Notes: ADMIN TOOL FOR WWJ

Airtel ADMIN TOOL

Month	No. of Active Subscriber (Cumul.)	No. of Registrations	No. of Regn & SMS	No. of Job Alerts sent	No. of "Apply" SMS
Feb	-	51	-	-	-
Mar	-	7	-	-	-
Apr	-	-	-	-	-
May	8	33	18	46	41
June	20	67	63	-	32
July	20	2	5	459	9
Aug	22	4	11	4330	11
Sept	23	33	6	4982	1
Oct (upto)	23	12	-	755	-
Total	23	209	103	10,572	94

Conclusions:

fracApplyJob Alerts = frac{94{10572 approx 0.009

Definitions of

No. Active Subscriber to

No. of "Registrations" to

(Note in margin: P.t. Redesign to reflect current 4-10-05)

Abhi to Rajeev to Rejji

24-08-05

How many job-aduts are needed to send 1 job-alert?

For Aug. 2005, See enclosed tabulation and corresponding graph. We could not download any jobs on 5^th & again today. Readings for 17^th & 18^th are FREAK - obviously something terribly went wrong! (also no record for 1^st Aug!).

One can easily calculate sigma (standard deviation) for the remaining 20 readings and then plot a normal curve, with

to Mean + 3sigma =Upper Control LimitUCL

toMean - 3sigma =Lower " "LCL

begin{array{l+3sigma Mean -3sigma end{array

[Sketch of a Control Chart with UCL, Mean, LCL lines. The x-axis is labeled "DATE to" and the y-axis is labeled with the sigma values. The chart is labeled "CONTROL CHART".]

If the process is statistically stable, then each day's RATIO shall fall, somewhere betweenUCL &LCL.

(Note in margin: 1/3)

I am sure STANDARD statistical s/w packages (to calculate Mean / std deviation of a given set of "Numbers") are available FREE from Net. Such a package can automatically Generate the CONTROL CHART and display in our ADMIN Tool.

As more & more "readings/records" (of ratio) accumulates, the software will also automatically revise the "Mean (upward or downward) - and also continuously/dynamically re-calculate the standard deviations. These are all, very established techniques. Someday (when we are a little free from other priorities), we should implement this.

But this ratio does not depend upon "No-of jobs downloaded" only.

It equally depends upon "No-of Active Registrations".

So, instead of just 306 registrations, If this no. goes upto 3060, then, even with daily download of 2000 job-adits, we may be able to generate/send out 2000 job-alerts, giving us a ratio of frac{2000{2000 = mathbf{1.0

To bring-in the effect / impact of "No-of subscribers" may be, we ought to calculate / plot

{No. of Subscribers times frac{B{A =Box

AUGUST 2005

Ratio = fracJob AlertsJob Aduts

How Many Job-Aduts Are Needed to Send 1 Job Alert?

[Graph showing a time series plot of a ratio vs. Date, labeled 1 to 25 on the x-axis (Date to) and 0 to 15 on the y-axis (Ratio).]

Data points shown on the graph (Date, Ratio):

Date	Ratio	Date	Ratio	Date	Ratio	Date	Ratio
2	6	8	3.7	13	8.8	21	6.5
3	8.3	9	5.6	14	7.5	22	4.4
4	-	10	5.5	15	4	23	4.3
5	9.5	11	10	16	2.6	24	-
6	11.4	12	8.6	20	8.9	25	-