A regional distribution operator with eleven sites needs about 300 people between late September and the middle of December. The recruiting team writes one advert, localizes the address field, and pushes it to the careers page and four aggregators. Applications arrive in the thousands. Offers go out. Then the real number shows up: how many of those hires are still working in week three. That number, not the application count, is what the advert actually controls.
Most of the fixes people reach for are complicated. Assessment tools, structured scoring, a longer screening call. The simpler question is whether the advert told the truth about the job in the first place, in language a tired person reads on a phone at 9pm.
Ten years ago the advert did less of the filtering
In 2014 a seasonal posting at a large employer reached a much narrower pool. It went to a job board, maybe two, and applying took twenty minutes of typing. That friction was doing filtering work nobody had to design. A recruiter then ran a phone screen and covered the awkward parts out loud: the shift starts at 5am, the freezer aisle is 34 degrees, the assignment ends December 22 and most people are not kept.
That conversation has largely disappeared from the front of the process. One-click apply from an aggregator, resume parsing, and automated scheduling mean a candidate can go from seeing a posting to holding an interview slot without a human saying anything specific about the work. Volume went up and the first honest conversation moved later, often to orientation. An employer that discovers a mismatch at orientation has already paid for the background check, the badge, and the trainer's morning.
So the filtering has to move back into the text. Not as a warning label. As a description precise enough that the wrong person self-selects out and the right person recognizes their own schedule in it.
What a large employer can now say that it used to hedge
Pay transparency is the biggest practical change. Several states now require a pay range in the posting, and a multi-site employer with locations in even one of them usually ends up publishing ranges everywhere rather than maintaining two templates. This was treated as a compliance burden. In seasonal hiring it works as a filter, because the single most common reason a hire evaporates in week two is that the pay was discovered late and was lower than assumed.
The second change is that remote and hybrid labeling became standardized enough to be meaningful. Ten years ago "flexible" could mean anything. Now, for warehouse, retail, and logistics work, saying on-site and naming the site is unremarkable rather than defensive.
The third is that the Department of Labor oversees the classification and wage rules that sit under seasonal work, and the distinctions it maintains between employee and contractor, exempt and non-exempt, are now well enough known that candidates ask. Answering in the posting costs one sentence and removes a whole category of late-stage surprise.
The four lines that do the work
For a peak-season role at a large provider, four specifics carry nearly all the filtering weight. Everything else in the advert is packaging.
- The actual shift pattern, with hours and days. Not "flexible scheduling available." Write: four days a week, 5:00am to 3:30pm, including Saturdays from November 1. A parent with a school run reads that once and decides. So does the person who wanted it.
- The physical and environmental reality. Standing for the full shift, lifting up to 50 pounds repeatedly, refrigerated areas, concrete floors. This is the line most templates soften, and softening it is what produces the week-three attrition.
- Start date, end date, and the conversion odds. Seasonal means seasonal. If the assignment runs to December 23 and roughly one in six people is offered permanent work, say the shape of that honestly. Candidates with other options are deciding between offers and will reward the one that told them.
- The first-week logistics. Where to park, whether steel-toe boots are required and who pays for them, when the first paycheck lands. Large employers know these answers and rarely put them in the posting. They are the questions that make someone not come back on day two.
Running one honest template across eleven sites
The failure mode at scale is not writing a bad advert. It is writing a good one and then letting each site rewrite it. The version that survives is a locked template with a small number of required local fields: address, shift pattern, pay range, start and end dates, boot policy. Site managers fill the fields. They do not edit the body.
That constraint is unpopular for about one season. It stops being unpopular when the turnover report separates cleanly by site and the sites that used the locked version are holding more of their hires. A large employer has an advantage a small one does not here: eleven sites running the same text produce a comparison worth reading.
The advert is the cheapest screening instrument in the process and the only one every applicant reads. Writing it plainly costs an afternoon in August and shows up as bodies on the floor in November.