Target B-School Estimator

Question

Much like the GMAT Score estimator, we are attempting to develop an application that would help with the following common question:

"What schools should I target based on my profile?"

The output is expected to give 1) broad match such as "Ultra-Elites", "Near Elite" etc and 2) a specific match depending on how much detail the user is willing to provide.

Another form of output is probability of a admit.. The program will scan the database for the closest profiles in the database and report the proportion of successes. We need more data for this, but this is an option.

It looks like a fun multiple regression model and we are seeking your advice on what parameters should be chosen as the explanatory variables.

The data would be drawn from the vast database of posts in this forum, Ask Hjort, Ask Accepted.com and MBA Game Plan. I am sure we will need Hjort's help at some point.

The most important question we have is : Is this kind of analysis helpful tool or will this just be more confusing? I hope it will not be too bad:oops: 

I list the usual suspects here, but we will certainly need your feedback.

- GMAT Score 
- GPA ( thanks Oasis.. mea culpa)
- Age
- Years of Experience 
- Years in a Leadership Role (too specific?)

We can go in for nuance also. For example, if you are a career switcher, the model will have a suitable penalty for that. Further, we can include some intangibles such as Quality of Community service ( scale of 1 to 10). 

What we will do is to compare the results from our model to the ACTUAL results posted by the members... This will serve to validate the model. The sample size can be an issue, but we will see.

This is probably something months away from completion, but we will see what we can do.

OasisNYK · Answer

Very cool idea...definitely should include GPA on there - and maybe if you are taking additional coursework?

riverripper · Answer

I am not sure how much help it would be with the ultra elites since it seems a lot of that is random or based heavily on essays.

Here is what I would find the most useful....though it might be too much to do specific schools but it might be great to have one that you enter the information you stated. GMAT, GPA, Age, Years experiences, rank leadership, and extra curriculars. Then your desired field post graduation. Then give a list of 5 schools that meet their goals ranging from a reach school whether its H/S/W or a school in the top 20s...all the way to a safety school for them. Eventually it would be great if you could include factors like desired location, whether work experience is name brand, quality of undergrad, and even where the person is from (we have all heard rumors of certain schools being very hard for Indians to get into).

Probably far to much to ask for unless someone around here is a software engineer with way too much freetime and access to lots of stats for tons and tons of schools...might want to limit it to top 50s or something haha.

GMATcram · Answer

I've seen that nationality plays imporant role on admit. Female is also an advantage when applying to MBA program.

yogeshsheth · Answer

Prat.. If you could get this working. It would be of great help to us 2008 applicants.

Praetorian · Answer

Sure, we can speed this up if we have some help. If you could help us organize the data into a spreadsheet, it would be a great help. Start with the Fall 2007- Comprehensive List. https://www.gmatclub.com/phpbb/viewtopic.php?t=40260 Another useful place would be to glean this data from BW Forums You could have columns for each of the entries as suggested in the list above.

kripalkavi · Answer

Extra curriculars. You need to have a scale for that as well.

aaudetat · Answer

I'm the last person to ask about probabilities and the like. I barely know what regression is. It seems that the intangibles are SO important; how can we predict? I think of the butterfly in Australia causing the tornado in Iowa -- so many variables between Sydney and Cedar Rapids: how can you ever imagine how each will play off the next? I leave it to you all, of course, but coming from a customer service standpoint, I would hate for applicants to over-apply or under-apply based on an unproven estimator.

I think a more basic "basic stats comparison" tool would be an easy place to start. During my search, I made a table of about 10 schools and color-coded it to show how I compared to the school's averages: green for above average (ie my 710 GMAT compared to School X's 680 average), yellow for average, and red for below average. Then I assigned each school a rating of stretch, possible, and likely.

As for the intangibles, I made my own judgement on the relative value of my essays, experiences, interview skills, etc.

Hjort · Answer

Neat idea- I am happy to help out.

A few additional factors:

1. Perceived Quality of Undergrad Institution/Major
2. Graduate School

I find the EASDL "Risks" framework a useful model to have in mind but there are clearly other ways to view admissions.

https://www.gmatclub.com/phpbb/viewtopic.php?t=19450

lepium · Answer

I think this estimation will be much more fallible than the one to estimate GMAT scores. 

I seem to be the quite alone (based on the stats obsession all around us), but I buy into the holistic nature of the application, so I don't think the predictor would be of any help other than to give some ballpark odds (probably aligned with Hjort's cluster system).

Anyway, it's always better to have some indication than nothing at all.

Cheers. L.

