impressions on its website are served smoothly and to their target audiences. Jonathan McCauley-Oliver, online sales manager at NRE, explains how programmatic buying improves conversion for brands advertising on the site: “If I go to a marketer and say that I can deliver your message to the one person who is likely to buy your product, that is worth a lot. If I can do that in five seconds to 8 million users, they’ll bite my hand off. This one-to-one relationship is afforded by these advancements in technology.”

Indeed, despite the widespread wastage in programmatic campaigns identified by Audience Science, some brands are enthusiastic about the possibilities. Voeller at MGM notes he is taking advantage of the ‘test-and-learn’ environment afforded by programmatic buying. Its ability to deliver campaign analytics in real time allows marketers to adjust targeting mid-campaign, to shift its focus to more effective inventory.

“We worked with Facebook, using its custom audiences and exchange function to target certain segments and through using social media we’ve been able to clean our data. We do a lot of behavioural targeting, which has broad triggers, but through Facebook it was very specific.”

Virgin Media’s Amartey suggests that brands need to resist the temptation to attempt to deploy all the data they own: “You’ll often be able to realise your business goals without needing to engage every single element of the chain or looking at every set of data.

“For programmatic buying, we reduced the amount of third-party data we use and mainly rely on our own server-level data. Identifying the right data doesn’t happen overnight and there is no silver bullet as the answer will often be unique to your business or campaign.”

One prerequisite of programmatic buying is that data not only has to be fit for purpose, but that companies have to think about their most important data segments in new ways. Clothing chain Reiss instigated a programmatic buying campaign to attract new upmarket customers to the brand through a platform provided by Turn software and bidding through agency Threepipe. Ross Loughlin, head of ecommerce at Reiss, states: “Layering third-party data over the top has given us the best results and our objective is acquiring ABC1 users.”

Using targeted third-party data, acquired through predictive and ‘lookalike’ modelling that identifies consumers similar to an intended audience, the brand incorporated time of day, geographic and frequency optimisation to deliver a 200 per cent uplift on return on investment and 63 per cent savings on the target CPA.

Data used by companies to create the programmatic buying campaign is only part of the equation. Brands need to integrate ad performance data fed back during the deployment of a campaign to refine and improve performance.

“The most important thing is the transparency of data and continual optimisation. It’s passed back and forth. I’m very ROI-oriented and that level of data and continued dialogue gives us the ability to optimise,” Loughlin adds.

But transparency often appears to be lacking in practice. Audience Science’s research found that it’s the norm for more ad impressions to be served than the advertiser’s desired frequency cap specifies, meaning that brands are spending more than two-and-a-half times what they should. The report concludes that this often happens as a result of publishers ignoring frequency caps in favour of delivering the full amount of impressions covered by the advertiser’s budget.

NRE’s McCauley-Oliver suggests that programmatic buying should be incentivised according to the client’s key performance indicators and not purely the number of clicks.

“A credit card company we work with looks for measurement based on the cost per approved application. With cookie-based data we are able to measure which ad formats across the site drive applications, but these are one step behind the final approvals that actually deliver revenue to the client. If our target CPA is £850 and we find ourselves over this, we have to use the technology within programmatic buying to adjust the targeting based on the client’s data to improve that.”

Many publishers are not quite so assiduous in prioritising advertisers’ needs, Audience Science’s research suggests. 

For all the good work brands and their agencies are doing in optimising their data for programmatic buying, this is one attitude that will need to change if it is to become as valuable a tool to marketers as the hype suggests.

For more on programmatic buying read our essential guide.

Case study

Sky upgrade campaign 

Subscription TV broadcaster BSkyB was looking to increase the number of people upgrading to its premium channels, including Sky Sports, so the company engaged Infectious Media to develop a targeted real-time bidding (RTB) campaign to deliver messages to appropriate segments within Sky’s existing customer base.

The campaign uses two distinct strands within Sky’s database: activity on the website, primarily around the shop, and the full database across the company’s whole product range. The data strands are then segmented according to whom the existing customers actually are and what products they do or don’t have. This data is then enhanced with other streams including geo-targeted, contextual, internet service provider, time of day, day of week, general behaviour, frequency and recency data.

Joel Christie, online marketing controller at Sky, reveals that preparing the data for deployment was a significant task: “It’s a lengthy process in a number of ways. The legal implications are big, as is adhering to the company’s data policy. A lot of these checkpoints exist in offline marketing, but this is a new thing for online marketing. Programmatic buying has driven more focus onto what we can do with online targeting.”

Christie notes that the company had to invest in data resources to fulfil its task, using Facebook’s ‘custom audiences’ function to match databases and then regularly check the data as a whole to make sure it kept its relevancy. The company uses Sky IQ, a consumer intelligence agency owned by BSkyB, to enhance its segmentation opportunities online.

Making sure data was clean and fit for purpose at the start of the campaign was only half the picture. The real-time element of the campaign meant that Infectious Media could collect information on how Sky customers were interacting with the ads and use the information to optimise the campaign.

Budgets were varied according to the best-performing campaign segments, and ad spend limited to optimal times of day when customers were more responsive. Ad creative was also tested to find out which messages and ad sizes drove the greatest conversion. This data was shared with Sky through a campaign dashboard that showed the brand where conversions were taking place and further segmented this data by product purchased, gender, geographic location, campaign and creative.

Christie reveals that being able to see who was transacting as a result of ads, and when, allowed Sky to manage the campaign to reduce wastage: “The global frequency cap ensures a single target will see an ad a maximum of five times regardless of which campaign partner is serving it.”

The result of the programmatic buying campaign saw Sky reduce its previous cost per acquisition by more than 80 per cent and beat the target it had set of 16 per cent.