The Off-Site Plan: Segmenting Audiences For Extension

Minimal Effort for Serious Revenue Gains

First-party data is oil of the digital media landscape. While all kinds of publishers can claim it, the refinement process, not the drilling, makes the big bucks. Fortunately, technology has improved publishers’ refining abilities, not just in driving revenue from first-party data through on-site advertising as well as selling data to third parties and applying them to CRM and email marketing.

In addition, tech has vastly improved publisher content and audience extension efforts, so much so that CPMs and performance are comparable (in some cases better) to on-site advertising. According to a recent survey by Digiday and Audience Accelerator, Rocket Fuel’s business unit for ad sellers, 77% of publishers say they are “very familiar” with audience extension; 52% are offering the service to their advertisers, with 33% adding it to most or all of their campaign proposals.

There are no universal guidelines to segmenting audiences or approaches to targeting – even best practices vary from company to company. The majority of segmenting is done with on-site inventory top of mind, but having an offsite segmentation strategy can boost extension performance with minimal effort on the publisher’s part. The tools are available to reach advertiser audiences and “model audiences to scale” (lookalike and act-alike audiences blended with historical and engagement data to build a behavioral blueprint) across devices, formats (e.g., long-form video), channels (e.g., social inventory) and more.

Segmenting for extension can come in quite handy for reaching even more specific subsets within niches. One of the biggest hassles of targeting inventory onsite is that you can only get so granular before running out of impressions. Finding women in their 40s with teenage children, large SUVs and an interest in fly-fishing may yield scant impressions on your own properties. But there’s a great wide Internet out there, traversed by exchanges that can put you in front of your elusive target – in a brand-safe environment, if you use the right tools.

“With extension there are larger opportunities from an impression standpoint so we can allow extra granularity in the segment definitions,” one publisher commented.

Building the Plan

The first commandment for publisher revenue efforts: know thy audience. Positioning the uniqueness of your audience is essential to the extension value proposition – why a client should trust your audience and audience models rather than go hunting down audiences on the exchanges themselves. This includes making sure your data ship is plugged from leaks and being careful about just what data clients and service providers take away during campaigns.

My sources unanimously reported that on-site and off-site segmentation are planned together, but who is involved differs from organization to organization, seemingly based on structure. A large, data-based publisher has a centralized strategy desk that aids its numerous local properties while another said that segmentation strategy is developed in collaboration between product, sales and ops. One other mentioned a dialog between the sales and account management team, the audience platform campaign managers, and the client or agency – sounds like quite the full conference room.

However, the typical procedure for determining off-site segmentation started with on-site audience segments, which were then customized based on client KPIs and further optimized to the performance of the campaign. One publisher I spoke to actually uses an “audience guide” that details the extent of its scale and its prefabricated first-party audience segments for on- an off-site targeting. The guide is intended to simply open the conversation to building customized segments: “We use these pre-defined segments as a starting point and look to our data science to help identify and refine the most responsive consumers,” the publisher said.

So what data goes in the pot? It’s more like what data doesn’t – browsing interests, visit frequency, time of day, search data, types of device used, mobile carriers, purchase behavior for e-tail sites and even social and engagement data. Registration data is also an option for some publishers, though its use is typically limited to model-building to avoid privacy issues.

In that case, third-party data is employed to beef up the segment, but such data can also come in handy when building audience models and hunting down similar users in contexts suggested by the client. In addition, publishers reported that pixel data from the client site can be quite beneficial, and that this data should be circulated into the campaign for optimization. DMPs offer the ability to accept siloed advertiser data, as well as syndicate and facilitate data sales. 

“Flexibility is the friend of optimization,” one publisher suggested. “The advertiser may know his audience, but we know how to use our data to reach that audience.”

At same time, another publisher said it will work with the advertiser to build audiences in its DMP, but the offer is rarely taken up. The involvement of the client will vary based on how audience-savvy it is or the level of trust in the pub-client relationship.

