Use Case: Telco customer data for advertising in OTT video

In our upcoming report on Two-Sided Business Model Use Cases in Advertising and Marketing (part of a new Advertising & Marketing 2.0 research practice), we examine more closely how telcos can leverage their data assets to improve their positioning in the advertising and marketing value chain via a platform strategy.

As a taste of our ongoing research, we presented a short example from the over-the-top video world at the 5th Telco 2.0 executive brainstorm at the beginning of the month, to test out some of the concepts with the audience. Interestingly, the presentation seems to have resonated with participants, who were asked to rate current telco strategies for supporting the advertising and marketing industries.

They responded with an overwhelming bias towards "weak". Commentary in the feedback session included observations that telcos are in an ideal position to add insight into their customer base, but have failed to data-mine properly or develop monetization strategies. The audience also identified three next steps for the industry: understand the data, respect the customer, and open the system.

enck-eventNov08-1.png Below is a use case scenario for telco customer data in the context of an over-the-top video and advertising platform:

The event participants' feedback really goes to the heart of our approach. A central tenet of our work around the Two-Sided Business Model is that telcos could, and should, do much more with customer data beyond the purely operational contexts in which it has been used historically (see article on the Customer Data Revolution).

This requires that telcos understand not only what data they have, but also what significance that data may have for an upstream partner or customer. This may not always be obvious, and indeed, much of this data has been viewed by telcos historically as the by-product of delivering connectivity, voice and data - the wood shavings on the workshop floor, left to be swept away at the close of business. The opportunity, as we see it, is to harvest, analyse and put that data to work.

First of all, we need to briefly revisit what data assets we are talking about here. As may be familiar to frequent readers, we have previously defined these along eight category lines, illustrated below. While each of the individual elements of the eight is of varying levels of value to the telco in provisioning and maintaining the services underlying the customer relationship, how might an upstream partner or customer view the same, and what value might they have for them?

enck-eventNov08-2.png

The answer will vary depending on the objectives of the upstream player in question, but broadly speaking, we have defined seven questions which upstream customers might commonly seek to answer, and aligned against them telco protocols and assets which might hold the answers, or at least give additional insight as to what we can infer about likely answers.

enck-eventNov08-3.png

Because the range of potential platform customer needs is so vast, let's narrow the discussion down and look at a specific potential upstream customer of telco data assets and consider what it might seek to gain from having access to them. We have chosen to look at a theoretical use case involving Move Networks.

Move Networks is a venture-backed private company based outside Salt Lake City, Utah in the U.S.A. With investors including Comcast Interactive Media, Cisco, Microsoft, and Grupo Televisa, and clients including ABC, Fox Networks, Televisa (see interview here) and ESPN.com, Move Networks might be considered as a poster-child for how traditional media and players in the distribution value chain will seek to harness and monetize internet video in future. Move Networks' offering to content publishers is in fact both a distribution and advertising platform, with a number of differentiating facets:

  1. Upon content ingestion, content is split into many smaller segments, with each segment encoded at a variety of bit rates to suit publisher targets for viewer experience across a range of access technologies (dial-up, broadband) and end user devices (mobile, computer, television). This content is stored on multiple standard HTTP servers rather than proprietary media servers.
  2. Rather than a dedicated desktop client or media player, Move Networks uses a lightweight browser plug-in, the presentation of which varies depending on content partner, and can be branded and re-skinned as required. The Move Media Player plug-in constantly provides the HTTP servers with a view of the network conditions, throughput and latency at the client side, allowing the servers to seamlessly synchronize and insert lower or higher bit-rate content segments dynamically, as required. This is Move's solution to buffering, and reflects the company's strategy to satisfy traditional media partners' demand for a broadcast-like experience in consistent near-HD quality - a key consideration given that some of its media customers are offering content on a subscription basis.

