Reach & Qualifywhat the calls actually screened

The report below stops at contact and qualification economics on purpose — that is where call data ends. This estimator translates those economics onto a desk like yours. Change any input. Every default is labelled with where it came from, and nothing here is another customer’s result.

Each figure carries one of three marks. The mark is not a footnote — it is the difference between a number you can put in a client QBR and a number you cannot.

Sparklines run March to August. Both cost lines divide screening-programme spend (₹1,86,264), not total period spend — the reminder programme in section 12 books no new interviews, so folding its cost in here would flatter the screening funnel. Section 11 carries both bases.

Every rate above is dominated by one variable: how recently the candidate chose to be in this process. A fresh applicant answered a form ninety minutes ago and is expecting a call. A record last touched nine months ago may already be working somewhere else. That is what the three stages below separate, and it is why a single blended connect rate is close to meaningless when the stage mix changes.

The cost column divides each stage’s own spend by that stage’s bookings, so the blended cell is the true weighted figure: ₹1,86,264 of screening-programme spend over 2,112 interviews. The post-interview stage books no new interviews — it recovers slots that already existed, at ₹29.85 each (section 12).

Why stage and not role type

Role type is the obvious alternative cut, and it is a real driver — a salaried role takes longer to screen than a shift role, so it costs more per call. But it moves cost through duration, and duration is the smaller lever. Stage moves cost through reach, and reach compounds. Both cuts are in the export; this report leads with the one that explains more of the variance.

Role type spreads cost per interview 1.33×. Stage spreads it 1.77×. Where the two overlap — white-collar records sitting cold in the database — cost per interview runs past ₹150, which is the cell worth acting on first.

Every row is a label a language model assigned by reading the transcript. The duration column is not a house average applied to everything — it is the mean handled time for calls carrying that label, and the rows multiply out to the billed minutes in section 11. A screening that runs the full criteria set takes five minutes; a reminder that is answered and acknowledged takes forty seconds. Averaging those together would misprice both.

Each of these is the first criterion a candidate failed, in the order your script asks them. That ordering matters commercially: a criterion asked late is a criterion paid for in full talk time before it disqualifies anyone.

The two variables a staffing desk actually controls: how fast the first call goes out, and how many times you are willing to try. Both are exact — timestamps, not inference.

Time from application to first dial

Fresh applicants only, n = 4,320. Median 4 min 09 s. The 95 beyond an hour all applied outside the calling window and were dialled at 09:00 the next morning — a policy outcome, not a queue delay. One honest limit: the clock starts when your system posted the application to ours. Any lag upstream of that webhook is invisible to this figure, so treat it as our latency, not the candidate’s experience end to end.

Dials to first right-candidate contact

6,940 of 9,600 candidates (72.3%) were reached at least once, at a mean of 1.85 dials. The tail is the part a manual desk cannot afford to chase and therefore never chases.

The 2,660 never reached cost nothing directly — unconnected attempts are not billed. What they consume is dial capacity and days. 1,208 of them absorbed four or more attempts without a single answer, which is the tracing candidate list, not the redial list.

Connect rate for every hour inside the 09:00–19:00 window this deployment holds. Shift workers are unreachable during a shift and reachable on the way home, which is why the evening block outperforms the morning by a wider margin than a white-collar list would.

Best slot Wednesday 18:00 at 53%; worst Sunday 09:00 at 28%. The 13:00 dip is the lunch break on a shift floor, and it is deep enough to be worth scheduling around rather than through. Reweighting the dial schedule costs nothing and changes no script.

Language is assigned per candidate from your application feed, then confirmed or switched on the call. Where a candidate switches mid-call, the conversation is counted under the language it ended in. For a blue-collar list this is not a courtesy setting — it is the difference between a completed screening and a hang-up.

Bengali sits 4.8 points below Hindi on screening completion, on 560 calls — above the reporting floor, below the threshold where one month’s gap justifies a script rewrite. Section 15 proposes reading twenty transcripts, not changing the prompt.

Nearly one connected call in six did not reach the candidate at all. That is a data problem, not a calling problem, and it is the cheapest thing on this page to fix.

Two bases, stated separately, because mixing them is how a cost-per-hire figure quietly becomes wrong. Total period spend covers everything the deployment did. Screening-programme spend covers only the work that produces new interviews.

Billing basis. Total period spend ₹1,94,892 — 32,482 connected minutes at the ₹6.00 Growth rate. Connected minutes only: ringing, voicemail and unanswered attempts are not billed, which is why section 10 shows no cost against the silent candidates and why no recommendation in section 15 claims a cash saving from dialling less. Average handled time is not assumed — it is built from the per-outcome durations published in section 05, which sum to the billed total above.

Why the two bases differ. On total spend, cost per interview booked reads ₹92.28 rather than ₹88.19 — the gap is the ₹8,628 reminder programme, which recovers existing slots and creates none. Quoting ₹92.28 makes the screening funnel look worse than it is; quoting ₹88.19 while pocketing the reminder saving would make it look better. Both are on the page.

What this is a floor on. ₹88.19 is the floor on cost per placement, not the figure itself. The true number is ₹88.19 divided by whatever share of booked interviews becomes a start. At a 25% interview-to-placement rate that is ₹353; at 12%, ₹735. We do not know that rate on your desk, so we publish neither.

This is the section most likely to be misread, so it is separated out. A reminder call can prove a candidate said they would attend and can prove a slot was moved. It cannot prove anybody walked through a door. Attendance itself arrives here as a flag from your scheduler — your data, passed to us so we can act on it, not a measurement we produced.

Recruitment calling in India is governed differently from debt collection, and the difference matters. The 08:00 to 19:00 window people quote is a lender fair-practices rule for recovery and has nothing to do with hiring. Nor do the telecom rules supply a substitute. TCCCPR 2018 carries no blanket calling-hours prohibition — its time restrictions run through customer preference registration, and the express 9 p.m. to 9 a.m. bar people remember belonged to the superseded 2010 regulation. Separately, a call to a person who submitted an application is service traffic rather than the commercial communication those rules govern, so they are not the operative constraint either way. Which means the window below is policy, not compliance: we hold 09:00–19:00 because a candidate on a shift floor should not be called at 21:00, and because a self-imposed rule that is measured every month is worth more than a statutory one that is not.

Every line is produced from the call record itself; the two audited lines state their sample size.

These are absent by construction, not by omission. Every one of them is an event that happens in your system after the call ends. Any vendor quoting them from call data alone is quoting a model, not a measurement.

Every report ends with recommendations drawn from its own numbers, each tied to the section that produced it. These are the ones the sample month produces — a desk with a different stage mix, role mix or criteria set would get different ones.

A figure you cannot check is a claim, not a measurement. Everything on this page reconciles to a row you can pull yourself.