Markus Malti, CEO (left) and Dr. Steffen Wachenfeld, CPO of WeQ, look at some of the myths around ad tech, offering their views on the technologies and trends that deserve the hype, and those that don’t.
There is no denying that mobile marketing is becoming an increasingly complex affair. Some of the most talked-about topics – including data privacy, the competitive global marketplace, fraud issues and quality of traffic – remain largely confusing and concerning across the industry. To circumvent these complexities, companies are turning to technology, hoping for quick-win solutions. Crucially, while there is a huge shift towards adopting new technologies to simplify and improve processes, few understand how they can be applied in practice to their advantage.
Some technologies are particularly hyped: think of the more exciting buzzwords like AI and VR. With all the talk about these new great advancements, there is still a wide gap between the theoretical and practical that needs to be bridged. Although advertisers see the value of these technologies, it remains difficult to distinguish exactly which ones can add value and which ones don’t, at this point in time.
To further exacerbate the current state of affairs, there are many “myths” circulating about how these technologies can be applied, that result in more confusion. As we work with mobile advertisers globally and are aware of their biggest concerns, we’ve identified some of the biggest misconceptions about hyped technologies that, although they sound great on paper, haven’t yet produced much value and are often misunderstood. These include myths around segmentation, automation, APIs and machine learning. So we asked ourselves: which of these technologies are actually adding tangible value and generating real results?
Myth 1: Hyperlocal Targeting Technology is Ready for Advertisers
Targeting is a necessity in today's mobile advertising industry. Leveraging parameters such as an operating system, geolocation or demographics and more, mobile advertisers are able to deliver the right ad to the right user at the right time.
This has given rise to the buzzwords “hyperlocal targeting,” also known as geotargeting with very high accuracy or a “segment of one,” meaning a single-person target group. These imply a scenario where it is possible to advertise to a group of people in a very specific place, or to an individual mobile user. For example, reaching someone who is in front of a particular physical store and offering them a coupon in real-time. Sounds amazing, right? But it isn’t quite like that. The reality looks a little different, since almost no one is using targeting that can be accurately called “hyperlocal” – and the reasons for that are simple.
First, it is very difficult and costly to obtain location information that is accurate enough. In fact, the standard way of inferring the location based on the IP address can be inaccurate by as much as 100 miles. Second, if an advertiser has access to this information, they need to have the technology in place to make optimal decisions on what to advertise, based on the additional location information. If both information and technology are there, the observations that can be made per location are too few in order to derive the required level of statistical significance.
So, while hyperlocal targeting is possible, it’s not entirely accurate today. While there are some limited technologies that offer this option, it's not something that is as readily available as it appears.
Marketing is a game of numbers, where computers repeat the same action, like showing a million impressions of a certain ad and using machine learning to predict the outcomes. The more parameters that are taken into account for the decision-making – such as the operating system, time of day, day of week, etc. – the smaller the number of observations for each combination of parameters. Adding hyperlocal information causes the number of observations to shrink, such that predictions start lacking statistical grounds. That’s why it is used only in special cases, e.g. where 100,000people are in a stadium and listening to a certain artist at a certain time, but even this cannot be really considered as hyperlocal targeting.
Technically, most advertisers can do hyperlocal targeting, and spend a smaller budget manually in this manner, but if a technology provider claims that they can do it on a larger scale in an automated fashion, be skeptical.
Myth 2: Automation is Not Worth Investing in Right Now
Many campaign management tasks can be automated, yet account managers are still setting up thousands of advertising campaigns manually, including checking the cryptic tracking links. When optimizing campaigns, setting up different parameters for A/B testing is tedious.
With automated workflows, multiple A/B test scenarios can be created in parallel and modified over time. Machine learning methods allow for automatically assigning more budget/traffic to better performing settings, thus automatically eliminating non-performing settings over time and ultimately streamlining processes.
Dynamic creative optimization (DCO) uses automation for testing dozens or hundreds of versions of creatives and banners in programmatic advertising. Things like automated A/B tests and automated link checks have become the serial letter function of advertising. While not many advertisers would replace their skilled designer, they would definitely benefit from having a system that combines dozens of call-to-actions with different logo versions on multiple backgrounds in an automated A/B test, to see which ones perform best, on which traffic and at which times.
Nevertheless, it is surprising how many companies haven’t considered the possibility of adopting automation technology yet – despite having the option – and still do everything manually. Companies that use automation scale better and produce less errors because their account managers can focus on actually managing and optimizing campaigns. As reported by Forbes, automation will save employees a whopping six weeks of time per year, and a full nine weeks of time for business leaders, which could all be invested in more rewarding tasks. Automation is here now and it's here to stay. The sooner you adopt it, the sooner you can start reaping its rewards.
