The adtech stack: a marketer’s guide to buy-side technology
Sound, top-level knowledge of the core technologies used to buy media programmatically will help marketers make sense of developments, have productive conversations and contribute to building an in-house adtech stack.
This article is an abridged version of the original 10-page feature from The Adtech Issue, our previous issue of Marketing mag produced in collaboration with TubeMogul and SpotX. For the full guide, which contains product and vendor guides for each category, statistical analysis and further reading recommendations. New issue out now »
Data management platform (DMP)
In the order of things, we’re placing the DMP at the very start of this list of components for one main reason: the DMP is the bridge that connects adtech with the other layers in the martech stack and external and internal sources of data.A DMP’s goal in life is to ingest all data at your disposal in order to paint as full a picture of the customer and their experience as possible. As the link to the rest of your martech stack, the DMP’s hunger extends to email campaigns, CRM, online behaviour and everything else. Every marketer dreams of having single, consistent experiences for members of their audience across devices and systems: email, website optimisation, programmatic advertising, SMS and the rest. Simplifying a little, the DMP is how anonymous users at a touch point turn into recognised users. DMPs should be able to provide high and low level insights, allowing exploration as well as optimisation during campaigns. Some can manage the day-to-day operation of all online advertising campaigns in one dashboard from campaign creation, audience profiling, media buying and targeting to optimisation, measurement and reporting.
So, the oft mentioned but little seen ‘segment of one’ becomes closer to a reality with a DMP. With the ability to recognise customers across devices, marketers will be better able to drive seamless customer experiences and deliver better-integrated campaigns. With a DMP, the martech stack can be the hub of enterprise intelligence, replacing the CRM at a job CRM was never designed for and connecting, if not breaking down, siloed data.
* * *
Sources of data
The more data available to use, the more effective is programmatic advertising. Marketers in some industries will have an abundant supply of first-party data – banking, retail and travel, for example – while many others won’t. But that’s where second-and third-party data comes in.
Data-as-a-service (DaaS) is a rapidly growing category with many types of organisations commercialising their data, and many that were already in the data game, such as credit rating providers and credit card companies, getting tech savvy to take advantage of the programmatic age of advertising.
First-party data is the data your organisation owns. It’s names, emails, CRM data and cookie data captured through your own website. It’s your secret sauce, the only type of data that can be your competitive advantage alone, and therefore the most valuable. “First-party data is queen, and it must be protected at all costs,” says Cameron Strachan, principal solutions consultant at Oracle Marketing Cloud, which includes Oracle’s DMP (you may know it as BlueKai). Strachan recommends marketers educate themselves about businesses that have successfully bridged the gap between first-party data and paid advertising channels.
As well as ingesting data back from advertising campaigns, a DMP should be continuously ingesting CRM data on the fly and on-demand within 24 hours, because CRM data is powerful in making strategic decisions on campaign direction and ROI.
Strachan suggests finding some quick wins. One way first-party data is used in programmatic advertising is by creating ‘look-a-like’ segments of people who have never interacted with your brand before but share characteristics with those who are customers.
Second party data is the second-best kind of data, but the rarest, and definitions vary. The important thing to remember is, while you don’t own second-party data, it’s almost as good at giving you an edge over the competition as first-party data because it’s unique data that comes directly to you from a source, such as a publisher or a company with whom you’ve established a partnership.
This is data proprietary to your partner and will cost you money to acquire – for example, in a direct deal between you the advertiser and a publisher, the publisher’s own data can be used to enrich the activity bought from that publisher.
Third party data is the most common type of data and has become an industry unto itself. It’s data that’s been aggregated from websites and offline information that data providers build into customised segments and make available through syndication. As with any data, it’s useful only if you understand how it can be used, so it’s important to understand its accuracy, how it is derived, whether deterministic or probabilistic, inferred or people-based, modelled or more. DMPs will usually have their own data partnerships, as will many demand-side platforms (DSPs). Ask what’s included and available for extra cost.
Oracle, which has DMP BlueKai as part of the Oracle Marketing Cloud, is expected to launch Oracle Data Cloud into the Australian market in 2017. It’s separate to Marketing Cloud because there are other applications for this kind of DaaS offering. Oracle Data Cloud claims to have more than 1500 data partners, such as Dun & Bradstreet, Visa, Experian, comScore, Forbes and 1495 other financial services companies, publishers, retailers, loyalty programs… you get the idea.
One new Australian data partnership is between Qantas Loyalty’s Red Planet division and mobile ad exchange APEX (an advertising marketplace that’s a joint venture between Fairfax Media and Nine). The deal combines data from the 11.5 million members of the Qantas Frequent Flyer program with APEX’s advertising inventory to enhance its value. If this data were exclusive to APEX, it may be categorised as second-party, but it’s pretty much third-party for you.
