Black Spruce Analytics

Classification:
Say you have large amounts of outcome data about your past and current students, clients, or employees in a particular program or department. You have intake surveys or early performance data about new students or clients.

You could build a predictive model that classifies your students or clients into two or more categories of expected outcome, based on your intake survey or early performance data. If they are predicted to fall into categories with negative outcomes, you can now effectively target interventions to mitigate negative outcomes.

Classification:
Build a classification model to predict concentrations of an element that isn’t commonly assayed, based on commonly assayed element concentrations and other non-assay metadata (e.g. bedrock/surficial characteristics, other lithological factors). These models go far beyond a correlation matrix, and can process thousands of samples with hundreds of predictor variables (e.g. 10,000 samples with 80 assayed elements each).

You could use this model to process data from your own property, for instance using large-scale soil, chip, or water chemistry sampling methods, where you are interested in minimizing costs per assay.

You could further apply this model to publicly available data (e.g., from Natural Resources Canada, Geologic Surveys) to highlight areas where there exists potential for a similar deposit hosting your target element/mineral.

You have 50 years of water main break data, and want to answer the following questions to optimize maintenance and capital replacement:

• Are the mains susceptible to infant mortality or do failures peak around the expected life of the asset? Is there a different mode of failure?
• Is there a statistically significant relationship between maintenance intervention and subsequent breakage?
• How do the different types of pipe installed over the years perform?
• With mathematical answers to these questions, you can then look at existing installed mains and gauge which are likely to fail in your time horizon of interest, based on their type, installation date, mode of failure, and maintenance history.

Preference Analysis:
Consider a situation where there are several design options for a revitalized downtown, or several re-purposing options for the under-used community pool. Council or a community planning agency is charged with determining stakeholder preferences for each option. Each option consists of various attributes, for example “annual cost”, “more green space”, “tennis courts”, “studio space”, etc.

When dealing with stakeholder preference, opinions are often sought through surveys. A common mode of determining preference is to rank a number of options from most preferred to least preferred. On the surface, this mode of survey doesn’t provide any information on the attribute preferences of the stakeholders.

Using a conjoint analysis model, you can tease out the preferences of stakeholders for each of attributes represented in the options (technically called “part-worths”). When applied at an early stage, this can allow for optimization of the re-design project to include those attributes which are most important to the community.

This level of quantitative analysis of preference can be further used to inform decision analysis if it is applied to the selection process.

System Dynamics:
A high-leverage policy is a policy that achieves your desired outcomes with minimum intervention. Often a holistic perspective is required to understand feedbacks in the system—a perspective offered by system dynamics models. When feedbacks are understood, you can interrupt or amplify them to achieve your goal.

Consider the following apocryphal tale of a high-leverage policy.

Locks, socks, and costs:
Betty is a homeless person. In Betty’s operating environment there are many risks to the feet, especially for someone with diabetes. These risks can become severe or life threatening: pain, absceses, even necrotizing fasciitis.

If a homeless person has hurt their feet, they have both the right and the option to attend the local ER and seek medical attention. Health care costs and ER wait times are politically charged topics. We should seek to minimize both. And we should seek to improve the wellbeing of those who are marginalized.

What does it take to help Betty and other homeless people reduce the risks to their feet, thereby improving their wellbeing and reducing the pressure on ERs due to preventable foot-care issues?

The answer, as related in this tale, is relatively simple: showers and clean socks. In many large centres there are free showers available to the homeless. But they aren’t using them. Why not?

Because most of these facilities don’t offer any way for homeless people to secure their personal effects. Imagine if you had to risk everything you owned just to take a shower.

What is the solution? Provide some personal storage with locks. Provide some clean socks. This will promote healthy feet, and work toward minimizing the extra health care costs from preventable foot ailments among the homeless.

Now… “what would this cost!?” will be the cry. Well, the real question is what will it cost versus the costs we are already incurring. What is the cost of a simple visit to the ER? How many of those can we prevent with free locks and socks? Rather, how many preventable visits to the ER does it take to pay for the locks and socks?

If the savings are several million per year, and the cost is a few hundred thousand, that’s a high-leverage policy. When we think holistically, we can find a win-win-win.

System Dynamics:
Building a structural model of your organization’s supply chain can help you to understand the many variable, dynamic processes that cause you headaches. Understanding these processes can help you manage your business more effectively by being prepared for upstream shocks.

By simulating your supply chain and your organization’s policies, you can see what happens when dynamic events occur, and test policies to mitigate their effects on you and your customer.

• Bullwhip: How does an X% supply shock in the first tier (raw materials) amplify itself by the time it reaches you?
• Hoarding: How do human-induced effects like panic buying, hoarding, and speculation play into your input price?
• Policy Tests: Sometimes, people behaving rationally, in a silo, can produce strong negative feedbacks. What are your proposed policies to mitigate these supply chain effects?How do these policies interact inside and outside your organization? In what ways can you optimize your policies?

System Dynamics:
If you’re stuck in a reactive maintenance regime, you can build a model that shows where the feedbacks are that are keeping you there. This model would include “soft” variables such as morale, as well as models of asset reliability.

A properly calibrated model can let you estimate both the cost and time required before senior management sees a turnaround.

Go further and show what happens if the funding for proactive maintenance is cut in the future, and how this affects the bottom line through decreased downtime and decreased production.