RE: Assignment 2- Executive Director Foundation for Sustainable Agriculture

Dear Akshay,

You got me stumped !

But that was expected , considering that the last time I opened a book on Statistics , was 60 years ago !

In any case , Mitchelle and other colleagues are working on your suggestions and expect to have some results available by Monday evening

I wonder whether you can visit us on Tuesday morning ( 10 am ? ) , when we can share with you what we find,

And , I can share with you my notes / thoughts on Resume Parsing

Incidentally , I sent to Rohit , our source code for www.RecruitGuru.com , which was the Resume Parsing web site that we had launched some 13 years back

Did you get a chance to look at this ?

Warm Regards,

Hemen Parekh

+91 98675 50808

3P Consultants \ Strategic Human Resources Consultants

From: Akshay Sehgal [mailto:akshay.s@ipredictt.com]
Sent: 05 May 2017 20:47
To: Rohit Verma
Cc: Hemen Parekh; nirmit@3pconsultants.co.in; Mitchelle; Prashant Sharma; Manoj Pandey; aneree92@gmail.com
Subject: Re: Assignment 2- Executive Director Foundation for Sustainable Agriculture

Hi Hemen,

Thanks for the feedback and ideas. I am quite interested in understanding the method you talked about for increasing parsing accuracy. As it so happens our team is working on the same problem using NLP and should be releasing a better version of the parser soon.

Coming back to the correlation analysis between POD (probability of decline) and Employability score, I don't think utilizing 2 probabilistic KPIs to validate each other is a good idea. Following are my thoughts on the same - 

1. POD, which is still a beta KPI, only red flags are to be considered with a certain degree of significance parameters. In simple terms, there is a certainly acceptable bandwidth under which variation in POD values are not significant, however extreme POD values which lie outside this bandwidth should be of concern to the HR.

2. The features used in employability score are from historical behavior derived from the facts about the candidate. While POD is supposed to be a non-predictive essence of mismatch between candidates' opinion and job offer details (mismatch in CTC range and expectation, etc). It depends highly upon post resume data capture, where chatbot comes into play.

3. A correlation analysis between the 2 may give some insights but if used together as 2 separate KPIs let you validate both candidate applicability to the job at hand as well as cautionary feedback towards investing resources in chasing the candidate.

4. To put it simply, the underlying variables that are used to calculate the 2 variables will require having some amount of correlation to generate correlated KPIs. We ensure same variables are not used in both the KPIs so that both of them are relevant. A small example is that if 2 candidates are equally skilled, experienced and educated, their employability score will be the same. However, due to a mismatch in location preferences, CTC expectations, the frequency of job change and other factors, the POD of both candidates may be drastically different.

Moreover, the employability score that you see are using an unsupervised model which is primarily used for pattern recognition without any response variable. If we can get response variables, however, the second model will be supervised and the accuracy of that model is well defined. Unsupervised models require actual testing to check the validity of the model.

Let's discuss these in more detail soon.

Thanks,

Akshay.

Co-Founder, Product Strategy & Research

+91-991-659-4778 | Skype : akshaysehgal2005 | http://ipredictt.com/

On Fri, May 5, 2017 at 12:46 PM, Rohit Verma <rohit.v@ipredictt.com> wrote:

Thanks for the detailed input.

Including Akshay in this conversation to update you with our thought processes about KPIs and co relations.

Regards

Rohit

From: Hemen Parekh <hcp@recruitguru.com>
Date: Friday, 5 May 2017 at 10:23 AM
To: <rohit.v@ipredictt.com>
Cc: <nirmit@3pconsultants.co.in>, Mitchelle <mitchelle@3pconsultants.co.in>, Prashant Sharma <prashant@3pconsultants.co.in>, Manoj Pandey <contact@3pconsultants.co.in>, <hcp@recruitguru.com>, <aneree92@gmail.com>
Subject: FW: FW: Assignment 2- Executive Director Foundation for Sustainable Agriculture

Rohit,

After getting your results ( of “ Employability Score “ and “ Probability of Declining “ ), I wondered :

What could be the “ Correlation Coefficient “ between these two sets of numbers ?

Could such correlation , possibly reveal :

·         Higher the Employability Score , higher the Probability of Declining ? or vice versa ?

