Print Page  |  Contact Us  |  Your Cart  |  Sign In  |  Join IMC USA
Daily Tips for Consultants
Blog Home All Blogs
Search all posts for:   

 

View all (805) posts »
 

#442: How to Be More Accurate With Uncertainty

Posted By Mark Haas CMC FIMC, Tuesday, November 23, 2010
Updated: Tuesday, November 23, 2010
We all know that the only guaranteed thing about all models is that they are wrong. There are too many uncertainties in any model, both in its inputs as well as its design, to do more than give us insights. How are consultants to respond to clients who ask them to build or use a model to predict future outcomes?

While true that models are not, and cannot, be accurate in estimating the future, this does not mean that they are not extremely useful. A model is a representation of the real world as well as we can characterize it. In both the building of the model and in collecting data to support it, we develop greater insight into how our (modeled) piece of the world works. The best think about models is that they can be tested and improved until you have a model that is "good enough," although criteria for that judgment vary with the client.

Where most models fall down is that they try to deterministically predict an outcome when the system they are modeling is probabilistic. Say you are building sales model. Product prices may range depending on economic factors. Number of sale staff may vary with the season. Sales effectiveness may vary as a result of a training program that itself has uncertain outcomes. When you try to build a spreadsheet model whose inputs area all uncertain, the outcomes are barely useful. You are left with saying to your client, "Based on average expected inputs, our best guess of sales is $3.5 million in the fall quarter, but it could be more or less."

There is a powerful alternative: changing a spreadsheet from into a probabilistic model using an add-in called @RISK. This commercial software program allows you to specify a distribution instead of a fixed "best guess" or average input number (including various shapes of the expected distributions). If inputs are correlated, you can even link these distributions (e.g., if customer traffic increases, the available time spent with each customer decreases, which may not be a simple relationship). The more you can model, the better you can understand how your business works.

Tip: How great would it be to say to your client, "We have used historical data on all the factors that affect sales and modeled the effects of enhanced sales training. The most likely impact is a $1.2 million increase in sales, with a 20% probability of it being $1.9 million and an 80% probability of it being at least $1 million"? Then you can tell your client which parameters are hiding in the model that have the greatest impact on decreasing the uncertainty in results. Just using this add-in gives you far greater insight into the business you are advising. Check out the add-in at Palisade.com.

© 2010 Institute of Management Consultants USA

Tags:  assessment  consulting tools  decision making  innovation  risk analysis  technology 

Share |
Permalink | Comments (0)
 
Community Search
Sign In


Forgot your password?

Haven't joined yet?

IMC USA Calendar

10/5/2016
The Consultant's Workshop : Ethics in a Chaotic, Challenging World - Oct 5, 2016

10/7/2016
IMC SoCal Breakfast Meeting

10/13/2016
IMC Orange County Business Forum Breakfast

10/21/2016 » 10/24/2016
2016 IMC USA Conference

Message from the Chair
Don Matheson CMCChair and CEO, IMC USA
Gregory BrooksExecutive Director, IMC USA