* * *
Turns a website or app into source of data and feedback loop for your campaigns and DMP. Visitor behaviour is a significant source of first-party data that can be leveraged to power and assess online advertising. Not to mention a key step in retargeting and prospecting.
Google Analytics is by far the most well-known web and mobile analytics providers and the most used platform by the marketers we interviewed for this issue of Marketing.
But that doesn’t mean other platforms don’t exist to, if not replace, enhance or extend your capabilities. In fact, Avinash Kaushik, author of Web Analytics 2.0, says trying to find a single solution to answer all your questions is, at best, a waste of time. “The quest for a single tool/source to answer all your questions will ensure that your business will end up in a ditch, and additionally ensure that your career – from the analyst to the CMO – will be short-lived,” Kaushik says.
Many organisations use Google Analytics and something else. One thing that makes Google Analytics almost impossible to break away from is its integration with Google’s AdSense and AdWords. (The theory that whether you use Google’s analytics product or not will affect SEO has never been backed by evidence.)
Many large companies use Google Analytics Premium, which adds significant layers of support, consulting services, service level agreements, data integrations and closer-to-real-time refreshing of data. It’s not cheap so is typically used by organisations that would have at least one full-time role dedicated to using it.
* * *
Tag management systems (TMS)
Tags are small pieces of code placed on a webpage to enable tracking and reporting by a third party. They’re also referred to as ‘pixels’ or ‘tracking pixels’.
A very common example: for Google Analytics to monitor a website, a unique tag is placed on the site, giving Google Analytics a portal into your website. Tags are also used for retargeting, conversion tracking, affiliate marketing, CRM systems and many other services.
Tag management tools have become vital for marketers as the number of different pieces of digital marketing software has risen from the hundreds to the tens of thousands in less than a decade.
The benefits should be improved time to market and standardisation of data, as a tag management system will know you have a primary website, a microsite, an international website, a mobile site, an app, a specialised in-store kiosk, and so on, or some other combination.
In an enterprise environment where security and compliance are paramount, a marketer can lose advantage if they have to wait to update the content management systems associated with those sites and apps. A tag management system requires the placement of just one tag – itself – on all sites once and then the marketer can use it to control any other tags they’d like to add or update.
A good tag management system should allow the user to control the deployment of all tags through an intuitive web-based interface. Point and click, ideally.
For many organisations up to a certain size, Google Tag Manager will be the default tool to use. For others, the tag management product you use may depend on your web analytics platform. Some, such as Adobe Analytics and IBM Digital Analytics, will include their own tag management system, while other web analytics platforms can integrate with third-party tag management services.
* * *
First click, last click, even allocation, starter-player-closer, adjacency, latency, pathing…
Attribution modelling is nobody’s favourite topic. But understanding the role different touch points play in a customer or non-customer’s journey is vitally important to improving marketing effectiveness, and the idea that ‘last click wins’ is difficult to defend.
“Brands are attracted to digital advertising due to its transparency and programmatic advertising efficiencies,” says Ben Sharp, managing director of AdRoll in Australia and New Zealand. “However, the sums spent on digital are now so vast that any inefficiencies have become costly.”
Attribution models can go a long way to addressing such waste, but many marketers still struggle to understand the opportunities of attribution. In fact, in the IAB ‘State of the Industry’ report into adtech usage among marketers and agencies, more respondents said they use attribution than said they were familiar with it.
“I think what this suggests is that there are lots of marketers that are probably using last-click methodology for their attribution model,” Sharp says. “But there’s still a lot of education that is needed to be able to educate marketers on the options and the benefits of using different forms of attribution modelling.”
Even when marketers want to move forward, getting crucial budget support and business case approval from the C-suite can prove daunting. When looking to solve for attribution, you may find solutions in-built in a DMP or ad server, or you may look to a standalone service.
* * *
Data visualisation through the use of reporting dashboards is possibly the closest many businesses can get to the fabled ‘single point of truth’ for marketing performance.
A dashboard is not really a single point of anything, of course, but a visualisation of one or more streams of reporting data from other systems, such as web, mobile and social analytics, CRM, ecommerce, advertising campaigns and email databases, as well as pretty much anything else with an API that can feed data out continuously or at regular intervals.
Several other pieces of technology in the adtech stack, such as DSPs, not to mention in other marketing and business software, will have in-built reporting dashboards, but their usefulness varies. Numerous third-party dashboard providers have popped up to satisfy the need to see reports from campaigns running in different systems, if different DSPs are being used for different things, like display, video, search and social.
When running programmatic campaigns, reporting dashboards are very important. All campaign metrics can be seen in one place, alongside insights on audiences. The goal is for any insights gleaned to be actionable in optimising a current campaign. There are several dashboard providers in the market and several considerations to keep in mind.
