13 Jan
contextual-data

Data & Context in 2014

The dawning of a New Year brings a surfeit of “Predictions for 2014” posts. While there are numerous topics, I’m interested in discussing some that touch on the state of technology – specifically, how the driving force behind the internet (advertising) will change, and what that means for consumers.

I’ve seen a number of authors posit that i) data will drive more programmatic sales beyond the display banner; ii) the demise of cookies will result in additional 2nd-party data relationships; iii) large publishers’ internal IDs will silo campaign budgets; iv) econometric modeling will prove sufficient enough to formalize attribution models for omni-channel campaigns and budgets will shift accordingly; and v) wearable technology will be adopted by consumers providing yet more data for marketers.

I am not here to argue or further expand on any of the above topics. It is my belief that those who license, aggregate, and leverage data for paid media will shift the focus of contextual data from an input source for profiling to a qualifier when making bids in real-time.

Historically, datasets have been aggregated from multiple parties (1st, 2nd, 3rd), normalized, and segmented based on both static demographic and dynamic intent attributes. Data sources include search queries, browsing behavior, social interactions, and context. While traditional print – and even web – publications offer contextual targeting based on copy and audience, I think that context variables will become increasingly important as we move towards even more real-time information. Let me explain:

Location

The rapid adoption of mobile devices means wide availability of location data about consumers, their travel patterns, and shifts in those patterns. Businesses like PlaceIQ & Factual rely on location signals to target, build individual profiles associated with device IDs, and determine attribution for a media campaign based on shifts in consumer foot-traffic. The industry is embracing this shift, and, with the recent release of Apple’s iBeacon, we expect that location data will become more accurate and actionable.

Historic location information has been available for some time; but I think there’s much more value in having individual location information available real-time. A user’s location is relevant to determining if they are a qualified buyer – or, in more formal sales terminology, as a means of qualifying the prospect before investing significant energy in converting them.

For instance, Ford may be interested in targeting auto-intenders. As a segment, auto-intenders encompass a wide range of demographics that have leased a car ~30 months prior, or exhibited behavior online that make us assume they are in-market. However, location information – as an additional targeting layer used in real-time (post profiling) – can be leveraged to target on-the-go mobile users who are currently at a dealership fact-checking or those at a dinner party discussing vehicles with friends who are higher in the purchase funnel. It allows the advertiser to ultimately serve more relevant content, based on where the consumer is in the purchase funnel and what they’re doing. A discount for the premium trim on the newest model Ford Taurus is out of place at a dinner party, but tremendously relevant when they’re at the Ford dealer (or, better yet, the Nissan dealer across the street).

Patterns & Anticipation

Cezary Pietrzak shared his outlook on mobile and highlighted the “growing number of sensors on devices” and the unprecedented amount of data encompassing individuals’ usage. As a result “apps [are able] to anticipate a person’s behavior and quickly deliver the appropriate solution.” While some more established companies like Google Now, Foursquare, and Sunrise have made headway in the space, I am more interested in the potential of a few smaller ones – all focused on optimizing the mobile lock-screen. I’d watch out for Yahoo’s use of recently acquired Aviate, Facebook’s Home, Cover, and Locket.

Other, “more powerful and independent motions sensors will enable developers to use new gestures to augment classic touch interactions.” As Pietrzak discusses, simple, repetitive motions – like putting your phone to your ear – can be enough of an input to trigger your smartphone to pick up the call. Real-time data gathered via local sensors will soon be leveraged to further inform the use of data and optimize the user-experience.

Empathy

Context encompasses more than sensors and location data. Context can be based on time, holidays, and, perhaps the hardest to account for, real-world events. Om Malik said it best in a recent post on data’s meaningless purpose without a soul: “As we move towards a quantified society, one shaped by data, we start to dismiss things that are unquantified. Empathy, emotion and storytelling — these are as much a part of business as they are of life.”

To pull from Malik’s example, Uber’s surge pricing – while I completely agree with their Supply/Demand logic – should have been adjusted given the circumstances in the Northeast post Superstorm Sandy.

Weather

To expand upon the above, weather can be used as a contextual attribute to adjust pricing models, but I think there’s more potential in using weather as a contextual filter to optimize a paid media campaign. Anthony Lacovone, CEO of mobile ad network AdTheorent, mentions that “Weather is a big component of the data that we look at. Before we serve an ad, we always look at the weather condition of the region including humidity, temperature and precipitation.” He goes on to give specific examples of the effects temperature has on Click Through Rates (CTRs): “When it’s really hot, it affects people differently in different parts of the country. We tend to see that hot weather in New York drives a higher CTR, but high temperatures on the West Coast have an adverse effect on CTR.

Final Words

I am not arguing that contextual data should be ignored in the profiling process, but I believe that there is greater value in employing contextual queues as qualifiers – and then have profiles that predict customer CTR and conversion rate within those pre-qualifying buckets.

The ability of online advertisers to effectively qualify should lead to a more positive, helpful experience for the customer. Advertisers have known for a long time that additional exposure can have a positive effect in the long term, but that kind of nuclear approach is both expensive and insensitive. The approach should be to help the customer, in the most relevant fashion possible, and earn both exposure and a positive brand association. Hopefully we see datasets organized to function in this manner and vendors shift optimization algorithms in this direction in 2014.

[Thanks to Michael Griffiths for contributing to and editing this post.]