Praetorian · Answer

Very helpful info. thank you

Indeed. Thank you for this. aaudetat Good points. Your concern about members using the estimator to make such critical decisions is justified. The estimator will have user input by current admitted students. All admitted students can input data and rank the intangibles on a scale of 1 to 10. So, we are making recommendations based on data. By no means are we saying that this can replace common sense. I had a horrible teacher for Sarcasm 101 in undergrad, so forgive me if you were joking about Regression and Probabilities. What we try to do is to look at say, Admit to Duke (Variable Y, the "response" variable) and input known explanatory variables such as age (X1), GPA(X2) , ...Extra curriculars (Xn). We have data on the explanatory variables. So we can run a regression for each such school based on the data we have. About probabilities, simply compare the total number of students with a user specified profile who applied to Duke (The user can specify a range of AGE, GPA etc). Then go look for similar profiles in the database and determine how many such students were accepted. Divide (acceptances) / (total number of such profiles) gives you the probability. We will also look up recommendations made by Paul Bodine, Linda, MBA Game Plan and try to compare them with our own estimator. This will serve to validate our results. thanks

You are insightful as always. Two more factors to the list . Thanks.

This is getting interesting. Agreed. We want to emphasize the "estimator" part. It is not intended to replace common sense.

PaulBodine · Answer

Interesting idea. More fun than useful in my opinion though because at the top schools it's often the unique mix of factors that makes the difference, and the essays are absolutely critical. So how would your model distinguish between someone with a brilliant profile who executes poorly in the application and the motivated, savvy mediocrity who knows how to exploit the essays to maximize her distinctiveness? It sounds like fun, but unless your database is huge I think the predictive value will be weak or misleading. If you are going to create one then it makes sense to have as many variables in the pot as you can so you can better capture each applicant's profile. Only a minority of posters to Ask Accepted.com provide a really full profile though...

Good luck,

Praetorian · Answer

Paul

Thanks. Your opinion mirrors some of the concerns that others have expressed. One thing for sure is that this is not intended to replace expert or peer advice. This is also not supposed to replace common sense. 

We will think about this some more. In the meantime, if anyone else has any other opinions (feel free to criticize), let us know.

Praetorian

vikramjit_01 · Answer

Seems like you'll need to be careful not to make this thing into something that's too huge to handle. Here's one more opinion, maybe you could have this monster consider a few basic criteria that are easy to work on without going into too much detail. These basic criteria could be GMAT, GPA, Work Ex and age, and maybe a few more. Things like years in leadership role, nationality, gender could be saved for the switch to Monster 2.0

lepium · Answer

But, you see, the years in leadership role is not what matters. It's the quality of the leadership. Or even better, the leadership "acts" performed without the formal authority.

As I said, the project will be interesting as an initial indication of typical profiles but not that much as a personalized indication. 

I'm with Paul on this one, you need to factor the complete profile as a whole, rather than the sum of its parts (stats). And as Paul said, not many people share their complete profile that extensively. Plus the execution will always be an unknown. 

Cheers. L.

vikramjit_01 · Answer

Agreed. Considering this, maybe the estimator could use the information from the basic criteria and tell the applicant how he/she measures up in the basics department and suggest a level at which the applicant's other variables like leadership experience and extra-curricular should be so that the applicant has a good chance of getting in.

So someone with above average GMAT GPA for Harvard would be told,  Given your GMAT of 720... and that you are 24 years old etc etc, your chances of a Harvard admit are fair  (other options being very stong ,strong, weak...). It then would go on to say...  a desirable level of Work Ex for you would be......  and give a few examples of Work Ex that an applicant with such a profile looking to get into Harvard should have. And so on with the other intangible criteria.

If the estimator can't evaluate the intangibles then maybe this is the next best thing to do.

dukes · Answer

I feel like I've seen something like this on one or two other sites (sorry, can't remember where) and felt that they were pretty useless. I think its really tough to capture this stuff in any kind of model. My 2 cents...

paddyboy · Answer

Can we look at this from a different angle?

Let us prepare a database that would have the profile (GMAT, TOEFL, GPA, Age, Stream before joining b-school i.e. IT, finance, non-profit, marketing, armed forces, arts, medicine etc.). A link on the home page to a simple form should suffice.

A person wishing to apply could select from a drop-down list or search the school and view a graph that is, say, Y-axis (% accepted) and X-axis (One or combination of factors mentioned above).

The problem with percentages, of course, is that if one person with a 630 score applied to HBS and got selected, it would show up as a 100%. Need a huge source of data to get the results normalized... :roll:

Target B-School Estimator

Prep Toolkit

Top 5 GMAT Debrief Videos

615 to 715 in 15 Days (Ria's Strategies)

More than Quant/Verbal Abilities: Speed vs. Accuracy

745 with Only Self-Prep and GMAT Club

From 555 to 765: 210-point Improvement

Perfect 805 Score Debrief by Julia

My Rewards