Sam Watts Discusses Pathfinder, SMG's In-House Attribution Modelling Tool

by Ciaran O'Kane on 9th Aug 2011 in News

Sam Watts is Associate Director, Data & Analytics at Starcom MediaVest. Here Watts discusses Pathfinder, SMG's in-house attribution modelling tool, the value of the display impression, and moving beyond the last click/ last view metric.

Can you give some overview of the Pathfinder solution - and how it’s helping Publicis clients allocate media spend?

Pathfinder is SMGs in-house approach to online attribution modelling. Pathfinder uses a combination of a data driven approach combined with a classical media analysis approach. Rather than re-inventing the analysis wheel, we recognise that the skills and approaches that we have built up over the last twenty years, from the advent of adstock modelling in seventies through to econometric modelling can be applied to the online world. Using Pathfinder allows our clients to understand how effective all their online channels are, how they work together and from this make informed decisions on how best to allocate their media budgets.

There’s been a lot of research on the impact of offline media channels on online media performance. How does Patfinder discern these effects and apply data to client attribution modelling?

Bringing offline data into the attribution process can be complex but not impossible. By its nature, online data is available at an individual level whilst the majority of offline media is held at an aggregated and mainly inferred level. This leads to a two-stage approach to bring in offline data, for example, Direct Mail can be linked to the click stream and dealt with in the same way as an online channel. In order to include broadcast channels a different approach has to be taken; combining the results of Pathfinder with the outputs of a disaggregated econometric modelling project to give a complete view of media performance.

How is Pathfinder improving planning within Publicis?

The main aim of Pathfinder is not to be purely an insight report, rather to be an integral part of the monthly online planning process. With this in mind we have created a simple evaluation framework, which allows all placements in display and keywords in PPC & SEO to be quickly and efficiently evaluated.

The last click/ last view metric dominates the thinking of most media planner buyers when allocating reward for hitting campaign KPIs. Pathfinder is clearly trying to move up the purchase funnel when assessing all media touch points. What criteria are you using to assign exposures in the path to conversion?

We use two fundamental criteria when assigning value to exposures. The first fundamental assigns based on the likelihood of an exposure being served to someone who eventually goes on to convert, this allows us to reward those placements that are successful in targeting the correct audience. The second fundamental harks back to TV analysis in the seventies, in that we follow the principle that an exposure today is worth more than an exposure yesterday. From this principle we assign a modelled decay curve to each consumer path, ensuring that we credit the exposures closest to conversion.

Do you think that the value of display advertising is still unclear to some marketers? Where do you think it sits in the marketing mix?

I think the value of display advertising is still unclear in two ways. Firstly, the value of a display impression in direct response evaluation terms is still clouded by lack of a standard approach to post-view conversions, some of our clients give no value to an impression, others give full value within a given time. Implementing Pathfinder in combination with a display ad verification tool allows us to answer half of the display conundrum . The second issue is around the value of display advertising as a brand media in driving softer measures such as awareness and consideration, in order to answer this more consistent research is needed.

A lot of agency attribution models crunch the numbers based on post-campaign performance. Will display buys via RTB enable Pathfinder to deliver real-time analytics to inform media buyers during actual campaign?

We currently see Pathfinder being used as a strategic report to evaluate performance post campaign, which is broadly in line with the rest of the marketplace. We are currently working with the AOD team at Vivaki to develop a solution that will allow Pathfinder attribution to be plugged into AOD.

Can you give some overview on the Pathfinder "investment optimisation app" – and how it is helping media planner buyers make informed buying decisions on behalf of the client?

The Pathfinder investment optimisation app provides everyone involved in investment
decision making with a tool to assess what is working best. We recognise that all clients are different, therefore each app is custom built to meet the needs of the end-user. Within the app the user can see all findings at many levels of granularity, from the CMO who just wishes to see top line performance by channel to the analytics team who go down to PPC keyword or display creative to make small adjustments to the media plan.

How does the process work from collecting data on media performance to optimising the media buying process? Can you walk through how a typical attribution cycle works?

Pathfinder is used as an integral part of the monthly strategic planning process, so at the end of every month, Pathfinder is applied to the previous month of data, once the attribution has been run, a cross-channel planning meeting with members of the PPC,SEO, Display and Affiliates teams sits down and makes a list of strategic recommendations based on the findings. Our client is then taken through these recommendations and they are implemented for the next planning cycle. Whilst much of digital optimisation happens at higher frequency almost real time, Pathfinder serves as a monthly milestone to ensure online investment is being optimised effectively across the board.

How would you say Pathfinder’s approach differs from other agency attribution solutions?

There are two key differentiating factors to Pathfinder, one technical and the other organisational. The technical difference is the fusion of advanced statistical modelling with digital clickstream data. The organisational difference is the deep integration within our planning teams, all planners and buyers are fully trained in Pathfinder and how to use it, ensuring complete confidence and reliance on findings from both clients and internal teams.