* * *
Demand-side platforms (DSP)
A DSP is the decisioning engine for buying programmatic advertising. Some DSPs specialise in certain formats (such as video) or channels (such as mobile). This means many advertisers will be utilising multiple DSPs, although the trend in recent times is for DSPs to expand to offer more and more complete offerings. A single unified platform for all media types may still be a little way off, and more media types are being readied for programmatic as we type this, so the situation will evolve for a long time to come.
Of all the DSPs available, Google’s DoubleClick Bid Manager is the only must-have. Not necessarily because it’s good, but because it’s currently the only software allowed to purchase advertising on YouTube. (Yes, the only way to buy some of Google’s best inventory is with Google’s own purchasing tool.)
But YouTube aside, marketers will need another DSP that can cover off everything else. The DSPs listed in the breakout support most major programmatic channels and devices: desktop display, desktop video, mobile web display, mobile web video, in-app display, in-app video, social and native. Depending on your needs, you should check a vendor’s capabilities carefully when procuring DSP technology. DSPs should also have decent pre-existing integrations with data provider, such as Nielsen, as well as allow the use of third-party verification tools such as Integral Ad Science and Moat.
Sources of inventory
DSPs can provide a high degree of scalability and reach almost all of the world’s top publishers through open marketplaces. Advertisers often still need the ability to connect directly with publishers to get the absolute best inventory at prices they can better forecast against. Anything that involves direct negotiation between an advertiser and a publisher is known as ‘programmatic direct’.
- Auction: RTB, public
- Programmatic direct: No
- Guaranteed impressions: No
Ad exchanges open to anyone in which instantaneous auctions are facilitated. Not many (smart) advertisers buy from open exchanges anymore without the help of a DSP, as it controls for targeting, brand safety and ad fraud. Ad private exchanges may specialise in different formats such as video or channels such as mobile
Examples include: DoubleClick Ad Exchange (Google), OpenX Ad Exchange, Microsoft Ad Exchange, ONE by AOL, AppNexus, Rubicon Project Exchange.
It’s being described as the future of programmatic video as the marketplace trends towards a hybrid of open and private. Read more »
Private marketplace (PMP)
- Auction: RTB, private
- Programmatic direct: Yes
- Guaranteed impressions: No
Invitation-only marketplaces created by one or more publishers with inventory set aside for select advertisers. PMPs have arisen so that premium publishers can manage their margins. As with open marketplaces, private exchanges may specialise in different formats such as video, or channels such as mobile.
Some will partner with data providers to make their inventory even more valuable. If an impression isn’t sold in the private marketplace, it may then be made available in an open exchange environment, although header bidding eliminates this sequential process and PMP buys can win out even over direct deals.
CPMs from private marketplaces can be more than three times higher than open markets. Private marketplace buys are around a third as a proportion of all exchange buying, and rising.
Examples: SpotX (video), APEX (mobile, joint venture between Fairfax Media and Nine).
‘Real-time guaranteed’ (RTG) or ‘private marketplace guaranteed’ (PMPG)
- Auction: RTB, private
- Programmatic direct: Yes
- Guaranteed impressions: Yes
A private marketplace deal in which an advertiser and publisher commit to a certain number of impressions that are then served in a real-time environment. This kind of arrangement is made even better when the advertiser and publisher sync their data.
It’s similar to a PMP, but it involves only one buyer and one publisher making a deal, which guarantees a total spend and audience, and therefore guaranteeing the delivery of impressions.
Automated guaranteed (AG) and automated performance (AP)
The automation of workflow in a traditional, direct deal that was previously a set of manual processes, such as RFP and trafficking. Direct integration between an advertiser’s ad server and a publisher’s ad server makes impressions available in real time.
* * *
For advertisers and publishers alike, ad servers are part of the foundation of their digital advertising operations.
In adtech, the advertiser’s ad server is the ‘buy side’ ad server. We’ve seen contradicting usage of the terms ‘first party’ and ‘third party’ when it comes to ad servers, so we won’t use them.
Typically, ad servers have had basic request-response logic and rely on being told what to do in relation to trafficking ads, but they’ve changed. Modern buy-side ad serving platforms offer capabilities to plan media, create ads, activate data, manage assets, traffic ads and evaluate results.
But as DSPs – which all incorporate some aspects of ad serving – rise in prominence, does the need for a standalone ad server disappear? Not really.
While many DSPs are now marketing themselves as one-stop shops, most advertisers will need to use more than one. Therefore, the ad server acts as a hub for creative, cross-channel management and consolidated reporting.
DSPs will encourage clients to run as much of the client’s activity through them as possible, so an independent ad server can be an important step to verify what the DSP is saying about your campaigns. Additionally, as mentioned on the previous page, ad servers can be used to automate the workflow of pre-arranged deals with publishers. By integrating an advertiser and publisher’s ad servers, the need for human intervention in trafficking and reporting can be reduced.