Finally, careful thought should be given to campaign context. No fan of the “spray and pray” school of chasing cookies across the exchanges, one publisher cited compiling a network of qualified sites that enables the company to take subject matter into account during extension campaigns. However, this does cut into available inventory and may have to be balanced with the granularity of the targeting.  

Execution + Optimization

As AdMonsters has detailed before, there’s many a way to execute extension – with the help of a DSP or SSP, through a DMP or, for the DIY-segment, via seats on the exchanges. There’s no right setup, particularly since combinations of the above can prove fruitful (e.g., a certain DSP may perform better on a certain segment or provide a bountiful complement to a DMP). At the same time, one publisher cautioned against “cherry-picking” based on platforms or inventory sources, as it can dilute the true value in favor of piecing value together.

Using a full-service solution allows for straightforward campaign setup – “equivalent to setting up an exclusive ad in the ad server,” one publisher said. The hard work then occurs on the service provider side – pubs can lean on their DMPs and DSPs for active managing and monitoring, continuous optimization and the application of machine learning for tasks such as striking off users that have already converted.

Full-service solutions enable ops heads to put extension workflow into the normal routine of their ops team rather than rely on specialists, though some organizations still prefer an “extension manager.” Especially for data-heavy publishers, full-service providers can drive revenue with minimal effort.

Certain publishers have the resources and incentives (i.e., margins) to build out their own extension solutions and dedicated teams. Here the optimization relies on ad server campaign reporting, inventory reports from different ad servers and platforms and analysis from the DMP. One publisher mentioned his team is working now to automate optimization at the ad server and data system levels, which he said would “let the computers help determine the model after we found the right starting point.” However, for now the optimization process is quite manual and time-consuming, though worth the effort.

In addition, certain organizations have applied tech development from other areas of the organization, such as search or referral systems, to integrate machine learning tools into extension optimization. As you might imagine, this requires massive infrastructure and data management that goes well beyond the advertising department.

Bumps in the Road

As the industry has been on high alert for fraud for at least the last year, its not surprising that the chief concern for extension is the quality of RTB-powered exchange inventory. Jumping out of the U.S. market into international waters, there’s an even higher level of invalid traffic on top of country-specific quirks.

Pubs can layer on some filters, the first and easiest being a network of trusted sites; a whitelist if you will. Of course, this limits the inventory available for targeting. If you need a wider pool, you’ll have to trust your service provider to weed out the ugly stuff. You should get a thorough explanation of their invalid traffic and brand safety, including anti-fraud technology, keyword and domain blacklists, and third-party verification partners.

Forecasting can be a chief source of grief for extension, especially for new segments without proper “bake” time. One publisher griped that ops has been asked for forecasts within two days of creating a complex segment. “Most of our systems have not had time to calculate overlap, which is crucial for an accurate forecast,” he said.

And remember that I said extension allows you to get super granular? Well, over-zealous planners will fall into the hole of over-specification. Figuring out the bounds of granularity will require some trial and error. One publisher saw success early on in super-focused segments, but found a broader approach identified areas that had been overlooked in planning. 

“You never know which combination will work, so it’s important to fail fast and change the DNA of an audience if needed,” he commented.

Trends at a Crossroads

At the most recent Publisher Forum, one attendee commented, “So not only does my media company need an internal agency now, I have to become a media buyer as well?” 

It was an interesting juxtaposition of two big trends in digital advertising: native and extension. A recent PubForum presentation went into the intricacies of using media buying strategies to boost traffic to native content – extension is an extremely useful tool for bringing additional eyeballs to your native campaigns, and ultimately can bring you host of new users.

Sure, some pubs may lament that it feels like they have to do everything around the digital advertising space these days, but others have seen the opportunity behind the “burden” – buying third-party inventory for advertisers can derive additional value from their heralded first-party data. In effect, pursuing audience segmentation for extension is a low-effort way to boost performance via reaching both larger audiences through modeling and finding more granular users off-site. 

Forget “Drill, baby, drill!” – it’s all about “Refine, baby, refine!”