In this issue, I have marked:
- In Red ✓adverts of Public Ltd. Companies
- In Greennames of Recruitment/Placement Companies who have advertised on behalf of their clients
I presume all of these adverts are on naukri.com website, so Sajida can easily download directly those which I have marked in Red.
Of course, she will need to “convert” these using “Advt. Convert” software given to her. And when she does:
→ Actual advertiser’s name/contact data will get hidden.
→ Our email ID (Apply@3pjobs.com) will replace original email ID.
She must, of course, send to us one copy of each advert with FULL/ORIGINAL content.
When we have compiled 100,000 such job-adverts (over next 12 months?), then using JOBMINE software (almost same as our Advt. Convert with a few small changes), we would process/datamine these 1 lakh job-adverts, to uncover/discover following patterns/trends which would help us to carry out an aggressive/pro-active…MARKETING DRIVE:
Industry-wise frequency distribution
Which are hot/sunrise sectors?
Which industries have maximum demand of executives?
Function-wise frequency distribution
Which “functions” are most sought after?
Age-wise frequency distribution
What is the “prime” age-group?
Designation-Level wise frequency distribution
At what “level” most recruitment is taking place?
Edu-level/branch wise distribution
What educational qualifications are in greatest demand?
City-wise distribution
Which cities/regions have maximum job-offerings?
Industry wise
Function wise
Desig.-Level wise
We will do the same with “keywords” (as opposed to “key-phrases” & “key-sentences”).
These analyses would help us in automatic “matchmaking” of resumes which also contain →
Same keywords / phrases / sentences (as found in job-advt).
Then you don’t need a consultant to manually enter “SEARCH PARAMETERS” into our “Resumearch” & then wait for results to appear & then,
One by one, open each resume & read it to decide how well that resume “matches” the UN-EXPRESSED criteria specified by the client — but which cannot be entered as “Search-Parameters”!
(Two Bell-curve graphs drawn — first showing population of resumes = 95,000 in SALES, second showing Job Ad population = 15,623 in SALES drafted by VOLTAS)
Could it so happen that if we (JOBMINE) process 100,000 job-advt., we could see a frequency distribution such as following emerge?
If VOLTAS recruitment manager “composes” a job-advt. using our WEBSERVICE (Advt.-Compose Tool) & suddenly sees above-mentioned GRAPH emerge in front of his eyes, he would know (although he may not admit it!), that he has done a LOUSY JOB in drafting/composing the advt!
Now, from the dropdown LIST-BOX, provided by us, he can choose/pick…
Some more
Keywords
Key phrases
Key sentences
… pertaining to Sales function, and try again.
Now, he sees, “revised” graph as follows:
(Drawn graph with Job Advt. No, Drafted by VOLTAS)
This would be a tremendous Decision Support System
Based on “Datamining” of 1 lakh job-advt. (as it will go on improving as we JOBMINE more & more).
If we can pull this off, recruitment would never be the same again!
!!
Remember, what we can do, others can do even better! – but “first mover” has better chances of survival.
Abhi / Deepa
Job Advt. Datamining (Job-Mining)
Creation of a Job Advt. Database
You should also borrow from me and read folders/notes on:
JAWS
JobAlert as SMS, etc.
But these folders need not be read right now, because, in enclosed notes, I have briefly covered these Concepts/Usage/Applications.
We can/will take-up for development each of these “applications” (one-by-one).
Auto-Converter
In course of time. But the first step is to modify/redesign “Auto-Converter” in such a manner that, it is capable of:
handling/processing job-adverts (digital) received from any source and in any format
doing this “processing” automatically & without any human intervention
creating a structured database of job-adverts, using which, we can launch a series of “applications”
Whereas, it is desirable that such a modified Auto-Converter is capable of processing job-adverts from any source and in any format, the danger is that such a universal tool may run too slow, trying to search for all possible permutations/combinations.
If you feel the same way, you may consider developing 6 different tools, each exclusively meant for processing job-adverts from one/given source only. This can speed-up processing.
(Signature)
-
citywise
-
functionwise
-
industrywise
-
education-wise
-
age-wise etc.
→ Clearly define, the JOB-MARKET in India.
Now, who/what “supplies” executives? (Supply-chain!)
For fresh/entry level, these are Colleges/Training Institutions (NIIT, APTECH etc.) – Instt. of Chartered Acct / Instt. of Cost & Works Accountant etc. etc.
At Experienced Executive Levels, the “Source of Supply” is Thousands of Companies (where these executives are working).
Obviously, we cannot email specific job details to Corporates (like most jobsites – it would be too obvious – in respect of our “hidden intentions”!! (viz: to get their employees to submit their resumes).
BUT, if Abhi can produce such HR Reports (once-a-month), Inder can cleverly push (email) such GRAPHS to Corporate HR Mgrs with an “educative” covering letter.
Let us do.
Abhi – Kartavya – Vicky – Inder – Deepa
Job Adverts. Summary is quite like a TV Guide
At micro-level, individual jobseekers are interested to know of jobs having specific “Position Names/Actual designations”, so they can apply.