My grand daughter Aneree helped to answer this question with her attached statistical analysis . Thanks , Aneree !

But , this current analysis , may not be conclusive , considering that our dataset was small .

Another question :

Does such analysis ( to find any hidden pattern ) have to be confined to EACH mandate / assignment , separately or could it be done for all 1275 mandates , together ?

I believe , we need to do this separately for each assignment

In the first instance , what is however much more important from 3P’s needs is to find out :

“ To what extent the software was able to mimic the human brain , in selecting the best candidate , for a given assignment ? “

I suggest , Prashant re-arrange the shortlisted ( 50 ? ) candidates in the descending order of the “ Employability Score “ – then locate where ( in which percentile ) is that BEST ( meaning , selected / appointed by the client ) candidate is located

If there is a HIGH DEGREE of correlation between the “ Algorithmic Score “ and the “ Nebulous Score ( of the client ) “ , then it is a proof that iPredictt is on the right track !

Of course , if we wish to establish such a PROOF beyond any doubt , then we need to carry out this exercise for ALL 1275 mandates and establish a Correlation between the two !

As far as extraction of “ data fields “ from the text resume and creation of a structured / searchable , database is concerned , I find a few data missed out during the parsing process

I have a solution for improving the accuracy of parsing , which I can discuss with you / your software guys . I can visit your office if desired

Warm Regards,

Hemen Parekh

+91 98675 50808

3P Consultants \ Strategic Human Resources Consultants

From: Aneree Parekh [mailto:aneree92@gmail.com]
Sent: 05 May 2017 09:50
To: Hemen Parekh
Subject: Re: FW: Assignment 2- Executive Director Foundation for Sustainable Agriculture

Hi dadu,

Attached is the output of the Pearson's Correlation r. There appears to be no sig. correlation between Employability Scores and Probability of Decining (r = .087, n =  42, p = .585)

On Thu, May 4, 2017 at 8:53 PM, Hemen Parekh <hcp@recruitguru.com> wrote:

Warm Regards,

Hemen Parekh

+91 98675 50808

3P Consultants \ Strategic Human Resources Consultants

From: Rohit Verma [mailto:rohit.v@ipredictt.com]
Sent: 04 May 2017 18:12
To: Mitchelle
Cc: Hemen Parekh; Nirmit Parekh; Prashant Sharma
Subject: Re: Assignment 2- Executive Director Foundation for Sustainable Agriculture

Hi Mitchelle,

We have created a 3P Consultant account for you on the platform.

Please login to enterprise.careerletics.com

UserId: mitchelle@3pconsultants.co.in

Pwd:    3p@789

Following are the advised steps to experience the features –

1.       Review the JD (You can edit or add information  - like Skill sets)

2.       Review the PARSED resumes (You can edit or add information)

3.       Open the dash to see the trends

4.       Download Excel sheet for quick reference (attached)

5.       Review the top listed resumes to decide on next steps

6.       Provide us feedback

We currently ran the parser to collect Resume information and also ran various algorithms to get results for Employability score., Probability of Decline .

Please study the accuracy and let us know patterns you see.

Also share details of selected candidates you have originally have for this role. The funnel should be

1.       Candidates selected for interview

2.       Candidates made offers

3.       Candidates hired

Further, Our Data Science team will work on this to improve results which we will discuss as we progress.

Should you have any issue or need clarity, please call me

Regards,

Rohit

From: Mitchelle <mitchelle@3pconsultants.co.in>
Date: Wednesday, 3 May 2017 at 12:36 PM
To: <rohit.v@ipredictt.com>
Cc: Hemen Parekh <hcp@recruitguru.com>, Nirmit Parekh <nirmit@3pconsultants.co.in>, Prashant Sharma <prashant@3pconsultants.co.in>
Subject: RE: Assignment 2- Executive Director Foundation for Sustainable Agriculture

Dear Rohit,

Many thanks for your response.

As agreed ; we shared with you two descriptive JD’s along with evaluation criteria at our end on the basis of which candidates were shortlisted and resumes (50 batch size/assignment).

Basis these documents shared; we expected the iPredict software to accurately pars these resumes and derive an accurate shortlisting with an employability score.

The objective of this exercise was to then match the candidate shortlist derived out of the iPredict automated algorithm along with an Employability score vis a vis the manual shortlisting done at our end.