Myth 3: A Full Stack of APIs Will Solve All Your Needs
One of the first questions a publisher will ask a potential provider is: “Do you have an API?” as API integrations have become the industry standard for technology partners to work with content publishers. While there is a perception that “the API” can perform on every level, there are in in fact multiple APIs: to be deemed as “well connected” in the industry, industry players typically have dozens of APIs. They are all very different.
Additionally, there are administrative APIs, such as those for third-party CRM systems for billing and finance tools. It’s evident that the sheer amount and variety of APIs can be confusing. As a response to the question “Do you have an API?” platforms shouldn’t offer a simple “yes” or “no” answer, but rather a well-informed conversation of which APIs are available and for which purpose. This is a good starting point to understand which ones can be employed to serve your end goal, and ensure that your partner has the solutions you need for your business.
Here are some examples:
Multiple dozens of APIs, that connect to the campaign management systems and allow for automated import and updating of campaigns. These API connectors are self-written, including the internal logic, e.g. in which cases manually agreed payout should prevail over updated values coming in via API, or vice versa.
A feed API to the publishers, that propagates the available offers to all publishers and constantly provides up-to-date information about payouts and remaining budgets. The challenge here is to provide the internal tools that allow precise control over which publishers or traffic sources down to the sub-IDs are allowed and on which campaigns.
Multiple API connections to external third party data-providers, starting with services that allow accurate geo-location services, over services that provide meta-information about mobile apps, and ending with services that allow to check the quality of tracking links or traffic sources.
Real-time APIs, often called native API or dynamic API, which allow mediation frameworks, in-app SDKs, or other forms of native traffic sources to directly connect to the ad network. This enables the ad network to do the work of recommending campaigns for given users from a specific source and at a specific time. These are the most advanced APIs, which require a sophisticated machine learning capability, in order to guarantee optimal recommendations.
Myth 4: Compliance is Impossible
The advertising ecosystem is complex and often lacks transparency. Ad tech players and technology providers such as Google, Facebook and Amazon, are investing in technology and legal expertise to increase transparency and control of data. With GDPR – the legal change that is heavily re-shaping advertising in Europe – there is a need for technical solutions that manage data and user preferences correctly, or to exchange data in a compliant way.
Besides laws and regulations, there is a general call for compliance from industry players on all fronts: advertisers want to know if the users they pay for are real and if their brand ads are visible; media buyers want to make sure they are not buying BOT traffic and RTB auctions are fair; and publishers want to make sure that no inappropriate ads are shown.
For many years, industry players have invested a lot of effort to control traffic, incentivize proper behaviour, and build tools that allow for monitoring and reporting. Currently, the existing tools are rewrapped and promoted as anti-fraud solutions, but what is under the hood is what’s important – and here lies an opportunity for machine learning to solve these challenges.
Specifically, machine learning can:
Ensure that users are real users with post-install KPIs, such as retention and in-app activity. These KPIs are monitored and publishers are committed to certain retention and activity levels, or even paid based on post-install events, such as in-app purchases.
Filter out BOT traffic. Many industry players have built strong intellectual property in the form of refined detection methods. In sophisticated cases, an automated decision is made based on a cascade of analyses, such as to let users pass or not.
Ensure RTB auctions are fair. This is a huge problem and the ultimate solution is to have code-audits of DSPs and Exchanges.
Machine learning solves many issues around compliance by being able to observe over time, learn and decide based on statistical distributions. Renowned distributions include click-to-install-times (CTIT), where too long is unnatural and too short is impossible; device distributions, where especially high concentration of certain devices is unnatural; and over time distributions of impressions to clicks from the same IPs and user agents. This results in more streamlined checks and balances for meeting compliance and managing security of private user data.
Despite the “hype” about certain advertising technologies, segmentation, automation, APIs, machine learning and many more can actually be of real assistance and value to mobile advertisers globally, especially when managing hundreds or thousands of campaigns. We are firm believers in the power of applying automated technologies and machine learning algorithms to simplify and improve the user acquisition journey in mobile advertising, particularly as advertisers face fierce competition within the walled gardens of Facebook, Google and Amazon, and so must now manage hundreds of other supply channels in order to efficiently acquire high quality users.
It’s just as important to keep your finger on the pulse of the latest ad tech developments, as it is to decode and analyze which technologies can truly benefit your business, before investing time and effort into anything that wouldn’t yield concrete results. Take educated risks. Work with trusted expert partners, who have the cumulative experience of working with hundreds of publishers, because this can help mobile advertisers understand how to leverage technologies to their advantage and weed out the hyped “myths” from the tangible, actionable solutions. In the end, the facts should speak for themselves in the form of successful results.
This article first appeared in the September 2018 print edition of Mobile Marketing. You can read the whole issue here.