Blog/AI In HR

Campus recruitment: a dated fall and spring plan for small teams

Flowmingo Editorial TeamFlowmingo Editorial Team19 মিনিট পড়ুন
Campus recruitment: a dated plan for fall and spring: a calendar from October to September: requests open 12 Oct, a fair in January, scoring in February, 29 offers in March, 20 start in September

You need 3 graduates by next September, with no campus recruiter, no travel budget and no applicant tracking system. That is campus recruitment for most small teams, and it is not what most graduate hiring benchmarks describe.

Those benchmarks come mostly from surveys of far bigger employers. In the 2025 benchmarks survey of the National Association of Colleges and Employers (NACE), answered mostly by its members, more than half of the employers had over 5,000 staff.

The UK figures describe big programmes too. The Institute of Student Employers' 2025 UK survey polled its employer members, mostly large businesses with formal graduate programmes, and they averaged 140 applications per graduate vacancy.

The same survey found that 61% of those employers reported candidates using AI in interviews without permission, and 45% had given applicants no guidance on when AI use is appropriate. So say in the posting what AI use you allow.

Then judge everyone on 1 step they all do the same way, so the comparison does not rest on a written application that AI may have helped write.

Each shortcut a small team reaches for has a cost. Miss the schools' calendar, and the students you want may sign elsewhere.

Cut the pile on GPA, and you filter on grades, which a review of 114 samples links to only about 4% of job performance (section 7). Set a 48-hour deadline, and you may break a school's own rules, such as Duke University's 2-week minimum (section 9).

The calendar comes first, because every other date in this plan is counted from it. At the 3 US and Singapore universities in section 1, the main fall fairs were over by 7 October 2026.

So this month's job is to pick 2 schools and book spring dates with their career offices (sections 2 and 3). In the UK or anywhere else, check each school's own dates.

Key takeaways

  • When should campus recruitment start? About 3 months before the fair. Vanderbilt University opens spring interview requests on 12 October, so start now.
  • Can a campus recruitment ad ask for recent graduates only? Avoid it. The US calls it an age-law violation, the UK likely indirect age discrimination, and Singapore says avoid age.
  • How do I screen 500 near-identical applications without a GPA cutoff? Score each on 4 to 6 written criteria, then give everyone who passes the same task or interview.
  • How long should graduates get to accept an offer? NACE calls 1 to 2 weeks common. Because some graduates decline and some later back out, send about 29 offers for 20 starters.
  • How many interns should I convert to full-time hires? In NACE's latest survey, 63.1% of interns became full-time hires, so keeping 3 of 5 is typical.

1. When should campus recruitment start, and what does a fall and spring calendar look like?

Start about 3 months before the fair you want, and try not to hire for a job that starts more than 12 months away. That limit is NACE advice, not a rule.

NACE's reason is practical: offers made too far ahead leave an employer open to more students reneging and to fewer new hires staying.

Stony Brook and Vanderbilt in the US and the National University of Singapore (NUS) each run a fall season, roughly August to December, and a spring season, roughly January to April. The table shows their 2026-27 dates, which shift each year, so confirm them with each school.

When What the schools publish Your move
12 Jun to 10 Dec 2026 Fall: NUS Career Fest registration 12 Jun to 10 Jul for 23 Sep. Stony Brook 6 fairs, 26 Aug to 2 Oct. Vanderbilt season to 10 Dec, main fair 22-24 Sep, law fair 13 Oct, finance fair 14 Oct Main fall fairs are over. Already interviewing? Close offers before each school's season ends
12 Oct 2026 to 26 Apr 2027 Spring: Vanderbilt opens interview requests 12 Oct, season 11 Jan to 26 Apr, fair 26-27 Jan. Stony Brook 5 fairs, 27 Jan (general) to 24 Mar (volunteer) Now: email 2 career offices and request interview days, then run the fair and offers

"About 3 months" is a pattern rather than a rule, because it rests on only 2 schools. NUS closed fair registration 11 weeks before its fair, and Vanderbilt opens spring requests 13 weeks before classes.

To plan spring backwards, start from how long employers take to decide. NACE's 2025 average from first interview to an offer or rejection was 27.3 days.