At macro-level, individuals are interested to know:
-
Which Industry is advertising most (or least)
-
What Functions are in great demand
-
How many jobs are there at various Desig. Levels
-
In which Cities are most openings
-
What Edu. Qualifications are in demand etc. etc.
Then, we will make this content available to one top-ranking media in each category viz: Newspapers / Magazines / TV / Jobsite etc.
That should give us a lot of PUBLICITY.
Job Advt Analysis Tabulation
Function:
| Advt. No | Position Advertised Name | Jobsite Uploaded on | Resumes Received (Descending order) | Cumulative (No. of Resumes / % of Resumes) |
|---|
Abhi → Refining "Project Manhattan"
-
During last 6/8 months, we must have uploaded some 3000/4000 job-adverts on different jobsites (Monster / JobsAhead / JobsDB) and received over 40,000 email resumes.
-
It is now high time to carefully analyse what has been happening before (blindly) continuing with Project Manhattan.
Such an analysis becomes all-the-more important/urgent, considering that we are:
A very quick (coarse?) analysis of data you gave me (for last 15 days), shows that:
• Ave. Resume/job-advt = 5.3• Highest No. of resumes/advt = 392!• A mere 4% of advt (60 advt) yielded a whopping 79.5% of all resumes recd. (5887)
For this group, ave. resumes/advt works out to:
Approx. 100! (Pretty Good).
-
We need to do similar analysis for all the 4000 job-adverts & all the 40,000 resumes received during last 6 months (ever since we launched Project Manhattan) & do it FAST!
-
We should also analyze this entire database (4000 advts / 40,000 resumes),
• Industry-wise• Function-wise• Position Name-wise• Jobsite-wise (from where resumes came)
-
It is unimportant as to from where we had “downloaded” a particular advt.
Since we are not revealing the “Name” of the advertiser, there is no question of that influencing the quantity/quality of response.
The underlying assumption is that, even on a given jobsite, there is no uniform distribution of resumes amongst all the functions.
By conducting some very simple statistical analysis (A/B/C analysis = 80-20 ratio etc.), we would be in a position to discover PATTERNS / PROBABILITIES.
-
With help of these emerging “patterns,” we would be able to predict (with reasonable accuracy),• Expected Response (No. of resumes that are likely to be received) from each website, for each function, if we were to upload a given job-advt on that website.This is merely an “extrapolation” of past into future.
-
A tabulation such as following will emerge:
Function: Mktg.
| Sample Size / No. of Adverts | No. of Resumes |
|---|
Jobsite on which advt is uploaded
| Jobsite | Total Advts | Total Resumes | Probabilities of Getting (5 / 25 / 50 / 100 resumes) |
|---|---|---|---|
| Monster | 0.90 / 0.70 / 0.60 / 0.50 | ||
| Naukri | 0.55 / 0.50 / 0.40 / 0.40 | ||
| Jobs Ahead | 0.30 / 0.25 / 0.20 / 0.10 | ||
| Jobs DB | 0.10 / 0.08 / 0.07 / 0.05 | ||
| Grand Total |
Such tabulations have to be constructed (from existing 4000 advts / 40,000 resumes recd) for each Function!
Of course, as more & more job-adverts get uploaded daily & more & more resumes keep arriving daily, the software should automatically re-calculate the “Probabilities” & re-populate the tables.
-
Now it is easy to develop “Decision-Rules” which software will automatically apply (to decide on which jobsite to upload a given job-advt), the moment Sayida downloads a job-advt.
In fact, the software will, on its own & without human intervention, actually upload each downloaded job-advt on the BEST / MOST APPROPRIATE jobsite!
We must make Project Manhattan to “Graduate” to a “Pin-point / Targetted / Laser-guided bomb.”
How soon?
Analysis of Resumes Recd. Between 15/04 & 30/04/03
| No. of Job Adverts in the Slab | No. of Resumes in the Slab | No. of Resumes in this Slab | Cumulative No. of Job Advt | Cumulative No. of Resumes | % |
|---|---|---|---|---|---|
| 3 | 1131 | 3 | 1131 | ||
| 11 | 1991 | 14 | 3122 | ||
| 16 | 1141 | 30 | 4263 | 57.6 | |
| 13 | 915 | 43 | 5178 | 69.9 | |
| 17 | 709 | 60 | 5887 | 79.5 | |
| 22 | 563 | 82 | 6450 | 87.1 | |
| 36 | 424 | 118 | 6874 | 92.9 | |
| 64 | 297 | 182 | 7171 | 96.9 | |
| 137 | 176 | 319 | 7347 | 99.2 | |
| 1076 | 55 | 1395 | 7402 | 100.0 |
Total: 1395 job-adverts, 7402 resumes
This analysis shows that just 4% of the total job-adverts uploaded (i.e., 60 advts) produced nearly 80% of the resumes (i.e., 5887).
That is, 96% of our time/money/effort was WASTE!
So, what we need to figure out is:
-
What Industries / Functions / Design levels did these 60 job-adverts belong to?
-
Which jobsite gave best response against which job-advt?
No comments:
Post a Comment