* * *
Note: Some DSPs will now incorporate retargeting into their capabilities, but standalone platforms exist, too.
These are the ads that follow you around the internet because they know you visited a particular website. Also referred to as ‘remarketing’.
It’s bought programmatically, but it involves buying an audience that satisfies a very strict condition: the person has visited the advertiser’s website, or even a specific page. “As a user browses the web, the sites they visit and the pages they scroll through provide clues about what products and services they’re looking to purchase,” says Ben Sharp, managing director of retargeting software vendor AdRoll in Australia and New Zealand.
“Retargeting was one of the first tools that let marketers analyse how a customer interacted with their site, identify that customer’s objective or product preferences, and use programmatic ad buying to serve them content accordingly.”
While originally billed as an effective way to win back people who abandon a full cart on an ecommerce site, retargeting has been drafted into other functions. B2B brands use retargeting to run content marketing, lead generation and lead nurturing programs. Educational organisations use it to build awareness among potential new students. Retail brands use retargeting to increase customer lifetime value with loyalty campaigns and targeting by email address. Driving sales is still the number one objective for using retargeting, but branding is gaining ground.
Most advertisers already use it (83% according to the IAB’s ‘State of the Industry Marketing & Advertising Technology Survey’, October 2016) and very few plan to decrease its use (8% according to AdRoll’s ‘State of the Industry 2016’). Retargeting is a major reason display ads are not as dead as they could be, but retargeting doesn’t only apply to display ads.
Video ads can be used to retarget, as can mobile web, mobile in-app, social, native and search. (Even email can be retargeted, according to a new offering from AdRoll, although it’s not strictly advertising and relies on capturing email addresses as well as cookies.)
One major complaint audiences make about retargeting is that it’s not as smart as it could be. Frequency can be very high and retargeting often continues even after a purchase has been made. In his manifesto for the industry to improve the experience of online advertising, Scott Cunningham, general manager of the IAB Tech Lab, says the industry must address frequency capping on retargeting. “And we must make sure a user is targeted appropriately before, but never after they make a purchase.
* * *
Viewability’s one of the six current major headaches of the adtech industry »
As investments increase in adtech stacks and programmatic media buying, the tolerance for wastage understandably decreases.
For advertising to be effective, it needs to be seen. It doesn’t sound like much to ask for, but almost half of all online ads aren’t viewable. Viewability is difficult to predict in advance – impossible even – because an ad is only ‘viewed’ when a user keeps more than half of the ad on their screen for more than one second (two seconds for video). This is one reason for the rise of larger formats, as users may scroll down before the whole page has loaded.
Keep in mind that all ads – desktop, mobile web, mobile in-app, display, video – have different ways to measure viewability. While DSPs and ad servers will report on campaign performance and take steps to optimise for viewability, the nature of this issue means it’s a problem only solved by employing independent third-party verification technology. The good news is advertisers will only ever need one viewability tracking tool.
While there will always be variance between tools, picking one as your brand’s source of truth is best practice.
The bad news is that third-party verification of viewability in walled gardens isn’t possible. This includes some of the internet’s biggest sources of advertising supply – YouTube, Facebook and Twitter – so is a significant issue. While some vendors claim to measure viewability in these platforms, the platforms themselves don’t actually allow impression-level data out, so those vendors will be reporting limited data.
* * *
With human and non-human traffic fraud costing advertisers around the world billions of dollars (US$7.2 billion to bots alone in 2016) this is another front in the battle to reduce wastage in digital ad spend. Fraud detection is closely linked to viewability – as well as brand safety – because tools often offer monitoring both or more factors to give comprehensive ratings of ‘quality’ for impressions and campaigns.
DoubleVerify and Integral Ad Science offer competing solutions, both providing data on the quality (or ‘authenticity’) of activity by tracking viewability, ad fraud and brand safety.
Tools like these show advertisers how many of their ads actually had a chance to be seen by a real person. It’s a simple thing, but incredibly complex to ensure in the world of online advertising. Video advertising is most at risk. It attracts high CPMs and is therefore a more lucrative target for fraudsters. If you’re paying CPMs on video inventory higher than around $15, the prevalence of bot fraud almost triples compared to CPMs under $15 (White Ops and ANA ‘Bot Base- line Report’ 2015).
Unfortunately ad fraud will never be stamped out entirely, but it can be reduced. Your DSPs and/or ad server should have an integration with one or more independent fraud and viewability tracking tool. If they don’t, watch out. So far, one DSP has created a refund scheme. TubeMogul will now automatically refund its clients for fraudulent impressions above a threshold of around 3%.
* * * * *
Subscribe to Marketing mag, visit our online store »
* * * * *
Image copyright: retrorocket / 123RF Stock Photo