Here is the spring back-plan, our arithmetic on that average. Take Vanderbilt's 26 Jan fair with first interviews on 27 Jan (our assumption), and offers go out about 23 Feb.

A 14-day deadline, inside the 1 to 2 weeks NACE calls common, then puts answers due about 9 Mar, 42 days after the fair. For a September start, a 23 Feb offer is about 6 months ahead, inside NACE's 12-month limit.

Campus recruitment timeline: spring fair on 26 Jan, offers about 23 Feb and answers due about 9 Mar

Worked timeline: a 26 Jan fair and first interviews on 27 Jan, with the same questions for everyone. Offers come about 23 Feb and answers are due about 9 Mar, 42 days after the fair (our arithmetic).

NACE's 27.3 days is an average, though, not a target. Our hiring process steps guide says to offer within 1 week of the final interview, because faster offers won more acceptances.

So count from your last interview instead. The 23 Feb date assumes that interview is on 16 Feb, and if your last interview is on 3 Feb, offers can go out about 10 Feb.

2. We're a 50-person company with no recruiter. How does campus recruitment work on a small budget?

Pick 2 schools, email each career office, then run 1 virtual talk, 1 set of same-question interviews and 1 offer window. A small team needs no travel budget and no applicant tracking system (ATS) for that.

For example, here is one way to lay it out as a 4-week campus hiring plan (our plan, not a study):

  1. Week 1: pick 2 schools by programme fit, and ask each career office for its rules, dates and fees.
  2. Week 2: write 4 to 6 criteria and a posting with no age words and an adjustment line, 1 sentence telling candidates how to ask for extra time or a different format.
  3. Week 3: open applications with that line on the page, and run 1 virtual talk or class slot.
  4. Week 4: run same-question interviews with the line in every invitation, then send offers with a 1 to 2 week deadline.

The adjustment line can be as short as this: "If you need extra time or a different format, email [name] before [date]."

The 4 weeks are working weeks, not 4 in a row. Week 1 starts about 3 months before the fair, weeks 2 and 3 fit in between, and week 4 starts with the fair and the interviews.

Offers then go out within 1 week of your last interview, as section 1 advises. For Vanderbilt's 26-27 Jan fair, that puts week 1 in late October, so email the career offices this month.

3. How do I choose schools for campus recruitment when I can only visit 3 or 4?

Choose schools by whether their programmes teach the skills the job needs and whether you can show up there, not by rank. Start with 2 schools, add up to 2 more once the first cycle shows yield, and drop any that give you no applicants.

A simple way to compare them is to score each school 1 point for every yes. For example, the table scores 2 schools for an operations analyst role (illustrative):

Criterion School A School B
Programme fit (a course teaching the job's 3 skills) A data analysis course None
Access (dates, fees and rules published) Dates and fees published Dates and fees published
Reach (can you visit in person?) 1 hour away 2 hours by train
Yield (past applicants and hires from here, and how they scored) 4 of your last 20 graduate applicants None
Score 4 of 4 2 of 4

With no past applicants yet, every school scores 0 on yield, so the other 3 criteria decide. Prestige is not one of them, because the research on it reads as a warning, though it covers elite firms only.

Lauren Rivera's 2012 paper in the American Sociological Review, now more than 10 years old, drew on 120 interviews at elite US banks, consulting firms and law firms.

The paper found that shared culture, such as leisure pursuits and self-presentation, often outweighed concern about productivity. So a list built on prestige can let fit stand in for skill.

Grades can still help, if you pick the right ones. A December 2024 study found that grades in job-like courses link to job performance more closely than grades overall, and section 7 has the numbers.

So ask for 2 grades from courses like the job, taken from an official transcript. Use them at the screen's third pass, described in section 6, as evidence and not a cutoff.

4. Are on-campus fairs worth the cost for 3 hires, or should campus recruitment be mostly virtual?

Spend on in-person events only at the 2 schools you picked, and use virtual sessions for the rest. The reason is a wide gap in how employers rate the 2 kinds of event.

In NACE's 2025 internship survey of 247 organisations, over 2 in 3 rated in-person fairs effective, while under 1 in 5 rated virtual ones effective. Virtual events are rated lower but cost far less, so the split between the 2 comes down to money.

