DSPs and marketers: in this, the second instalment of Marketing’s serial feature on automated trading, we look at what the rise of data-driven online advertising is going to mean for your brand, your department and you.
We highly recommend you read Part One (or at least the glossary of terms at the end) before delving into this instalment.
Complexity and uncertainty are looming for marketers and advertisers entering the daunting territory of automated trading. As a payoff, however, there also exists great opportunity: to take advantage of the efficiencies and targeting precision brought about by data-driven advertising; opportunity to measure, analyse and adjust campaigns on the fly to maximise ROI; and, opportunity to streamline previously disjointed business processes, minimising wastage of information, energy and money.
As Garth Agius, media director, Neo@Ogilvy, explains, one of the obvious benefits to marketers is the ability to communicate, cost-effectively, with clients past and present.
“If you can identify people who have bought your product before from cookie IDs, DSPs (demand-side platforms) make it incredibly easy to speak to that person again, with tailored creative, through retargeting. It is the next step in taking traditional CRM direct mail online.”
It is in this intersection of CRM and advertising that Julian Tol, CEO of Brandscreen, sees the chance for businesses to combine two misaligned marketing functions. “Previously, you’d have the folks in the bank or the travel company or the credit card company, and they would run their customer database… And that customer database would be the the arteries, the bolts, the crown jewels of the company. And what we had then is the separate activity in a company: advertising. Advertising grew up, from TV and print and radio, and the advertising guys would walk around and do their thing and create lots of wonderful television commercials and brand communication.”
The problem, of course, is that the CRM and advertising functions hadn’t been formally introduced. For CMOs and chief executives, the greatest opportunity in automated trading is to leverage their companies’ existing customer data to guide online advertising campaigns.
A data game
“The best data sources are normally from a client’s own assets,” agrees Agius. “But marketers often don’t realise the extent or value of the data points they have available. Every time your business speaks to a current or prospective client online it’s a source of data that can help with targeting and optimisation.”
Through a DSP, customers can then be targeted based on their online behaviour, purchase history, or any other information collected.
“You can also buy third party data sources, which can be very valuable,” he says. “For example, a car insurance advertiser can today buy data regarding people who searched for new cars on a classified site in the last 30 days. This can be broken down to the location and value of the car.”
The quality of third-party data available in Australia, however, is still somewhat patchy. It’s an area in which we lag behind other markets, but the importance of quality data sources cannot be understated for audience targeting, as it puts power in the hands of marketers. For each piece of ad space available online, the marketer can use data to make decisions on price and creative, in real time, and in doing so collect more data to continually optimise their campaign.
Clearly, this is going to put pressure on marketing departments to evolve.
Carolyn Bollaci, country head of Australia and New Zealand, MediaMind, says this will have a significant impact on a marketer’s role. “DSPs put more analytical and decision-making power into a marketer’s hands than ever before, allowing marketers to dive far deeper into campaign results. DSPs essentially give marketers the insight they need to tinker and adjust live ad campaigns – or their media buys – to optimise results.”
And so the skills required of the marketer of the future are evolving too.
“If you’re a global or even a local advertiser, you need people in there who can process data, and you need people who can come up with creative and innovative ideas,” says Paul Fisher, CEO of the Interactive Advertising Bureau (IAB) Australia.
Cue the entrance of tomorrow’s marketer: the data-loving ‘creative technologist’.
“I suspect that we’ll probably start out by embedding data specialists within teams of marketers,” Fisher predicts, “and eventually all marketers will have to have at least a ‘data awareness’.
“But the psychologist out there will tell you that you’re either analytical or you’re creative, you can’t be both, and I don’t know if that’s true, because I’m not a psychologist. But the businesses that will win in this space, or that will grow quickly, will be the ones that realise they need a combination of those skill sets, and they hire, they train, they develop, they retain, and, more importantly, they build their product offerings around that combination of analytics and creativity.”
Everything in its right place
Last month’s instalment on this topic (the first in a four-part series continuing here) provided background on the automated trading landscape here in Australia and overseas, demystifying the swathe of new acronyms that are befuddling even the most tech-savvy marketer. Reading that article, and other press on the topic, it’s easy to get the feeling that automated trading is turning the media buying industry on its head.
While that is partly true, as a matter of context it should be noted the role DSPs play in a campaign is not simply to replace existing methods of buying inventory, but to add to them.
“The role of a DSP line in a media schedule is there to complement, but it depends on the objectives of the campaign,” says Danny Bass, chief digital officer, GroupM.
For advertisers focused on direct response, buying through DSPs is attractive due to the ease of tracking. A click is much easier to measure than brand sentiment.
“If you can get 30% off your cost of a credit card acquisition, you will make that decision in a heartbeat,” says Tol. “If you have to start considering how Facebook Likes are going to factor into your brand preference decisions as a marketing person, that’s a bit harder.”
But that’s not to say there’s no place for branding, and Fisher says it’s a mistake to say otherwise. Bass agrees, explaining that in situations such as branding campaigns and product launches, DSPs play an important role.
“The DSP component, from the client’s perspective, might be the ‘always on’ component, in a similar vein to search,” Bass says. “That might be the thing that is continually on and continually optimising, but for other parts of the campaign, we absolutely need to ensure that we’ll look at traditional marketing and metrics in how you build a brand.”
