How to frustrate an engineer

[Anonymous]

Just tell him that he has to get the highest possible result by plugging both controllable and uncontrollable (historical) data points in to a formula that he is not allowed to see the equation for.  I need a coefficient set.  I get the need for keeping proprietary data proprietary, but it is more than annoying to chase a moving target that you can never get a good look at.

[takeshi74]

I can understand those who feel the need to have a complete understanding but it is what it is and we're stuck with what little we have to work with.  We are not FICO's customers -- the creditors are.  We've just been able to get better access to credit scores since it impacts us as well.

Maybe it's because my nature is different but while I can't say that I can calculate a score for different models out there just by looking at a report I'd say that one can get a pretty good feel for assessing a report with some understanding of what information is available.  I'm no engineer but I do work in IT and my work constantly involes identifying causalities with limited information and ruling out coincidences so I can determine the cause(s) of problems and address them.  There's quite a bit that can be deduced by playing with inputs and monitoring outputs for black boxes whether a given black box is software, scoring models or anything else.  Granted, quite often I'm wrong as there are pitfalls to that approach but I'm right enough of the time to do what I do very well as a career.  As for credit stuff, I'm definitely still learning even though I think I am getting a good feel (I know engineers generally can't stand relying on "feel") for it.

These are only typical weights but it's a good resource if you haven't already taken a look.  It is, however, just a starting point.

http://www.myfico.com/crediteducation/whatsinyourscore.aspx

You may also want to visit the Understanding FICO Scoring subforum.

Even if you did understand all scoring models and could determine what scores would be that's still an incomplete picture.  It's never just about score.  A score is just one tool that a creditor uses in making the decision.  The creditor/product has underwriting criteria that will never be fully disclosed and one's entire credit report is considered along with income and possibly other factors in certain situations with all of that compared to the underwriting criteria.

Don't assume that you have to have the highest possible result.  For example, with FICO 8 best terms are generally offered around the 740-760 range.  Anything above that is gravy.

[BallBounces]

---

@Anonymous wrote:

Just tell him that he has to get the highest possible result by plugging both controllable and uncontrollable (historical) data points in to a formula that he is not allowed to see the equation for.  I need a coefficient set.  I get the need for keeping proprietary data proprietary, but it is more than annoying to chase a moving target that you can never get a good look at.

---

Let me solve your problem for you then, Textret:    You don't have to get the highest possible result.

I love solving problems for you engineer guys.   On to the next one! 

[Thomas_Thumb]

Ok - Let's not generalize with respect to engineers.

Some want to know how things work, some care mainly about the end result and others just don't give a damn anymore. 

[RobertEG]

It is not simply a matter of wanting to keep things secret, it is a business requirment that they do so.

After developing a scoring algorithm, any party has only one year after putting it into use in which to apply for a patent.  That is a requirment of U.S. patent law.

Applying for a patent requires full public disclosure of how to make and use the invention, and thus after the non-renewable patent expires at approx 20 years, it becomes public doman for anyone to use.  Some companies, such as Coca-Cola, hedge that they can keep it secret longer than 20 years, so choose to keep it a trade secret rather than disclose it now via applying for a patent.

If they choose not to apply for a patent within the one year period after first using it, they lose the right forever, and must then keep it a trade secret or it will become useless (from a money making standpoint) it it leaks out.

They cannot thereaafter release information to the extent that it would permit reverse engineering of the algorithm.

Thus, keeping the black box closed is mandatory.  The existence of their business depends upon it.

I too am an engineer, but am not too frustrated by not knowing the formula, only the results.