The table shows Handshake's estimates, and Handshake sells the platform that registers employers for fairs, so these are a vendor's figures, with no sample size.

Cost Virtual event In-person event
People time $456 $3,491
Travel None $3,017 to $5,046
All costs combined As little as $1,000 $10,000 or more

Most of the gap is people time and travel, as the table shows. The school sets its own price too: NUS Career Fest, for example, starts at $1,100 excluding 9% GST, and its page shows no currency.

After your first event, measure cost per interview, which is the total event cost divided by the interviews it produced. A $1,000 virtual event that produces 10 interviews costs $100 each, for example, so a $10,000 in-person fair needs 100 interviews to match.

5. Can a campus recruitment ad ask for recent graduates only in the US, UK and Singapore?

Avoid it: US rules name "recent college graduate" as an ad term that violates the age law, and guidance from Acas, Britain's workplace advice service, calls it likely indirect age discrimination.

Singapore's Tripartite Guidelines, from TAFEP, an alliance for fair employment practices, say to avoid age in ads. This is general information, not legal advice, so check with counsel.

Most of the table below comes down to 3 habits: describe skills rather than age, write down the job reason for any school or grade bar, and tell every candidate how to ask for an adjustment.

Where Rule Your move
US, age 29 CFR 1625.4(a) names "young", "college student" and "recent college graduate" as age-law violations unless an exception applies Describe skills, not age
US, selection 29 CFR 1607.16(Q): educational requirements are selection procedures Write the job reason
US, adverse impact 29 CFR 1607.4(D): a pass rate under 80% of the top group's is generally regarded as evidence of adverse impact (race, sex or ethnicity) Compare pass rates at each school or GPA cut (example below the table)
US, disability 29 CFR 1630.11 and 1630.14: tests must reflect the skill, and medical questions wait for an offer Tell every candidate how to ask for an adjustment
UK, disability Acas: reasonable adjustments at any stage. Equality Act s.60: no health questions before an offer, except to arrange adjustments Name 1 adjustment contact
UK Acas: a "recent graduate" ad is likely indirect age discrimination. Equality Act ss.19 and 39 Be ready to justify requirements
Singapore TAFEP Guidelines: avoid age as a criterion Keep age words out, and keep interview, assessment and offer records at least 1 year (TAFEP)

The US regulations in the table date from 1978 to 2007 and the UK's Equality Act from 2010, so each is more than 15 years old. Acas says "likely" because section 19 of the Act allows a requirement the employer can show is proportionate.

The US age law, the Age Discrimination in Employment Act, covers people aged 40 and over at employers with 20 or more staff (29 USC 630(b)).

Singapore's Workplace Fairness Act is passed but not in force, and it is expected to take effect by end-2027. Firms under 25 staff are exempt at first, according to Singapore's Ministry of Manpower (MOM).

Those small firms must still follow TAFEP's Tripartite Guidelines, which are not law but are expected of all employers.

The 80% check compares pass rates. If 10% of one group passes a school or GPA cut but only 6% of another, 6 is 60% of 10, under 80%, so review the cut. With small numbers, a gap can also be chance.

In the US, the rule itself is under challenge. On 9 June 2026, the Justice Department's Office of Legal Counsel, whose opinions bind the executive branch, said the Uniform Guidelines that set the rule rest on an unconstitutional reading of Title VII, the US law they interpret.

The Equal Employment Opportunity Commission (EEOC) lists rescinding the interpretive parts of the guidelines on its agenda for November 2026. On 16 October 2026 it sent the final rule to White House review (Ogletree).

No final rule had been published by 20 November 2026, though, so the Uniform Guidelines were still in the US Code of Federal Regulations. Title VII's disparate-impact liability and private lawsuits both remain, so keep the check as a rule of thumb, not a safe harbour.

5.1 Can I hire only from certain schools or degrees?

Nothing we checked bans it, but a school list needs a job reason. In the US, a school list is a selection procedure (29 CFR 1607.16(Q)), so it is tested for adverse impact like any other test.

