Embedded versus Consultative Decision Analysis

At a recent DAAG conference in Boston, I chaired a session on Embedded DA where over 100 Decision Analysts got together to reflect on the contrast between Embedded and Consultative DA. Three perspectives from Decision Analysts with experience in making DA routine in organizations.

From his experience doing decision analysis at Eli Lilly over several years, Jay Anderson shared that what has ultimately survived is their portfolio decision analysis process. In the traditional consultative approach, we normally present a map of the full process so people get a sense of what lies ahead. In contrast, Eli’s portfolio effort took an entirely different approach.

First, the decision analysts involved did not try to sell the organization on Decision Analysis. They just tried to facilitate a conversation about their portfolio. In doing so, they applied decision-focused thinking, focusing on things that mattered to the organization.

Second, they focused on producing relevant, reliable information. Their finance and marketing groups at that time were deterministic and uncertainty was seen as a sign of weakness. Jay provided a more insightful approach, that included uncertainty.

Third, the team’s unadulterated comments got preserved and shown. In previous decision making process, information had been edited at each level in the hierarchy as it moved to the top, resulting in significant misunderstandings about what the project participants said about the project. Jay created a system in which project participants comments appeared on one page, with management interpretations on a facing page. This greatly increased the information transfer and value of both participant and management perspectives.

Fourth, Jay’s team took a holistic view of the entire portfolio of the organization. This completely changed the frame of their business conversations. At first, the head of R&D was dead set against doing this but he later changed his mind. At the end of the day, lots of what has stuck from that work to this day is what drove discussions.

Matthew Kurtz from Chevron talked about the disconnect that happens between day-to-day decision-making and longer-term decisions in an asset-base business where there are lots of daily decisions. It is only the longer-term decisions involving large scale, complex projects that gets DA consulting attention, and Matt needed to link up small capital programs to longer-term priorities. This led him to come up with a scaled approach to DA.

He rolled up small capital decisions into a portfolio of capital to get an integrated perspective. Decision quality was achieved through open dialog. He looked at different tools and prepared a Decision Quality continuum. At the light end were traditional forecasting methods. At the heavy end was the use of DA consultants. In the middle were functional groups using fit-for-purpose tools. An example of a decision in the middle would be “Our water system broke, should we replace or upgrade in anticipation of future need?” or simple low-mid-high forecasts.

Matt’s effort changed the way middle management made decisions. Traditionally, the petroleum engineer would make a model that was the basis of a pitch. Such a model was very deterministic, myopic and introspective. With the new approach, decision makers could now get visibility to long-term viability of an asset. Engineers started having conversations with each other about the implications of their decisions over time. Earlier, old assets were being reactively fixed, and now, decisions started to be made ahead of failure.

By embedding DA in this way, mid-level decision-makers didn’t have to wait for the DA specialist – they were empowered to go ahead on their own using a fit-for-purpose tool.

Somik Raha from SmartOrg spoke contrasted the Consultative and the Embedded worlds of DA from a Systems Design perspective. He first pointed out the ground principle that the purpose behind Embedded DA systems was not to replace humans but to augment them. Consultative DA required a “Knobs for Nerds” approach where custom models are created to meet particular needs in a situation. In this worldview, it is a bad idea to reuse models as every situation is different. Embedded DA on the other hand relies on “Lean Models” which rely on the deployment of common simple templates. In this worldview, it is a good idea to reuse models to aid in the spread of a disciplined culture. Several implications of this different emphasis:

The user-interface (UI) of systems in Consultative DA  are more control-focused, allowing analysts powerful control over every aspect of their model. On the other hand, in Embedded DA, the UI is more role-focused, catering to different stakeholders, ranging from decision-makers, SMEs to Analysts. Somik highlighted the different concerns of different roles through a video vignette of a finance director using an Embedded DA tool, saying, “Well the good thing about something like this (embedded DA tool) over Excel is that people always go and change the bloody formulas, don’t they!”

The workflow in Consultative DA is facilitated through an expert analyst who sends out requests for data, carefully reviews it and plugs it into the model for analysis, whereas in Embedded DA, it is mediated in real-time by the system. This supports distributed work across roles and geographies, and the system turns into a storehouse of useful information. The pedigree of the data is assured by the analyst in Consultative DA, whereas it is assured by the system in the Embedded DA.

Finally, decisions are one-off or unique in the Consultative world, whereas in the Embedded world, decisions are made in the context of a stream of similar decisions.

 

The audience reflected further and contrasted Consultative DA to Embedded DA, and below are their contributions: