Asia-Pacific comprise many different countries and markets, resulting in multiple data inputs and a fragmented publisher community. This has made it challenging to gather data and build audience segments relevant for marketers in the region.
In this week’s Q&A, Eyeota’s CEO and co-founder Kevin Tan and Skimlinks’ head of audiences Ed Thomas discuss challenges the industry still face in establishing accurate user profiles and managing misconceived mindsets that audience data should be provided free with inventory buys.
The two ad tech vendors signed a data partnership agreement in December 2016. Skimlinks boasts a purchase intent database of more than 75 million users in the region, with the majority located in Australia, India, Singapore, Malaysia, the Philippines, and Hong Kong.
ExchangeWire: Discuss the biggest change in how data is collected, and user profile built, over the last three to five years.
Kevin Tan: One of the biggest changes is in the types of data that are collected. There has been a maturity and sophistication of what and how data is collected. Going back three to five years, there was a lot of behavioural data and data inferred based on the content of the pages. That data, however, was not accurate. For example, if I’m looking at a page about horses, that is perceived to mean I have intentions to buy a horse, which is not accurate at all.
Data quality has significantly changed over the years, but so has data privacy. In the past, many companies were collecting data from a lot of sources without getting approvals or consent from the end-users to either collect or use their data. Much of that is changing with new and stricter legislation, as well as broader consumer awareness of data rights and data usage.
Ed Thomas: The most meaningful change is the move from data mostly inferred from opaque seed data, towards declared data which original buyers are known. It came as a result of a lot of data asset struggling to beat random performance – let alone justify its cost.
How much improvement have we made, in terms of audience data accuracy, over this period?
Tan: Accuracy is increasing because data is being attributed to more devices with cross-device matching. We’re seeing accuracy based on where the data is collected and how it’s being attributed. If you look at signals, for example, coming from a company like Skimlinks in terms of intent data, they’re accurate because they’re coming from exactly the places where you’re seeing the intent or the intent to purchase. In comparison, in the past, we’ve seen the data inferred on a specific type of content on a page that people are looking at to buy that product, which may or not may be correct.
The use of VPN (virtual private network) is common in some Asia-Pacific markets, particularly China, and this masks the real location of the user. How do you address this to ensure you build an accurate user profile?
Thomas: We see many from our global footprint of 1.4 billion user identifiers visit multiple countries as they roam freely, including people going on family vacations, business trips, or switching countries, through VPN tunnels. These are patterns Skimlinks is accustomed to understanding through our publisher monetisation technology.
Invalid clicks pose a threat to our publisher and retailer customers all over the world; and understanding swarms of synthetic behaviour is an important part of our service. When resolving a user’s country, we look at their IP history and work out the most likely ‘primary’ country for that user. For example, a user who is 100% in the US on a residential IP address is going to be classified as in the US. However, a user who is 98% in the US from a non-residential IP, but 2% on Chinese residential IP would be given a home country of China.
Once our profiles are delivered to an advertiser, they then can use their choice to geotarget on the DSP.
What remains a challenge in data collection and audience segmentation?
Tan: The challenge in data collection across Asia-Pacific is that publishers are fragmented. There are so many different countries and markets, which means you have a lot of different inputs. The main challenge for data worldwide is that data is local. In order to actually collect, understand, and segment out the audiences specifically, you need to look at it on a local basis. You need to go on a country-by-country basis.
Growing data in Asia-Pacific will require you to build out data from local publishers and local sources in each market, and then apply an understanding of what the consumer behaviour is in each of those markets to create segments that are market-specific. You can’t take a US-based segmentation framework and apply it to local markets in other countries, or in Asia, because consumer behaviour is very different and the data becomes irrelevant.
You need to make sure you have collected data on a local basis, from local data sources, and that you have a local data understanding.
Thomas: The tradeoff between scale and accuracy is often a challenge. Occasionally, an advertiser may define a target audience so small, it would be more cost-effective to take a taxi to each consumer’s home with a hand-delivered letter than target them with digital advertising. Alternatively, others define audience sets that are so broad they include most of the population; at which point the cost of data becomes inefficient.
We encourage advertisers to think about their target audience as individuals – imagine a child drawing a picture of their life. Where do they live, how do they live, what brands matter to them, with whom do they share their lives, and what is the need the advertiser wants to fulfil? Answering such questions will clear the fog and it will become obvious your ideal target audience, for instance, is not ‘Male 30-55 with above-average disposable income’. Suddenly, it’s ‘a guy with at least two children, who shops in the top quartile for price-point, but always shops around, wishes they had more space at home, is the main buyer of electronics for the household, and is anxious about retirement income’.
Where do you see opportunities, as well as challenges, in Asia-Pacific in terms of building user profiles?
Tan: Asia-Pacific has been a challenging market in terms of data. Historically, there hasn’t been much data available, so Eyeota have been focused on building out audience data across Asia and working with publishers to monetise their data. The opportunities we’re seeing are from the changing awareness among publishers in the region that their data and audiences are valuable assets. By providing their data into the digital ecosystem, publishers are gaining a new source of revenue. In the past, there have been a lot of fears that if they contribute data to the ecosystem, it would cannibalise their existing inventory sales. With the maturity of the Asia-Pacific market, this is not the case – it’s a second and very distinct source of revenue.