In the UK, a requirement that disadvantages a protected group must be a proportionate means of achieving a legitimate aim (Equality Act s.19).

The Singapore guidelines we read cover age, not school lists. Wherever you hire, keep pass rates by group, for example with the US 80% rule.

6. I have 500 near-identical graduate applications and thin CVs. How do I screen them fairly?

Score every campus recruitment applicant on 4 to 6 written, job-related criteria, then give everyone who clears the cut the same short task or interview.

The figure breaks that into 3 passes. The first checks hard facts, such as start date and location, which you can ask for in the application with pre-screening interview questions.

The second gives each applicant a weighted score, and the third gives everyone who clears the cut the same task or interview, with the adjustment line repeated in the task email.

Graduate hiring screen in 3 passes: hard facts, weighted score, then the same task or interview

Screening in 3 passes: hard facts, weighted scores, then the same task or interview for everyone who clears the cut (method, no data).

Write the criteria before you read a CV, because reading bends them. Uhlmann and Cohen ran 3 lab studies in 2005, more than 20 years ago. In the police-chief ones, evaluators shifted the criteria to favour whatever credentials the male applicant happened to have.

In the third, committing to the criteria before seeing the credentials removed the gap in favour of the male applicant.

Then add the scores with a fixed formula rather than an overall impression. A 2013 review of selection and admissions studies by Kuncel and colleagues, now more than 10 years old, found that a formula predicts performance more than 50% better than an expert's judgement.

Here is that method as a worksheet for a graduate operations analyst (illustrative, not customer data):

Criterion (weight) Thin-CV evidence Score 8 to 10 Score 3 to 5
Start date and location (pass or fail) A direct answer Yes --
Data handling (Must Have, 4) A course project, course grades or a part-time task Names the data, tool and finding Lists a course, no result
Written communication (Very Important, 3) A report or 3-sentence answer Clear, ordered, with an example Correct but generic
Ownership (Important, 2) A society role or project they ran Names what they owned and the result Took part, no clear role
Learning new tools (Good to Have, 1) A tool they taught themselves Names the tool and what they built Says "quick learner"

A criterion with no evidence scores 3 and is never skipped. The weighted score is (4 x data + 3 x writing + 2 x ownership + 1 x learning) / 10, so scores of 8, 5, 4 and 3 (no evidence on learning) give 5.8.

The 4-3-2-1 weights and the 3 for no evidence are this guide's convention, not a study's finding. Hand scoring also takes time: at 3 minutes each, 500 CVs take 25 hours (an assumption, not data), before any interview.

So set the cut from your interview hours, not from a fixed bar such as 8, which can leave you with no list at all (candidate shortlisting).

If an AI tool ranks the pile, test it first: change only the names on 20 screened CVs and log any rank change. In an October 2024 study, Wilson and Caliskan tested 3 text-embedding models, AI that retrieves the CVs closest to a job description.

On the same CVs under different names, the models significantly preferred white-associated names in 85.1% of 27 statistical tests. In other words, a name alone changed which CVs the AI picked.

How Flowmingo helps

Flowmingo, which publishes this guide, makes CV Evaluation. It turns a job description into criteria you edit and rate, then scores every CV from 0 to 10 on each criterion.

A criterion with no evidence counts as 3, as in the worksheet above, and nobody is filtered out automatically.

7. We screen on GPA. Does it predict job performance, or should campus recruitment drop the cutoff?

Skip the hard GPA cutoff in campus recruitment. Across 114 samples, grades explain only about 4% of the differences in job performance, and they say nothing about who quits.

Those figures come from a December 2024 meta-analysis, a study that pools many studies, by Van Iddekinge and colleagues in the Journal of Applied Psychology.

On a scale where 0 is no link and 1 a perfect one, the average for job performance was 0.21. For quitting it was -0.02, across 20 samples.

The 4% is our arithmetic: squaring 0.21 gives 0.044, the share of the differences in performance that grades account for. The same study found a closer link for grades in major-specific or job-like courses, which reach the 0.30s and 0.40s.

Fewer employers now screen on GPA. In NACE's Job Outlook 2026, 42% of employers use GPA as a screening tool, down from 73% in 2019.

