7 Goal Attributes (3 defined by you), 2 Competing Decisions
(recommended browser: Chromium, remember to re-run SMAU after changing inputs and before displaying results)
Attribute Importance weight information (entered below)
All required values are between 0 and 1; Beta distributions are assumed and rescaled so that weights sum to 1.0. Enter (small) standard deviation=.01 if you feel certain about a weight.
Why put probability distributions on Weights?
Your attributes' importances may depend upon uncertain outcomes such as upcoming Political/Regulatory/Trade environment or Financial Markets. Or there may merely be fuzziness in your valuation of attribute importances.
SMAU - What is It?... See Quickstart definition toward bottom of this page.
What's It For?
Decisionmaking.... making the Best Adjustments in Organization, Product Line, Product Design, Marketing Channels, Supply Chain array, etc.
• Also applicable to Lower Level (Departmental) decisions, as well as to Top-Level organizational decisions
• Decisions may also be over a continuum of levels or amounts (eg. product design/engineering parameters), instead of discrete disjoint ones.
• Impact Assessment: New Products, Policy Changes, Reorganization,... you name it! Use it to assess Total Improvement from Expansion, for optimizing/mitigation among reorganization alternatives, etc.
Not Contemplating any Decisions? Use SMAU to Establish Current Performance.
•You're always striving for improvement. SMAU gives you a probablistic, quantitative, unified measure of total performance (Monetary Equivalent of Total Utility, MEq), reflecting the relative importances of KPI's and all attributes, instead of separately looking at only a few KPI's and attempting to mentally weight them and mentally consider tradeoffs for different combinations of their levels.
•The effort of merely establishing the MultiAttribute topology has been found by others to be worth the expense of fully establishing the system.
• Current attributes' performances (baseline MEq curve) are also assessed as a benchmark for comparison - if prospective adjustments or decisions don't perform well enough (their MEq curves), best decision is status quo ie. no action.
SMAU Consulting Products and Pricing
You'll Run the System Too!  In each effort below I'll conduct the official elicitations and analyses, but you'll also do informal elicitations and SMAU runs via web interface apps, to gain familiarity with your decision framework, but also to possibly optimize prospective decisions or even formulate new ones.
Introductory Application for your Corporation - Full Analysis of Small Framework: ($9.9K) 8 Attributes (all in the same category), 2 Competing Decisions plus MEq Evaluation at Current Attribute levels, formulation of Decision Performance Requirements, 1 Decisionmaker, 1 Expert, Comprehensive Report with Decision Recommendations. Also an adjustment is elicited that allows the utility function of one attribute to change with earnings level and one other attribute, eg. allowing for increased possibilities when earnings increases.
Please Note: To protect the confidentially of your Decisionmaking and Business Intelligence, to prevent conflicts of interest on the part of the Consultant, and to increase your confidence in the Consultant and freedom in the valuation process, (with the exception of earnings, used in calculating Monetary Equivalent) all Attribute, Category, and Decision names shall be generically relabeled for Consultant, who will at no time have knowledge of their actual names or functions.
Stochastic MAU Videos
(demo.mp4 and short_demo.mp4: use Chrome Browser or save and open in VLC Player)
•Decision Impacts' Uncertainties were not made scalable for this example, but as Variation Scales on Weights and Utility Functions both become negligible, MEq curves steepen, reflect only Impact Variability, and are more distinct by decision.... Try it.
•Alternatively, as these Scales are increased greatly, MEq curves for decisions will tend to widen and overlap because the majority of variation is due to Decisionmaker uncertainty.... Try it.
•In addition to the Decision Admissability and Selection Criteria discussed in the video "demo.mp4", the ratio of MEq to Earnings is another way to assess the value contributed by non-Earnings attributes: if larger contributions are desirable, the exceedance curve for this ratio should lie to the right of 2 or 3 (ie. probability of the ratio's exceeding those values equals 1).
•To add interesting variety in this demo, attribute outcome distributions are randomly selected for each attribute, in each decision (they're actually elicited from experts in the project applications). So here sometimes decision 1 will be clearly the best by any definition (ie MEq exceedance curves have identical shape [same variation, etc] but decision 1's lies entirely to the right) and sometimes the converse will hold. Other times, if decisionmaker fuzziness is low, the decisions' MEq exceedance curves can be shaped differently. To see the effects of various attribute impact distributions (only - ie. decisionmaker fuzzinesses very low) try initially setting variations' scales low - at about 0.1 each - along with providing the weights' distribution info. Then repeatedly cycle thru "submit inputs" "run smau" "display MEq and Earnings results" (without reentering any inputs), in each cycle noting the attributes' distributions whose weight distributions tend to be high (most important attributes) then examining the resulting MEq curves.
In MultiAttribute Utility (MAU), each attribute's level is scored according to its utility function, then all attributes' utility scores are adjusted by their respective importance weights and totalled.
This Total Utility may be adjusted to a net positive utility (undesirable attributes' disutilities subtracted from total positive utility), then converted to a more useful measure - Monetary Equivalent - by inversion back through the Decisionmaker's importance weight and utility valuation, of money (Earnings).
Now, Stochastic MultiAttribute Utility Decisionmaking: Dealing with Uncertainty
In Stochastic MultiAttribute Utility (SMAU), importance weights and utility functions reflect the Decisionmaker's fuzziness, and each Decision's impacts on attribute outcomes are uncertain and described by Probability Distributions. ... So MEq itself has a probability distribution, assessed by Monte Carlo methods. This is Uncertainty Analysis applied to MultiAttribute Utility.
(see demo.mp4 for definitions & decision selection)
INPUT: Attribute Co-Tendencies, by Decision
In this demo, decision-specific Attribute impact distributions are randomly assigned for all 7 attributes (instead of being elicited), but here the user may still control outcome Correlations between Attributes 5-7.
Decision 1 Pairwise Correlation Coefficients
Decision 2 Pairwise Correlation Coefficients
When attribute outcomes are correlated for a given decision, it's important to incorporate their correlations so that impossible attribute combinations are not included in the Uncertainty Analysis of MEq, which would bias its exceedance curve from the true one.