  3. Given that the media player plug-in reports a variety of information back to the servers, Move has the ability to track and analyze viewing histories, day-part viewing patterns, abandonment rates, pause patterns, etc., which provides the content publishers and media buyers with a granular view into viewer behavior. This can be useful in designing and refining programming, and particularly, advertising formats. The Move player is also able to return high-level geo-location data about viewer location, and presumably high level browser data such as language settings and operating system.
  4. Beyond the granular behavioral analysis which Move affords advertisers, it has the added advantage of being "TiVo" proof, in the sense that viewers cannot skip ads. For "must-see" content, this gives advertisers the confidence to sponsor entire television episodes, as was the case with the original "soap opera" format in the early days of radio. One demo we have seen of Move advertising capability includes a "sponsored intermission" for Toyota, with one frame in the window reserved for an ad avail for the local Toyota dealer, as inferred from the viewer's geo-IP data.

enck-eventNov08-4.png

In some respects, Move's offering would naturally be of interest to traditional content publishers and media buyers, as it delivers something akin to a conventional broadcast-like experience, but with a greatly enhanced behavioral tracking and analysis capability, which gives them the ability to continually experiment with and optimize advertising formats. Moreover, it theoretically allows advertisers to more accurately infer user preferences in pursuit of more finely tuned targeting.

For example, if Move's analysis of a viewer's history shows frequent views of sports or automotive-related content, late at night, in a city which is home to a number of universities, it may
be a decent bet to conclude that the viewer in question is a male student, and Move can then serve a suitable ad from its inventory.

Move's platform is considered by many to be the envy of the industry, and it certainly seems to be further along the path of mainstream media world acceptance than many
of its competitors. So what, if anything, might a telco do for Move?

  1. Guaranteed QoS? - This is already the case to some extent, at least in the sense of traditional CDN services. Move has relationships with a number of CDN providers, among them two offerings from telcos AT&T and Level(3). However, we question whether any more is really needed within the individual telco networks (as in network-caching solutions). Given that the Move browser client is the intelligence which manages bit-rate dynamically, the current approach is probably acceptable for the foreseeable future, notwithstanding congestion of a very extreme and persistent nature.
  2. Co-marketing? - Not likely. In the absence of some revenue share agreement between telcos and publishers, there is no incentive for the telcos, and in the absence of any added value being delivered by the telcos, there is no incentive for publishers.
  3. Market intelligence? - Quite probably. As we mentioned above, at the moment Move actually only has a limited amount of information with which to infer viewer profiles. Undoubtedly, a bank of telco customer data parsed on a more granular basis (location data, payment and credit-related histories, calling patterns, web session data) could make Move's targeting capability more sophisticated, which should translate into higher CPMs on its platform.

It's important to note that we're not talking here about data identifying individuals which is handed over to Move, at least not without a very robust and transparent opt-in system. Sensitivities over privacy aside, it may be, as we heard from one company in the analytics space recently, that for some upstream customers, too granular a dataset introduces too much complexity and negates the benefit of the data.

Rather, we think it might be sufficient for Move or a similar player to have access through telco APIs to anonymized customer segment data which defines customer attributes into buckets or categories, which could then be matched against both Move's own analysis and third party data sources as appropriate to an advertiser's own targeting criteria. This approach could form the first step towards a telco-specific Mosaic of sorts.

Specifically, we think the following telco data assets would be relevant to Move Networks' targeting efforts, while keeping customer data in the non-controversial arena with regard to privacy (please note we have excluded mobile data from this as Move currently is only compatible with Internet Explorer 6 and 7, and Firefox 2 and 3, which makes mobile largely a non-issue for the immediate future - though a range of mobile data assets would definitely be attractive to a comparable player with a mobile delivery capability):

Product data - A telco API could allow the Move player to query a database of anonymized telco customer data sorted by product mix. Depending on the targeting criteria defined by Move or its advertisers, the range of product choices by telco customers can inform ad targeting decisions. What proportion of the customer base is top/mid/low tier? What proportion takes multiple services? What proportion makes use of advanced options like voicemail, ring-back, etc.? What proportion are pay TV households? Of those, how would the average channel bouquet be classified? What proportion of the customer base has made service changes recently (upgrade, downspin, or cancellations)? Married with other telco datasets such as location, this manner of segmentation could be very interesting to an advertiser in defining audience segments by inferring "wallet size", technology adoption, media consumption and other attitudes and preferences, as it might fill in some information blanks left by other targeting tools.