Screen on evidence of the work itself instead. A 2022 analysis by Sackett and colleagues, now more than 3 years old, revisited earlier estimates of how well hiring methods predict job performance.

On the same scale, it scored structured interviews 0.42, job knowledge tests 0.40, work samples 0.33 and cognitive ability tests 0.31. Those 4 scores and the 0.21 for grades come from different studies and corrections, so read them as a rough ranking.

Graduate hiring evidence: grades track job performance at 0.21, below structured interviews at 0.42 and work samples at 0.33

Grades against other methods: structured interviews (0.42) and job knowledge tests (0.40) track later performance more closely than grades (0.21).

In practice, replace the cutoff with evidence that tracks performance more closely than overall GPA. Score coursework or project evidence on written anchors (section 6), and give everyone a short work sample.

Then ask everyone the same structured interview questions, as section 8 explains.

8. How do I interview 300 graduates with the same questions and compare them fairly?

Ask every graduate the same questions, score each answer on a written 1 to 5 scorecard, and have interviewers score alone before comparing. That is a structured interview, which at 0.42 scored highest of the 4 methods in section 7, though question quality matters.

Unstructured chats often fail for a simple reason: interviewers rate the same people differently. In Noise, a 2021 book on human judgement that is now more than 5 years old, Daniel Kahneman and his 2 co-authors make that point about job interviews (NPR).

The interview's 1 to 5 scale treats missing evidence differently from the CV screen. A thin CV may lack evidence through no fault of the graduate, so the CV pass gives 3, but an interview question asks for evidence directly, so an answer with none scores 1.

Here is a kit for a graduate operations analyst (illustrative, not customer data):

Question Skill A 5 looks like A 3 looks like A 1 looks like
Describe a time you learned a tool fast. What changed? Learning speed Names tool, steps, checkable result Describes the project, thin on how No example
A spreadsheet of 200 orders has 12 wrong prices. How would you find them and tell the team? Data handling Sets a check order, confirms the fix, plans a message Finds the errors, no plan to confirm or share Cannot say where to start
Tell me about a group project that went off track. What did you do? Ownership Names own action and outcome Describes the problem and the group's action Blames others, no action of own

The spreadsheet question can double as the short task from sections 6 and 7. Set it as a 20-minute exercise, and score the 3-sentence note to the team for written communication.

On the first question, "I learned Python in 3 weeks for my dissertation and cut the run from 2 hours to 4 minutes" scores 5, and "I am a fast learner" scores 1.

Volume is the practical limit. At 30 minutes each, 300 live interviews take 150 hours, about 4 working weeks (an assumption, not data), which can push offers past the 23 Feb in section 1.

If you use an AI interview instead, give candidates 7 days to finish it, as our interview invitation email templates guide advises, and put the adjustment line in the invitation.

Flowmingo data · AI interviews begun 1 Jul 2025 to 27 Sep 2026 · 19,440 interviews in projects titled graduate, intern, campus, early career or trainee · 227 companies · invitations sent 1 Jul 2025 to 7 Sep 2026 · 14,109 invitations · 131 companies

  • 84.5% of the AI interviews begun at the typical company's graduate, intern and trainee projects were finished, against 84.0% at other companies.
  • About 36% of everyone invited to the typical company's graduate, intern or trainee projects finished the interview within 14 days, against 30% at other companies.

The 84.5% finish rate uses the 50 companies with 20 or more interviews. Projects are grouped by title words, so early-career hiring under other titles sits in the comparison group.

Begun means the candidate opened the interview, and about 6 in 10 unfinished ones show only 1 message. These are associations, not causes, and say nothing about candidate quality.

In other words, at the typical company, graduate interviews were finished about as often as others. The 14-day figure measures something different: it counts everyone invited, including people who never opened the link.

The gap in that figure did not hold in every period. In March to May 2026 both groups were at about 35%, so do not plan on a graduate advantage.

The invitation figures leave out invitations sent from a job post, because those cannot be matched to an interview, and other unmatched invitations understate both groups. The data also cannot show where graduate invitees came from.

The card counts each company once because the 2 largest programmes make up 45% of these interviews and 60% of these invitations.