This would help us correlate the effectiveness and shortlisting accuracy along with the employability score of the algorithm vis vis manual shortlisting.

We would thus request you to share with us the same for each of the two Assignment details shared with you.

Do also indicate a timeline for us to receive this.

Warm Regards,

Mitchelle

+91 98675 50810

3P Consultants \ Strategic Human Resurces Consultants

From: Rohit Verma [mailto:rohit.v@ipredictt.com]
Sent: Wednesday, May 3, 2017 10:55 AM
To: Mitchelle <mitchelle@3pconsultants.co.in>
Cc: Hemen Parekh <hcp@recruitguru.com>; Nirmit Parekh <nirmit@3pconsultants.co.in>; Prashant Sharma <prashant@3pconsultants.co.in>; Manoj Pandey <contact@3pconsultants.co.in>; Akshay Sehgal <akshay.s@ipredictt.com>
Subject: Re: Assignment 2- Executive Director Foundation for Sustainable Agriculture

Thanks, all for the time and patient hearing.

Hope some of the discussion points intrigued you and I hope we work together in the future to create value in not only recruitment space but also in areas of descriptive dashboards, performance, attrition analytics for companies who you have built relationships with.

As email ID is the key resume identifier, we are tweaking our processes to include this change. Will share the results once done.

My responses in line

Accurate Parsing based on which Accuracy in shortlisting at our end with minimal manual intervention / subjectivity – Currently available

Employability score – Available , needs data for further machine training.

Decline Probability – Available , needs data for further machine training.

How would these algorithms assess Cultural fit – Can be done on case to case basis by accessing the social media data (Facebook, Twitter, Linkedin etc) – refer to the attachment. Can be done on case to case basis

What additional trends / analytics can you predict based on the data input – This depends on the questions you have. For instance, if you are interested in looking at next job trends, can be understood by social media tracking.

Also we would like to see the algorithm extrapolate and show us a comparative table /ranking in terms of relevance of candidate profiles shared to the Data points / Job criteria – Will be mailed

We have intentionally not shared the candidate contact details – as we do not want to activate the chatbot tool at this stage to maintain confidentiality – Sure, but Chatbot only works if manually activated for each of resumes. Its not automated.

Any accurate inferences on ability to predict attrition levels / and when will which position come up for hiring (replacement / new), what kind of hiring trends / industry / function wise that are seeing traction etc.,- Probability of decline is an indicator to that respect/

Regards

Rohit

From: Mitchelle <mitchelle@3pconsultants.co.in>
Date: Tuesday, 2 May 2017 at 4:33 PM
To: <rohit.v@ipredictt.com>
Cc: Hemen Parekh <hcp@recruitguru.com>, Nirmit Parekh <nirmit@3pconsultants.co.in>, Prashant Sharma <prashant@3pconsultants.co.in>, Manoj Pandey <contact@3pconsultants.co.in>
Subject: Assignment 2- Executive Director Foundation for Sustainable Agriculture

Dear Rohit,

Many thanks for the interactive session this morning.

As agreed, enclosed please find the documents for your reference :

Position Specification –Executive Director for the Foundation for Sustainable Agriculture

Evaluation Criteria – Data Points comprising Personal and Technical attributes which formed the basis of our evaluation/elimination criteria

Candidate Profiles sourced for this position

We would be keen to view the following based on the data shared :

1.       Accurate Parsing based on which Accuracy in shortlisting at our end with minimal manual intervention / subjectivity

2.       Employability score

3.       Decline Probability

4.       How would these algorithms assess Cultural fit

5.       What additional trends / analytics can you predict based on the data input

6.       Also we would like to see the algorithm extrapolate and show us a comparative table /ranking in terms of relevance of candidate profiles shared to the Data points / Job criteria

7.       We have intentionally not shared the candidate contact details – as we do not want to activate the chatbot tool at this stage to maintain confidentiality

8.       Any accurate inferences on ability to predict attrition levels / and when will which position come up for hiring (replacement / new), what kind of hiring trends / industry / function wise that are seeing traction etc.,

Do feel free to speak to me in case of any additional inputs required.

Warm Regards,

Mitchelle

+91 98675 50810

3P Consultants \ Strategic Human Resurces Consultants