Applicant Tracking System Features: How an ATS Actually Works

The numbers behind this

How does an applicant tracking system work, step by step?

The flow runs in eight moves. A manager raises a requisition and it gets approved. The job publishes to a careers page and syndicates out to boards and aggregators. Applications arrive and land in one queue. Resumes are parsed into structured fields so they become searchable. A screening layer ranks or filters them against the requirements. Survivors move along named pipeline stages, each one a decision point rather than a folder. Interviews get scheduled, and interviewers file structured scorecards instead of hallway opinions. An offer is generated, approved, sent, and signed, while everyone else is closed out with a recorded reason. Reporting then reads that whole trail: where people came from, where they dropped, how long each step took. The Jobscan 2025 ATS Usage Report puts adoption at 97.8% of Fortune 500 companies, which is why applying to large employers feels so uniform.

How do resume parsing and AI matching actually work?

Parsing converts an unstructured document into fields: contact details, employers, dates, titles, education, skills. It is pattern recognition, and it performs well on clean, linear resumes. It degrades on multi-column layouts, tables, graphics, text inside headers, and unusual date formats, which is why career advice keeps telling applicants to submit plain files. Matching goes further and scores a parsed profile against requirement text using keyword overlap, skill taxonomies, or embeddings, depending on the vendor. That is genuinely useful for ordering a queue of several hundred. It is weak at judging seniority, at interpreting a career break, and at anything needing context about your particular team. Treat the score as a sort order, never a verdict. Ask how a strong candidate gets rescued from the bottom of a ranking, and how the model was tested for adverse impact. There is more on the mechanics in AI recruiting.

Which ATS features get used daily, and which only demo well?

The ones used daily are unglamorous. A fast apply flow. One inbox for every application. Bulk actions. Calendar-aware scheduling. Templated emails. Mobile access for hiring managers. A pipeline view someone can read in five seconds. The ones that demo beautifully and then gather dust are usually elaborate: assessment libraries nobody maintains, chatbots answering questions candidates never ask, forecasting dashboards built on data you do not have yet, and approval matrices that get bypassed within a month. SHRM cites an InFlight audit measuring 51 clicks to complete an average online application, and SHRM separately reports that 51% of candidates abandon long applications. That is where your money leaks. A shorter apply path beats every advanced module on the slide deck. Buy for the daily path first, because the feature list matters less than the friction it removes.

What are the different types of applicant tracking systems?

Six shapes exist, and the differences are structural rather than cosmetic. Small-business systems optimise for setup speed and a single administrator, often with a free tier. Mid-market systems add approval chains, custom fields, and reporting worth reading. Enterprise systems add configurability, security review, and suite integration, paid for with long deployments. Agency systems are organised around clients and submissions rather than internal headcount, with placement and margin tracking attached. Open-source systems hand you the source code and the hosting bill, which suits teams with engineering capacity and unusual requirements. AI-native systems put parsing, ranking, and automated outreach in the core instead of selling them as add-ons. Choose by hiring model, not headcount alone. The agency shape differs enough to need its own requirements list, and the self-hosted route is weighed up in open-source ATS.

What automation and integrations should an ATS include?

Automation should remove clerical work, not decisions. The rules worth having: acknowledge every application immediately, nudge an interviewer who has not filed feedback, advance a stage when a scorecard lands, close out remaining candidates when a role fills, and resurface a talent pool when a similar requisition opens. Anything that auto-rejects against a threshold needs a human review path, because that is where fairness complaints begin. On integrations, five connections carry most of the value: calendar and email, job boards and aggregators, background or reference checks, assessments, and whatever HR platform receives the hire. The rest can wait for an API or a webhook. Ask whether the API is publicly documented and whether webhooks fire on stage changes, since that decides what you can build later. Current connectors sit in the integrations directory, and workflow ideas in recruitment automation.

What reporting should you expect from an ATS?

Four reports answer almost every question leadership will ask. Conversion by stage shows exactly where candidates are being lost. Time in stage shows which step is genuinely slow, as opposed to which one feels slow. Source performance shows which channels produce hires rather than applications, and those are usually different lists. Interviewer load and feedback lag show whether the process is failing for internal reasons. Everything past those four reports is nice to have rather than necessary. What matters more than the chart library is whether the underlying data is clean, because if recruiters skip stages or backfill statuses every Friday, no dashboard can save you. Ask to export raw stage-change events during the demo itself. If the data cannot leave the product, you cannot audit it independently. Definitions for each measure are in the recruitment metrics guide.

How do you test ATS features in a demo instead of watching a slideshow?

Take control of the mouse and drive the demo yourself. Ask the vendor to build a live requisition using one of your real job titles, then apply to it from your own phone while they watch the clock. Upload the ugliest resume you own and inspect what the parser produced, field by field. Schedule a panel across two calendars in different time zones. Reject a candidate and read the email that goes out. Pull a report, then export it to a spreadsheet. Create a restricted user and try to see something they should not. Each task takes a couple of minutes, and each one has an unhappy path vendors prefer not to show. Score identical tasks across every shortlisted product so you are comparing behaviour rather than sales skill. The wider evaluation sequence sits in the ATS buyer's guide.

ATS features and how to test each one in a demo