# My thoughts on Quantrix

I have finished importing my Improv model under Quantrix, and have played with it a little (just a little, as you will see). Here are my thoughts :

The Quantrix interface is definitely better. There is no single, big improvement, but a lot of small things that add up :

– pop down menu to choose a single element in a category (when you are not displaying the whole category)

– a undo that works

– error messages are more explicit

– you can save/load using windows name rather than DOS ones

– model browser shows which matrix have unresolved problems

– a more modern look

– each formula displays not only those that it overrides, but also those by which it is overriden.

On the other hand, I miss the automatic display of the formula that calculates a cell when clicking on that cell.

The functions are better. Recursives ones in particular, but I also noted that sumif was far better than its IMPROV equivalents…

I am not too sure about your “formula overrides cell data” policy. IMPROV did things the other way, and while both have their drawbacks, I think I liked it better. Most noticeably, I could use general formulas to fill in missing data, or input a general trend, then manually input available data where it belonged.

For example, when using a prospective mortality table (mortality rates by age and year), I could write some general formulas to fill in missing years (year[THIS] = year[PREV]) or ages (age[THIS]=age[PREV]*1,1), and put in the data I had, and it worked. Now I have to think about exactly which data I’m missing, and put formulas specifically targeted at this data…

Or I might put in a general formula for average salary progression (say, +1,6% per year), and manually fill in the (known) past values…

Is their any way to reverse the priority in Quantrix ?

My biggest problem is the time it takes to (re)calculate the model, and the fact that this is blocking (I cannot do anything with it until it’s over, and often nothing else on my computer either).

My big Improv model used to load in under a minute, and recalculate in 20-30 seconds if I changed something that affected everything else (say, a mortality table).

Under Quantrix, it took 1h30 to import from XML after I did the modif there, and if I changed the mortality table it takes 30 minutes to recalculate. I need to rewrite them to account for the new priorities between data and formulas, but I can’t do that if it freezes for 30 minutes every time I do a little change.

Do you think you can do something about those times ? I really like the interface, but I need those big models, too. I think lookup calls might be part of the problem, I must have about 100 of them in my formulas, calculating 10000’s of cells. Formulas without them seem to go much faster. I might be able to trim their number a little using the better recursive functions in Quantrix, but I do need them…

My company has been looking for a successor to Improv for some time. It has some serious problems (most noticeably, it cannot save models above a certain size without corrupting them), and an older UI. Quantrix might be it, if it runs those big models with acceptable speed, and can go beyond Improv’s limitations on size…

Just a quick update : I have tested my model with the new alpha version, its performance is much better.

Loading time is down to 30 seconds, recalculating time to 2-3 minutes (for a complete recalculation). This is on a P3 800 with 386M RAM. And even better : Quantrix now lets you keep working while it recalculates, so if you have several things to change, you can change them all without waiting between each (or you could just turn recalculate off until you’re finished, of course).

Just to give an update to the forum community, here are a couple of specific performance enhancements that will be included on the next release.

* Quantrix Models will not go through an entire model calculation on open, unless a change has been detected in the structure of the file since the last save. This will dramatically increase the speed of opening models and the entire model file will not have to be calculated each time it is opened.

* In this particular model the lookup function is used heavily. We are currently optimizing the lookup function in the Quantrix calculation engine so it will be more efficient in its calculation. The same work has already been completed with some other intensive functions, including the RANK and the SELECT series of functions.

We are very pleased with the early results of our efforts. The initial benchmark testing is showing dramatic improvements in the application performance. We should be able to post in the forums some benchmark testing results when the optimization work is completed.

-Mike

Hello Emmanuel,

Thank you very much for your comments and feedback. It is always very gratifying to find previous users of Improv and to garner their perspectives on Quantrix Modeler.

In regards to performance, please know in our next release performance is a top priority – particularly with large models. We are pleased with the results of our early code changes and testing and we think users will see a significant performance improvement with the next release.

That being said, your extensive use of lookup in your model is something we would like to analyze and test our changes against. I will be contacting you off-forum to see if getting a copy of your file is a possibility.

Also good comments regarding “formula overrides cell data” policy. Both methods have advantages and drawbacks. Possibly, effective uses of the IN and SKIP statements could be of use to you. When I model, I will often group sets of items that I want to have formula control over, and leave the items I don’t want formula control over un-grouped. Then I will use the IN statement in my formula to restrict the formula only to the group, leaving the rest of the cells available for data input. Could be something to consider… I would love to hear other thoughts on this subject from other forum visitors.

Thanks again for the great comments and feedback!

-Mike