I'm Paying for Job Ads. How Do I Know If They're Working?

Start with what the ad is supposed to deliver: people whose experience and interest make a next conversation worthwhile. Then compare your spending with that result—not only clicks or the size of the resume folder.

An ad does not control your whole hiring process. A clear posting can bring suitable people who never hear back from you. An unclear posting can attract lots of people for a different job. Those outcomes need different fixes.

A small-business discussion about paying for job ads prompted this question. It is one employer's unverified account, not evidence that a particular platform works or fails in general.

Keep the platform report and the hiring record separate

Your advertising report describes traffic and spending. Your hiring record describes what happened to the people who applied. You need both.

For example, Indeed's jobs-and-campaigns documentation distinguishes clicks, application starts, applications and advertising cost. It also distinguishes sponsored from organic traffic. Check the definitions and traffic filter in your own account before comparing numbers; a started application is not a completed one, and combined traffic is not paid traffic alone. Documentation checked September 11, 2026; this is not an account audit.

In your own record, add three stages:

• Relevant application: the evidence warrants a first conversation against the agreed job criteria. This is a review finding, not proof the person can do the job.
• Completed first conversation: you actually spoke, rather than merely sent an invitation.
• Agreed next step: both sides want to continue after discussing the work and terms.

Define the first stage before looking at which channel supplied the applicant. Keep missing information separate from a known mismatch, and apply the same criteria across sources. An AI score alone should not define a relevant application.

Worked example: the cheapest applications do not win the comparison

The following figures are fictional. Both channels advertise the same vacancy and offer over the same period. All applications in this example are attributed to paid activity, counted once, and have had the same follow-up window.

Channel A costs $5 per application, compared with $12 for B. If that is the only number you read, A looks better.

But A costs $50 per relevant application, compared with $30 for B. Its cost per completed conversation is $100, compared with $40 for B. The arithmetic changes the question from “Which brought more resumes?” to “What useful progress did we get for the spend?”

It does not establish that B will always perform better. These are small counts from one fictional vacancy. Differences in review, response speed, attribution and timing can change the result. Nor is $40 a market benchmark or a recommended target.

Advertising cost per completed conversation is also not total cost per hire. It excludes your time, other recruiting expenses and everything that happens after the conversation. Until someone is hired, you cannot report an observed cost per hire for this vacancy.

Find where the money stops producing progress

Few people view the posting: check that it is active, discoverable in the intended location and category, and described with a recognisable title. Do not assume that more budget fixes incorrect settings.

Views, but few completed applications: inspect the advertised duties, pay, hours and location, then try the application path yourself without submitting a fake application. Look for unclear terms, broken links or unnecessary steps. The counts identify a question; they do not establish why people left.

Applications, but little relevant evidence: review the criteria and the actual resumes. If every preference became a requirement, a different channel may produce the same empty shortlist. Use the no-qualified-applicants check before buying more traffic.

Relevant applicants, but few conversations: check your first-contact message, response time and scheduling. Separate declined invitations, no replies, missed appointments and people still waiting for your response. Our applicants-not-replying guide covers that handoff.

Conversations, but nobody wants to continue: ask what changed when the role became clearer. Record the applicant's stated reason where offered; do not invent one. More traffic may not fix an offer or responsibility mismatch.

Make the next comparison interpretable

Use a short record for each vacancy and channel:

Vacancy and version of the advertised offer:
Channel and paid/organic attribution rule:
Advertising dates and actual spend:
Application cutoff and follow-up cutoff:
Completed applications, excluding duplicates:
Relevant applications and criteria used:
Completed conversations and agreed next steps:
Still awaiting review, reply or a scheduled call:
One change to try and the date to review it:

Someone may see an ad, later visit your website and apply there. Preserve “unknown” or a stated attribution rule instead of pretending every hire has one perfectly measured source. Keep identifiable applicant records access-controlled; a channel summary rarely needs names.

For a bounded next test, keep the job and offer stable and change one major thing, such as the channel or the posting's explanation of the duties. Set an affordable spending cap and a review date in advance. Do not compare a mature month of applications with yesterday's new campaign, or keep spending simply because hiring has not finished.

If a denominator is zero, report “no completed conversations yet,” not a fabricated cost. If only two people have reached a stage, show the count alongside any percentage. Waiting applications belong in a pending category, not the failure column.

What screening software can—and cannot—tell you

Resume Ranker can help review collected PDF resumes against a job description and written criteria. A person still needs to inspect the evidence and decide who merits a conversation.

It does not buy ads, attribute applications to channels or calculate your recruiting return. Keep those records separately. Your immediate goal is simpler than proving a platform is good or bad: identify which part of this vacancy's journey deserves the next change.