Programmatic ad buying has been growing year on year, despite being plagued with scandals.
Whether its click fraud, brand safety, data privacy, or viewability it seems that every time programmatic is mentioned that it's because it is causing some major headache for marketers.
Yet programmatic has remained a very popular ad-buying strategy. The market has been growing at an average rate of 71% per year for the past five years and is projected to be a $64bn dollar industry in 2018.
At a recent Econsultancy event in Singapore, Digital Outlook 2017 Part 2 hosted by NTUC, Hari Shankar, MD at Escelis (the performance marketing brand of Havas Group) told attendees about a few of the current issues programmatic is facing and what marketers can do about them.
Two of these are summarised as follows.
The 'machine problem'
Hari started by introducing the programmatic landscape, which can seem overwhelming and confusing.
But when programmatic is simplified down to how ads are served to web browsers, it is relatively straightforward. Publishers announce that they have an ad space to be filled and advertisers bid for the spot. The winning bid then serves the ad to the browser.
The 'machine problem' occurs due to how the bidding algorithms work. Instead of offering a free bidding market, programmatic buying auctions are held using a waterfall model. What this means is that while all advertisers can bid for the space, some bids take priority, even if they are not the highest bidder.
In the case illustrated below, Partner #1 loses the bid because it is below the floor price set for the publisher, but Publisher #2, being above the floor price, wins the bid despite higher offers coming in from Partners #3 and #4.
Hari noted that this model creates an unfair advantage for large bidders, such as Google DoubleClick Bid Manager, as they are often at the top of the 'waterfall'. Publishers suffer too as they do not receive the highest bid for their ad space.
Recently, however, a new technology called 'header bidding' has emerged as a solution to the 'machine problem'. Instead of using the waterfall model, header bidding takes in bids from all ad partners at the same time and serves up the ad with the highest bid.
While some estimate that many large publishers, including MailOnline and CNN International, are implementing header bidding, not everyone is yet on board with the new technology.
Regardless, marketers were advised to ensure that header bidding is on their programmatic roadmap and to lobby publishers to adopt the technology soon. Otherwise, brands will lose out on bids for quality ad space and will not be able to get their ads in front of the people they are targeting through their partner sites.
The 'people' problem
Another programmatic problem Hari brought up has more to do with people than algorithms.
He noted that a lot of programmatic is still bought via agencies and trading desks which operate using an 'undisclosed' model. What this means is that many brands are buying ads programmatically through a third party which doesn't disclose the fees they are paying for their services.
This might not be a problem if the fees were low enough to be negligible, but Hari drew attention to some recent data about the costs of programmatic ad buying.
Using data from the Association of National Advertisers, et al, Hari stated that for 3.9 billion display ads, the volume-weighted average cost was $3.30 per thousand impressions (CPM). Disclosed fees from agencies, however added on a further $1.49 per CPM, a 45% premium.
The researchers concede that many of these costs may be worthwhile as they encompass such services as sophisticated targeting and buying against KPIs, but Hari's point was that if disclosed fees saw a 45% premium then undisclosed fees are likely to be much higher.
The 'people problem' then leads to brands cutting back on spending by buying low-quality ad inventory which then results in issues with viewability, click fraud, and the brand safety problems currently plaguing programmatic buying.
Hari concluded by pointing out that if programmatic is going to succeed as a performance-based ad buying strategy, then brands need to take more control of the process.
He makes three recommendations:
1) Set up a transparent model for programmatic buying
This could be one of three things:
- A fully-in house programmatic buying team, or
- Direct deals with DSP / RTB vendors, or
- Disclosed trading desk model with agency
2) Set up a strong tracking and optimization system
Hari advises that brands should use a single vendor for ad serving and web analytics and track display using multi-channel attribution. Additionally, marketers should integrate their customer data (CRM) with programmatic, ideally using a data management platform (DMP).
3) Test, learn, scale
Marketers should use tools like Optimize 360 for A/B testing and, using their transparent buying model, conduct more controlled buying such as programmatic direct, placement-led programmatic, and first-party tagged targeting.
A word of thanks
Econsultancy would like to thank Hari Shankar, MD Asia Ecselis, Havas Media for his presentation as well as the delegates who took time out of their busy schedules to attend.
We hope to see you all at future Singapore Econsultancy events!