Candidates Are Not Answering Your Calls. Here Is Why, and What Works

The gap between applied and contacted

Ask a recruitment team where candidates are lost and the answer is usually sourcing. Look at the funnel and the biggest drop is almost always between applied and actually spoke to a human.

A candidate who applies to your role on Monday morning has applied to several others in the same sitting. Whoever reaches them first has a conversation with someone who still remembers applying and is still interested. Whoever reaches them on Thursday is talking to somebody who has already interviewed elsewhere, or has stopped looking, or cannot place which role this is.

Nothing about that is a sourcing failure. The applications arrived. They went stale on a list.

Why they do not pick up

Four reasons, and only one of them is about the candidate.

It is an unknown number. Indian mobile users screen aggressively, and with good reason given the volume of spam calls. A number that has never contacted them before, ringing during work hours, is very likely to be ignored.

You called during their working day. A candidate who currently has a job cannot take a recruiter call at 3 pm at their desk. They can at 8 pm, or on Sunday, which is exactly when your team is not calling.

You called days later. Intent decays fast. The half-life of a job application is short and the candidate is not keeping track of who they applied to.

One attempt, then the list moves on. Most candidates who eventually convert do not answer the first call. If your process is one attempt per application, you are discarding people who were interested.

What actually improves the connect rate

In rough order of effect, from experience and from what recruiters report:

Call within minutes, not days. This single change usually does more than everything else combined. A candidate contacted while they still have the job description open is a different conversation from one contacted on Thursday.

Call outside working hours. Evenings and Sundays reach employed candidates, who are the ones you most want. This is an awkward ask of a human team and a non-issue for software.

Try more than once, at different times of day. Three attempts spread across morning, evening and a weekend converts far better than three attempts on consecutive afternoons.

Send a message before or after the call. A WhatsApp naming the role and the company turns your unknown number into a recognised one for the next attempt.

Speak their language. For volume hiring - BPO, field sales, retail, fresher drives - a screening call in Hindi, Telugu, Tamil, Kannada, Malayalam, Marathi or Bengali reaches people an English-only process never gets a straight answer from. Candidates who are uncomfortable in English do not say so; they say yes to everything and then do not turn up.

The screening call is the automatable part

Worth being precise about what should and should not be automated, because the honest boundary matters here.

A first screening call is a short structured conversation with a known outcome: does this person meet the basic criteria. Current and expected CTC, notice period, total and relevant experience, location and willingness to relocate, and one or two role-specific checks. The same questions every time, the answers written down the same way every time.

That is a good fit for a voice agent, and automating it has a second benefit people underestimate: consistency. Every candidate gets asked the same questions in the same order, so your shortlist is comparable rather than a function of which recruiter made the call and how their afternoon was going.

What should not be automated is everything after that. Selling a role to a candidate who is on the fence, negotiating an offer, managing a counter-offer, judging whether someone will fit a team - these are the parts of the job that justify a recruiter, and a voice agent that attempts them will lose you good candidates.

Interview no-shows, which are their own problem

A booked interview that does not happen wastes a recruiter's time and a hiring manager's, and hiring managers remember it.

Most no-shows are not people changing their minds silently. They are people who forgot, who never had the details in writing, who could not find the location, or who accepted another offer and felt awkward telling you. Each has a fix:

• Confirm in writing immediately, on WhatsApp, while the call is still happening - role, time, place, contact person, map pin.
• Remind the day before, with a way to say no. A candidate given an easy way to decline will decline instead of vanishing, and that is a much better outcome - you get the slot back.
• Reconfirm on the morning. Most drop-outs are recoverable if you find out at 9 am rather than at 11.

None of this is sophisticated. It is just work that nobody has time to do by hand at volume, which is precisely the kind of work worth automating.

What to measure

Four numbers. If you are not tracking these, you cannot tell whether anything you change is working.

• Time from application to first contact attempt. Measure in minutes, not days.
• Connect rate - attempts that reached a live conversation - split by time of day.
• Screening-to-interview rate, which tells you whether your questions are filtering usefully or just filtering.
• Interview show-up rate, before and after you add confirmations and reminders.

Start with time-to-first-contact. In most teams it is the worst number and the easiest to improve, and improving it makes every downstream number look better.

When not to automate this

If you hire a handful of senior people a year, ignore all of the above. At that volume every conversation is a judgement call from the first minute, your candidates expect a human, and a screening bot would actively damage your reputation in a small market.

This is a volume tool. It earns its keep for agencies running bulk hiring, BPO and field-sales recruitment, fresher drives, and staffing desks where applications arrive faster than anybody can call them back. If applications are not piling up unanswered, you do not have the problem this solves.