Thomas: Our taxonomy of audience segments are oriented around brands, product categories, and retailer categories. The wonderful thing about these taxonomies is that they form a vocabulary with global meaning. Headphones are headphones everywhere in the world and a luxury French brand has the same fundamental meaning globally. Our language of purchase intent is universal, so understanding how shoppers interact with these brands is pretty straightforward for us. Price-point and affordability are much more variable by country so, of course, we take that into account when deciding how much a shopper is likely to spend on a particular item.
How does the region’s high mobile adoption and penetration impact the accuracy and ability to build comprehensive user profiles?
Tan: This is a good thing. Mobile data is tied to distinct user IDs and device IDs. High mobile adoption allows us to collect more data and more accurately attribute it to users. With the growth of cross-device capabilities across the region, we’re seeing the ability to go across both mobile and the cookie environment to get a much more accurate and much more comprehensive profile across the region.
Thomas: Almost 60% of our user profiles worldwide are associated with mobile devices. At 54% and 60%, respectively, our ‘home’ markets of the US and UK are about average, while countries like Indonesia are closer to 75% mobile. Meanwhile, countries like France at 33%, and Germany at 43%, lag behind.
With the anticipated adoption of IoT (Internet of Things) devices, what kind of dimension will this add or complicate data collection and the accuracy of user profiles?
Tan: IoT is just another source of data. Any device that is putting out data is providing additional information into our centralised user profile.
IoT also can generate a lot more ‘noise’ or ‘unclean’ data, which can make filtering more tedious. How can this be resolved?
Eyeota: I don’t think it’s generating ‘unclean’ data – just more data. The challenge is attributing data back from any source, IoT or otherwise, to a distinct user or individual against which you can target. The challenge here is going to be making sure we can connect those pieces.
Where do you see the emergence of machine learning, cognitive, and big data analytics in the IT realm being applied here?
Tan: These technologies already are being applied. Ultimately, we collect a ton of data. At Eyeota, we have three billion profiles globally and growing. When you have that much data, that you’re collecting on a constant basis, you have a lot of signals. Applying machine learning or AI (artificial intelligence) or any kind of cognitive learning across the top just allows you to mine that data and go even deeper. That’s a core facet of the way we create profiles and the way we look at audience data as a whole. The challenge here, though, is making sure you have enough data to apply those signals. If you try to apply any of those tools across a small dataset, it’s not going to be very interesting. You really need a large dataset to be able to do more.
What misconceptions do Asia-Pacific brands have about data and audience profile, and publishers about the content they own?
Tan: In the past, an advertiser would target a website or a newspaper or TV channel, based on the percentage of that property that fits into a certain audience segment. As programmatic and audience data-targeting capabilities have emerged, a lot of those traditional advertisers still get confused between inventory and data and think those are just two cost-line items. Ultimately, though, they are the same thing.
Some advertisers, particularly ones with traditional mindsets, think they’re buying inventory and the audience should come free. They don’t realise that the two things are decoupled and that by applying a separate audience buy on top of the inventory, they are exponentially increasing their efficiency.
This distinction and awareness have not emerged as quickly as we would like, but we’re starting to see some positive results. A lot of traditional brands and advertisers are caught up in CPCs and that’s an old way of thinking. For a branding campaign, reaching a specific audience is the objective and they need to look at the metrics as the combined data price, which is the data plus the inventory price. The ROI (returns on investment) on that is significant, as long as you’re not looking at it on a CPC basis, which is not the right metric.
If you’re looking at actual brand metrics – namely, awareness, conversion, and intention to purchase – and apply those metrics, then the targeting is very efficient. If you use CPC, it’s irrelevant. You can have shiny ads and people will click on it, but it doesn’t mean the ad was successful.
Thomas: Whilst true back in 2015, there now is a misconception about the scalability and value of targeting segments. As impression or click fraud is eliminated, the unit cost of media has increased, making targeting accuracy more important. In the past, banner media in this region tended to be too cheap to worry about wastage. This is no longer the case for leading advertisers.
Discuss 2017 trends you expect to see in Asia-Pacific that will drive the adoption of data-driven marketing.
Tan: First, the programmatic ecosystem will mature. We’re seeing more and more agencies and advertisers increasingly buying ads programmatically, including online, digital, and mobile. This has made it easier to adopt and deploy audience data. The growth and maturity of programmatic adoption in the region has been exciting.
Second, the availability of cross-device matching tools will increase across the region, which allows the interconnection between mobile, display, and video. You see companies making big entries in the market and it’s exciting to have those capabilities in the region.
Thomas: As more targeting data and analytics facts are made available to Asia-Pacific advertisers, we can expect to see programmatic trading continue its journey to full adoption. Right now, media is combined with data under the legacy ad network model. Allowing agencies and buyers to combine data and media on their own terms powers a dramatic shift towards programmatic trading, and the opportunity for greater campaign returns.