Session/call-flow data - For advertisers to whom factors such as sociability and "connectedness" are relevant, telco data is a treasure trove. For the entire customer base, or for specific sub-segments as defined by other telco datasets, what is the ratio of inbound to outbound calls? How large is the average household's calling circle? What is the frequency and duration of calls? What proportion of the customer base could be considered above average in calling habits? What is the time distribution of calls? What proportion of the broadband customer base could be regarded as "super-connected"? From each of these questions, which telcos can answer quite definitively, advertisers and brand managers can make inferences about sociability and "connectedness" which inform ad targeting decisions, particularly if cross-referenced with other datasets, both telco and non-telco in origin.

Payment history - Again, depending on the criteria defined by the advertiser, insight into the telco customer base's payment history and relationship with the telco itself can potentially yield valuable insight, particularly when linked to other subsets of data. What proportion of households pay by monthly direct debit vs. manually or by phone? What is the timeliness/reliability profile of the customer base or sub-segment? What are the correlations between delinquency rates and product mix? How has that trend changed over time? All of these can be mixed with other criteria, such as location, to triangulate customer segment profiles.

Loyalty measures - Similar to payment history, just as brand advertisers might be keenly interested in targeting based on attributes around ability and willingness to pay, loyalty is another key metric. Churning from telco services is much more painful for the consumer than changing brands of soap, and thus may not be comparable data from the view of some advertisers. However, as part of a broader consumer profile, data around length of customer relationship, attempted churn/churn pre-emption/win-back history and churn risk level (both as defined by telco CRM software) would no doubt be of interest to advertisers in other industries where the pain of churn is high for the customer, but equally devastating for the supplier (banks, insurers, utilities).

Location data - Married with all of the above, more specific location data is probably an El Dorado for a company like Move Networks. A telco API could allow the Move player to query a database of anonymized telco customer data by local exchange. This could then be referenced against either third party data (census, credit scoring, lifestyle surveys) or other internal telco datasets (product mix, CDRs, broadband usage intensity, payment history, loyalty measures) to create customer segment profiles by location, which could then be included in the targeting decision algorithm. For example, the output to the targeting algorithm might effectively say:
  1. "50% of customers connected to exchange #2008 exhibit high sociability levels (they make and receive lots of calls);
  2. 40% are classified as high spenders (they take top tier or multiple services, or recently added or upgraded services);
  3. 30% are super-connected (they spend significantly more time online than the average for this area);
  4. 70% are credit-risk code green (they've never missed or been delinquent with a payment);
  5. 35% are classified as brand loyal (they've been with us for more than three years).

The ad ultimately served will depend on a wide variety of criteria stipulated by the advertiser or media buyer, but we believe there is no doubt that the non-invasive, anonymous telco customer data employed in generating this decision has immense value to the decision makers.

There are, however, many open questions surrounding this use case, with which we grapple and hope to answer in the final version of our report. For example, would advertisers be interested in this sort of data if it only related to one operator in isolation? Our view is probably not. This begs the question of whether telcos need to federate their data to provide as representative a sample as possible by country.

This, in turn, begs the question of whether the new data platform in question should be outsourced to an independent entity, whether an existing player in the data management space or a newco joint venture between telcos. We welcome your comments, criticisms and suggestions on this work in progress, as we firmly believe it should involve those who live and breathe the complexities of telco reality on a daily basis.

[Ed. - The Use Case Report will be available as a large report in Feb 2009, or available in installments via the new Telco 2.0 Executive Briefing subscription service].