Analytics & Optimization Decision Fundamentals: Building Institutional Memory

Elephants
Jun212016

This is my fourth of five blogs expanding on the key points of my FICO World 2016 address. Through these blogs I’m exploring the fundamental ways organizations need to improve their operational decision-making. Organizations can reduce complexity and create competitive advantage through:

1. Capturing subject matter expertise
2. Intelligent solution creation
3. Faster insight to execution
4. Building institutional memory
5. Greater analytic accessibility

Let’s talk about institutional memory.

How Elephants Survive

In my FICO World presentation I included a slide with a photo of elephants—not the usual stuff of a corporate deck. But a tale about elephant survival in Tanzania’s Tarangire Park provides a valuable allegory for business.

The Wildlife Conservation Society examined patterns of elephant calf mortality during the park’s drought of 1993. The drought was the region’s most severe in over 40 years. Sadly, during a nine-month period of that year, 16 out of 81 elephant calves died. This mortality rate of 20% was 10 times higher than during non-drought years.

Researchers noted correlations between calf survivorship, the movements of the groups, and the ages of the female members of the groups. The two groups that left the park during the drought suffered lower mortality rates than the one group that remained.

The groups that left the park had the oldest matriarchs, elephants that may have drawn upon their memories a Tarangire drought that lasted from 1958 to 1961. The group staying behind had no elephants old enough to have remembered that event.

This story about drought survival in Africa is highly applicable in a business context…institutional memory brings tremendous benefits.

Building Institutional Memory

Most professionals have been in a situation where they had to examine decisions made by a group of individuals no longer with the company, or who had moved to other departments. There was no supporting data to be had, nor any documentation of the decision process. Just as in Africa, without any institutional memory, there can be dire consequences.

Institutional memory is critical in operationalizing data-driven decision-making. Unless you intend to rely on “matriarchs” who have been around for 30+ years, you’re going to need a system for this, and the system requires five things:

1. Inventory of all decision assets: The most important thing, when it comes to capturing and learning from institutional knowledge, is the creation of a single repository of decision assets that can be mined and searched.

This inventory of decision assets needs to hold information at all levels of granularity, and regardless of whether that asset was created in your shop, or bought or borrowed from outside. All of the following elements must be centrally managed:

  • The variables used to create a score characteristic or a business rule
  • The knowledge models that use those variables, such as a predictive model or a strategy tree
  • The decision models that define how that knowledge is applied to data to make decisions.

2. Decisions correlated with data: We need to correlate all those decision assets with decision data (the data from the decisions made with those assets). This can provide a full picture of what decisions were made, how they were made, and how they have impacted key performance metrics in the past.

3. Record of decision workflow: It’s necessary to know, at any given time, who did what, with what, to what and whom, with whose approval, and why it was done. This record of decision workflow puts you in a position to build and better leverage your institutional memory. It’s a good business practice that can unlock value, and is also increasingly mandated in regulated industries.

4. Transparency of governance process: Decision assets need to be managed across their lifecycle, with a transparent governance process establishing how each asset can be used, updated or replaced. This is especially valuable in highly regulated industries.

5. Validation and analysis: We must understand an asset’s role and constantly validate if its performance has deteriorated over time. In addition, we must proactively measure that performance to be able to anticipate issues (and opportunities).
Let’s see how all of these steps come together in the real world.

ANZ Builds Institutional Memory with FICO Decision Central Solution

Models are a challenging area for banks to build institutional memory. ANZ is an Australian bank with a complex inventory of 170 models across many portfolios and several countries. ANZ had many different individuals manually creating, modifying, monitoring performance and demonstrating regulatory compliance for those models, creating significant challenges, particularly in transferring knowledge of models if an individual changed roles or left the bank. ANZ recognized that it needed a new approach to help reducing the amount of time spent monitoring, documenting and identifying deteriorating models.

The bank implemented FICO® Decision Central™ to transform these models into a powerful institutional asset via automated reports, streamlined model reviews and simplified documentation, to sustain better model health and streamlined regulatory compliance. They created a “single source of truth” for multiple teams, which improves collaboration, enhances productivity and overall business results. Consistent analytic assets are now produced automatically across all portfolios, creating a long-lived institutional asset, which permits individuals to focus on innovation and insight generation.

In my next blog I’ll wrap up my discussion of decision fundamentals with a post about greater analytic accessibility.

Leave a comment