Buy the stage that is leaking, not a platform that covers all of them.
Most multi-site hourly employers start shopping after a bad month. Three locations are short-staffed, the same requisitions keep reposting, and somebody decides the hiring software is the problem. The search that follows is for a platform: post jobs, track applicants, screen them, book interviews, onboard. Six demos later the shortlist costs more than the problem does, needs a quarter to roll out, and replaces the one part of the stack nobody was complaining about.
The cheaper move is to find the single stage where candidates actually disappear and buy only that. In high-volume hourly hiring it is almost always the same stage, and it is usually not the one the team names first.
Find the leak before you shop
Pull four numbers for one role at one location over the last 60 days. Not the whole company, not the whole year. One role, one site.
- Applications received
- Applicants who got invited to a first step
- Applicants who started that step
- Applicants who finished it
Here is what that looks like at a six-store restaurant group hiring shift leads and crew. One store, two weeks, 180 applications. A recruiter splitting time across all six stores reviewed resumes and called 24 of them. Nine picked up. Seven got through a screen. Four went to the general manager, two showed for a working interview, one took the offer.
One hire is a fine outcome. The problem is the 156 applicants nobody evaluated at all. They were not rejected. They were never reached, and most of them took a job somewhere else inside a week. The posting reposts next month and the cycle repeats.
Now notice what that funnel does to a buying decision. Nothing in it is an applicant-flow problem. The store got 180 applications for one opening. Spending more on job boards makes the queue longer and changes nothing downstream. Nothing in it is a tracking problem either. The recruiter knew exactly who applied. The constraint is screening capacity: one person cannot have 180 conversations, so 156 applicants got sorted by whose resume looked tidiest and who happened to answer a daytime call.
That gap between applying and the first real evaluation is where most candidate drop-off in hourly hiring happens. Measure it with everyone you invited in the denominator, not just the people who started. Most dashboards report completion among starters because that is the number the tool can see, and it flatters every vendor in the category.
Match the symptom to the category
Seven symptoms, seven different purchases. Only one of them is a screening tool.
| Symptom | What is actually broken | What to buy |
|---|---|---|
| Hundreds apply, most are never contacted | Screening capacity | An async screening step |
| Nobody can tell who applied to what, offers get lost | Workflow and records | An applicant tracking system |
| Too few applicants at any price | Sourcing and distribution | Job distribution or recruitment marketing |
| Interviews never get booked | Scheduling handshake | Scheduling automation, or drop scheduling from round one |
| Hires fail on a specific hard skill | Skill verification | A task-based assessment |
| Every location screens differently | Standardization | One question set, scored on one rubric |
| Paperwork stalls on day one | Onboarding | Onboarding, I-9, and credential tooling |
Write your four numbers down before you read vendor websites. Demos are built to make every tool look like the answer to all seven rows.
What each category actually does
Applicant tracking systems. An ATS is a system of record: requisitions, applications, stages, offers, compliance reporting. Buy one when you genuinely cannot answer who applied and what happened to them. What no ATS does is the judgment step. Something still has to decide which applicants are worth a manager's hour, and in most ATS deployments that decision is a human on a phone. Knockout questions on an application form are a gate, not an evaluation, which is the distinction behind pre-screening versus screening.
Frontline hiring platforms. The all-in-one tools built for hourly and multi-unit employers bundle distribution, tracking, texting, screening, scheduling, and onboarding. When three or four of those rows are broken at once, the bundle is the right buy and you should take it seriously. The catch is that you are buying your whole process from one vendor, usually on an annual contract with an implementation, and the screening module is rarely the strongest piece of it. We wrote up that tradeoff honestly against Fountain and Workstream, including the cases where they win.
Assessment batteries. Skills tests and psychometric instruments measure a defined trait or ability under controlled conditions. They are the right tool when a specific capability predicts success and self-report will not settle it. They cost the candidate real time, which is why they sit badly at the top of an hourly funnel where applicants are comparing five employers from a phone. More on that boundary in our TestGorilla comparison.
AI voice screening. An automated agent calls applicants, asks your questions, and scores the transcript. It removes recruiter hours and it is the right medium when phone delivery is the job, as with a call center agent or a dispatcher. Its ceiling is structural: somebody has to answer. A CNA on the floor, a picker whose phone is in a locker, and a driver mid-route are the people you most need to reach and the least likely to pick up an unknown number. Model your own answer rate before you compare time-saved-per-screen claims. The full version of that math is in AI phone screening.
Async text screening. Every applicant answers the same job-specific questions by text, on their own phone, whenever they have five minutes. Nobody schedules anything. You read scored, ranked results instead of a queue. This is the category that fixes the first row of the table, and it is the only row that was broken at that restaurant group.
Five questions that separate the demos
Ask these of every vendor in any category, and ask for numbers rather than adjectives.
- Of everyone we invite, what share finish? Not completion among people who start. Insist on the invited denominator, and ask for it on hourly pools comparable to yours.
- What does a candidate need to complete this? An app download, a webcam, a quiet room, or a resume upload each screen people out on hardware rather than ability. In senior living, warehouse, hospitality, and retail hiring, a large share of applicants have no laptop.
- How long from application to first ask? Elapsed time predicts drop-off better than process quality does. A good process that takes four days loses to a mediocre one that takes four hours.
- Can all our locations run the same questions and the same scoring? If each site can edit the rubric, you do not have comparable results across sites. That is the whole point of multi-location hiring consistency, and of a structured interview generally.
- What is the real per-hire cost at our volume? Include the applicants who never finish and the ones a per-call or per-minute model bills for anyway.
One more worth asking about your own side: what happens to applicants you decline? In accommodation, food services, and retail, the sectors with the highest quits rates in the Bureau of Labor Statistics Job Openings and Labor Turnover Survey, the person you pass on this month is often applying again next month.
Where we fit, and where we do not
Prelim is the first row of that table and nothing else. Paste a job description, the AI drafts role-specific screening questions, you edit them, and you share one link. Candidates answer by text in any phone browser with nothing to install. You get scored and ranked results with full transcripts, so a same-day decision on everyone who finished is realistic. Embed the link on a career page, a job posting, or inside your ATS. It integrates with Ashby and Zoho Recruit, and there is a public API and an MCP server if you want to wire it in yourself. Free to start, and a fraction of enterprise pricing.
Buy something else if what you need is an ATS of record, job distribution, interview scheduling, onboarding paperwork, or license and credential tracking. We do none of those. We also do not do live or recorded video, resume parsing, or sourcing. If verbal delivery is the competency you are grading, screen on the phone and look at AI phone calls versus text instead. If a hands-on skill decides the hire, keep your working interview and use the screen to decide who earns one.
A text screen will not tell you whether a crew follows someone on a slammed Friday night. It tells you which 20 of 180 applicants are worth finding out about.
Pilot it on one role at one location
Do not run a procurement project. Pick the single role with the worst applicant-to-contact ratio from the numbers you pulled, run the new first round at one site for two weeks, and keep the old process everywhere else. Then compare four things against your baseline: share of applicants evaluated, hours of recruiter time spent, days from application to first contact, and show rate at the next step.
If the tool is working, the first number moves from a fraction of your applicants to nearly all of them and the second number goes down at the same time. That is the only result worth paying for. Everything else in the category is a different purchase for a different broken stage, and buying it will not fix the one you have. More on running the first round this way in high-volume recruiting and asynchronous screening.