As Agius points out, DSPs mainly trade in banner and pre-roll video ad units: “These ad units only account for about 30% of the campaigns we execute. Most campaigns call for more creative and integrated responses.”
For Bollaci and MediaMind, typical clients now include automotive and finance sector organisations. “These sectors shift high-value, low-churn products so it makes sense to optimise media buys based on audience demographics and live browsing behaviours. Retargeting audiences who visit their website or interact with an ad is a powerful resource and one of the real benefits of DSPs.
“Where they are less useful is when an advertiser wants to do a specific sponsorship or content buy that is better negotiated directly with a publisher,” she adds.
Learning to fly
For marketers, it’s easy to get caught up in the hype of expectation surrounding new technology, but using a DSP to buy ad space is not a turn-key solution. A certain amount of machine learning needs to take place (see breakout box).
“With many new DSP campaigns, the results are not good in the beginning,” warns Agius. “The systems often need a learning period to test different environments and build sources of data. Therefore it is important that marketing managers hold their nerve over this learning period.”
“The longer a campaign is in market, the more data and trading history will be built up and the better the results should be. Much like a paid search campaign, there are benefits of taking an ‘always on’ approach, especially if it is a direct response campaign,” Agius says.
“If you could actually lower your cost per acquisition by something like 30%, which is about what we find on average when people switch over, you’re lowering your cost of acquisition of a credit card or a flight booking by 30%… that is a game changing number,” says Tol.
Core features? Data ownership? Neutrality? Pricing model? Brand safety? These represent essential questions to ask of any prospective technology partner, going far beyond simply how many ad exchanges a DSP covers.
“The first thing every client should ask their agency or DSP partner is how they are being charged,” says Agius. “Overnight some media agencies became both media publishers and media buyers. This represents an obvious conflict of interest, with traders motivated, at least in part, in securing the most profitable inventory for their agency and not always the best inventory for the client.”
To make the task of choosing a trading partner easier, Agius says that while there are some significant differences in the functionality of the major DSPs, the main points of difference are:
- the quality of inventory available
- the accuracy of data sources, and
- the ability of the media trader using the platform.
Which raises an important point: when it comes to the most important functions, you can’t just rely on technology, and one of the most critical of those functions is brand safety.
“Protecting brands in automated buying environments is a massively important issue,” says Tol.
With the power to communicate so directly with individuals – so intimately – comes the responsibility of protecting your brand and the customer.
“The moment you are advertising to your customers and you are the Commonwealth Bank or you are the ANZ Bank and you have communication which is going to a consumer personally and precisely, just imagine if you started to do that on a porn site. Extremely humiliating, embarrassing and damaging,” cautions Tol.
“Every digital media planner dreads the thought of getting a call from a client advising their ad has appeared on a porn site,” adds Agius.
“You can’t just rely on technology when it comes to protecting brand reputation, and we invest a lot of time manually reviewing sites and adding them to our DSP network to ensure brand safety.”
With the ability to target so precisely, almost down to an individual, in an almost endless array of conceivable scenarios, another question arises: Who is responsible for producing the mountain of creative to be used in those scenarios?
Fortunately, it need not be so daunting, says Agius: “It’s possible to use dynamic creative and landing pages, where a core set of assets are developed and the creative messages are built and served in real time depending on who is viewing the message. This makes the creative and landing environment a lot more relevant to that individual user.
“DSPs aren’t a set and forget tool to automate an outcome, there is no formula to balance creative and volume of messaging. That’s where the quality of the media trader and analyst comes in,” he adds.
So the media planner’s death has been pronounced prematurely. Her skills are still relevant. Vital even. They are just changing, as are those of her colleagues. But we’ll be turning our focus to media agencies next month.
In a few short years, we won’t be talking about automated trading or real-time bidding. The specificity and savings brought by data-driven advertising are already making DSPs another tool in the belt of advertisers, one with its own strengths and weaknesses, that complements the others nicely.
As Tol puts it: “Why would you book an ad a month in advance to an audience that may or may not be there, and that you can’t identify and can’t measure accurately even if you did?… Of course you buy media in a way that is surgically accurate and in real time. Why would you not?”
Marketing Extra: Two examples of machine learning in an online advertising
Example one: Retail (pilot for a large global group shopping group)
- Starting point: The client was buying conversions from ad networks for $7.50 CPA, and had a bud- get of up to $10.00 CPA.
- Activity: The trading desk ran a trial with a budget of $100,000 over four weeks. CPA after one week was $9.00, and after two weeks $5.50.
- Result: The budget was repeated in month two, and by week eight CPA was $2.52, representing a fall in CPA of 66%.
Example two: Financial services (pilot for a new credit card from a major bank)
- Starting point: A major Australian bank was buying conversions from ad networks for $50 CPA.
- Activity: The trading desk ran a trial with a budget of $40,000 over four weeks. CPA after four weeks was $75 return on equity, but far lower for retargeting strategies.
- Result: A significant review of targeting and data management strategies was conducted and brought a steady decline in CPA, now at $32, representing a fall in CPA of 36%.