Candidate Management Software: Tracking People, Not Just Roles

The numbers behind this

What does candidate management software do that a spreadsheet cannot?

It keeps one authoritative record per person and survives more than one editor. Spreadsheets fail in three specific ways. Two people open the file, both save, and one set of edits quietly vanishes. The same applicant appears three times under three spellings, so nobody notices she interviewed last year. And when whoever built it leaves, the meaning of the colour coding walks out with them. Candidate management software fixes those by making the person the primary object rather than the row. Applications, messages, interview notes, files and stage history all hang off a single profile that several people can edit at once, with a change history behind it. That sounds like a small distinction until you try to answer a plain question, such as everyone we have ever spoken to about a backend role in Pune, and discover the answer is scattered across four files and two inboxes.

Why does one profile per person matter so much?

Because duplicates cost you people you already paid to find. When one human occupies three records, each holds a fragment: the rejection from two years ago on one, the strong interview feedback on another, the live application on the third. Anyone reading a single record makes a worse decision than the file collectively supports, and the candidate gets asked the same questions all over again. Deduplication has to work on more than an exact email match, since people change addresses and often apply from a personal account after using a work one. Sensible systems match on a combination of email, phone and name, flag the likely pairs, and let a human confirm the merge instead of doing it silently. At a demo, test the merge behaviour specifically: check whether notes, files, applications and stage history from both records survive it, and whether a wrong merge can be undone.

How do you keep candidate stages and statuses honest?

Define every status in one line and let the system move records for you where it can. Pipelines rot predictably. People sit in screening for six weeks because nobody wants to send the rejection. Active means five different things to five recruiters. A filled role leaves forty people in limbo. Three habits fix most of it. Give each status a written definition including what has to be true to leave it. Set an age limit per stage so anything untouched flags automatically. Close roles properly, meaning every remaining person gets an outcome instead of being abandoned in place. That last one is unglamorous, and it is where reputational damage accumulates quietly, given how many applicants report never hearing back at all. A pipeline you actually trust is also the only condition under which recruitment reporting means anything.

What belongs on a candidate record, and what does not?

Anything a colleague would need to pick the conversation up cold, written as though the candidate might read it one day. That means the source, the roles considered, the questions asked, the evidence behind each score, salary expectations as the candidate stated them, notice period, and the reason for every outcome. Keep it job-related and factual. Impressions about somebody's personality, appearance, family situation or health do not belong there, and in many places collecting some of that carries legal obligations, so take advice on what you store rather than guessing. The practical test is turnover. Recruiters move on and usually take the context with them, so the difference between a database and a filing cabinet is whether the next person can act on what they find. Structured fields beat prose for anything you will filter on later. Prose is right for reasoning.

How do you re-engage past applicants and silver medallists?

Deliberately, and with a genuine reason to make contact. A silver medallist is somebody who reached a late stage and lost to a stronger candidate, which is the closest thing to a pre-qualified shortlist any team owns. Most are never contacted again. Tag them at the moment of rejection, because nobody goes back through old pipelines later, and note precisely what they were close on. When a similar role opens, one short personal message referencing the earlier conversation outperforms a bulk campaign by a distance. The economics become obvious once you price a search properly, and cost per hire makes that argument for you. Two cautions. Ask before adding anybody to a marketing list, since consent rules differ by jurisdiction. And say so plainly if the new role is materially different, because somebody rejected once and then pitched something unsuitable will not reply a third time.

What makes a candidate database still searchable in three years?

Structure at the point of entry, not a cleanup project later. Search fails for boring reasons. The skill is buried in a resume PDF nobody parsed. The tag was freehand, so there are nine spellings of JavaScript. The location field holds Bangalore, Bengaluru and BLR as three separate places. Fix it where the data arrives: parse resumes into structured fields, keep a controlled list for skills and seniority, and let recruiters add freeform tags separately from the controlled ones. Then check what search actually reaches, because plenty of systems index names and job titles but not interview notes or resume text, which is exactly where the useful detail sits. The functional parts of an applicant tracking system vary most on this point. Boolean support, saved searches and the ability to keep a result as a working list turn a database into a sourcing channel you already own.

How does candidate matching work, and where should it stop?

Matching compares a role's requirements against parsed candidate data and returns a ranked list, usually blending keyword overlap with a model trained on earlier hiring patterns. It is genuinely useful for a first pass across a large database, and it is not a decision. Two limits are worth holding onto. The ranking reflects whatever is written on the profile, so a capable person with a thin resume ranks low regardless of ability. And a model trained on who you hired before will reproduce the pattern of who you hired before, which deserves checking rather than trusting. Keep matching to surfacing and leave judgement to people. There is a clean boundary with a neighbouring page here: filtering inbound applications for one live role belongs to resume screening software, while this page covers the standing database of people you already know.

The same candidate, tracked in a spreadsheet and in a candidate record