How Flowmingo helps

With Flowmingo's AI Interviewer, you send 1 link and each graduate does a structured, spoken interview in their own time.

Follow-ups depend on the answer, so only the base interview and criteria are shared. You get the recording, the transcript and a report scoring each criterion 0 to 10, and you decide who moves forward.

Name your adjustment contact in the invitation you write. On Free, the latest 300 videos stay for 30 days.

9. How long should graduates get to accept, and how many accepted offers still fall through?

Give graduates 1 to 2 weeks to answer, which NACE's guidance calls common, and plan for about 3 in 4 to accept. In the NACE survey in the table below, about 7% of accepted offers, roughly 1 in 14, still fell through, so send about 29 offers for 20 starters.

NACE's guidance adds that giving less time can be undue pressure, and some schools set their own rules. The Duke University Career Hub asks for at least 2 weeks on offers made in fall or spring.

Duke also asks employers to keep year-ahead offers open to 31 Aug, and return offers to summer interns open to 2 Nov or for 2 weeks, whichever is later. It prohibits exploding offers, for example a bonus for accepting within 24-48 hours.

The 2 benchmarks behind the 29 offers, in the table below, come from 1 survey of graduate offers, so they are not a trend.

Measure Figure Survey
Offers accepted 75% NACE 2024 report, Class of 2023, 235 employers, now more than 2 years old
Accepted full-time offers that fell through About 7% (10% for internships) Same report

To turn those 2 rates into an offer count, divide the starters you need by the share who accept and do not renege. For 20 starters, 20 / (0.75 x 0.93) = 28.7, or about 29 offers (our arithmetic).

Gartner's HR research puts acceptance much lower, at 48% of candidates in late 2025. It asks candidates about their latest offer, though, so it is not an employer's acceptance rate.

Graduate hiring offer funnel: of 100 graduate offers about 75 are accepted and about 70 start

Offers for 20 starters: of 100 graduate offers about 75 are accepted and about 70 start, so 20 starters need about 29 offers (our arithmetic).

Newer data points both ways. NACE's 2025 article says students accepted at a greater rate but also that reneges rose for full-time positions, and it prints neither percentage.

So treat 29 as a planning figure, and recheck it after your first round. For the 3 graduates in the opening, 3 / (0.75 x 0.93) is 4.3, so send 5 offers.

10. How many of our interns should we convert to full-time hires, and what rate is normal?

NACE's latest survey of 284 organisations puts the average at 63.1% of interns hired full time, so keeping 3 of 5 is typical. Size the intern class to the jobs you actually have.

Conversion means the share of the intern class that became full-time hires. It depends on how many offers are made and how many are accepted.

That 63.1% is for 2024-25 interns and the highest in 5 years, and NACE calls it "a steep climb of nearly 13%" from 2023-24. Acceptance of return offers reached 88.3%, so about 23 return offers keep 20 interns (20 / 0.883 = 22.7, our arithmetic).

Say you hire 5 interns and keep only 2, which is 40% (our arithmetic), below the average, so ask what changed. NACE warns that hiring more interns than you can integrate sets the programme up to fail, which is why the class should follow the jobs.

Write the return-offer rule before the internship starts, so every intern is judged on the same evidence. Use the intake's weighted criteria plus the manager's end-of-internship rating, so a return offer never rests on 1 person's liking.

How Flowmingo helps

Flowmingo is a free AI hiring platform, led by its AI Interviewer. Results, scores and transcripts stay on every plan, so you can compare each intake's scores with its end-of-internship ratings.

Your next round of campus recruitment then starts from what worked with Flowmingo.

11. Sources

Every study, law and survey in this guide links to a source below; Flowmingo figures come from Flowmingo's own platform data.

সাইন আপ করে প্রায় 60 সেকেন্ডে শুরু করুন। কার্ড লাগে না, কল লাগে না

বিনামূল্যে সাইন আপ করুন
Flowmingo Editorial Team

Flowmingo Editorial Team

We write practical guides for recruiters and HR teams who want to hire faster and more fairly. Each guide draws on hiring research, employment rules and Flowmingo's own data from real interviews, and lists its sources.

LinkedInYouTubeFacebook