The Target-School Tier Read: Core, Secondary, Digital-Only, or Drop

This is the scoring standard for deciding which universities stay on an early-career target list, how much to invest in each, and how to defend the list when an executive pushes for their alma mater. It is for university recruiting leads, talent-intelligence analysts, and operators sizing a campus market. It gives you five dimensions you can count, a rubric that turns them into Core, Secondary, Digital-Only, or Drop, and the specific ways each piece of evidence can mislead you.

What do Core, Secondary, Digital-Only, and Drop mean?

The four tiers are levels of investment, not prestige rankings. Each tier gets a different level of presence on campus, and a school earns its tier from evidence, not from which leaders went there.

How many schools land in each tier comes out of the scoring. You do not set it in advance. Public benchmarks are weaker than they look:

• An early-talent research firm reports that employers engage a median of 20 schools, with deeper investment concentrated in a smaller subset.
• A secondary source says average target lists shrank from 39 schools to 25, and that 63% of university recruiting teams now use tiers. I could not retrieve the primary report, so treat the 39-to-25 figure as unverified.
• The same source suggests 5 to 8 schools in the top tier, 8 to 12 in the second, and an open digital tier. That is one vendor's advice, not a measured norm.

Why score from the outside in? The standard advice is "use your historical yield," which assumes you have some. It fails for a new role family, a new geography, or a team with fewer than three cycles of hires. Outside-in evidence exists for every school whether or not you have ever hired there. Outside-in means federal completions data, public placement records, and campus employer rosters. When you set it against your own yield, the tier becomes a number you can defend.

The five dimensions and what each one proves

Score every school on five dimensions: graduate supply, peer-employer pipeline, employer crowding, internal yield, and reach cost. Anyone can count the first three from public sources. The last two come from your own records, and they are the only ones a competitor cannot see.

Each layer makes a narrower claim than the one before. A school can have a program but no graduates, graduates but no peer placements, or peer placements but none of your hires. The method works because it keeps those claims separate instead of blending them into a general sense of "strong school."

How do you score graduate supply from federal completions data?

Supply is the three-year average of first-major completions at the award level you hire, summed across the six-digit CIP codes that define the role family. CIP, the Classification of Instructional Programs, is the federal list of program codes that US institutions use. A six-digit code is its most specific level.

The source is IPEDS Completions, the federal survey of degrees awarded. It counts awards by award level, race/ethnicity, sex, and six-digit CIP code, and it splits them into first and second majors. Each reporting year runs from July 1 of the previous calendar year to June 30 of the current one. The final file comes out about two years after the degrees are awarded. Before each cycle, check the federal catalog for the newest final year instead of assuming you know it.

Start from the role-family codes:

The software codes are Computer Science, Computer Engineering General, Computer Software Engineering, and Computer and Information Sciences General. 30.7001 is Data Science. 52.0201 is Business Administration and 52.1301 is Business Analytics. You can browse the full taxonomy in the CIP 2020 browser.

Scoring:

• 0: no relevant program at your award level.
• 1, 2, 3: bottom, middle, or top third of the candidate list by three-year average.

Scoring against the candidate list instead of a fixed cutoff is deliberate. No published threshold says how many completions make a school worth a trip. Ranking into thirds lets a second analyst reproduce the same score from the same file.

College Scorecard is not a substitute for IPEDS. Its counts are by institution, four-digit CIP code, and credential level. It builds them by adding up related six-digit codes so that fewer cells need privacy suppression. That merges Computer Science with neighbouring programs you may not want to count, and its field-of-study earnings files use older, pooled award years. Use it to confirm a program exists, not to count supply.

Make two adjustments before you score:

• Count first majors only. Second majors are students already counted in another first-major column, so adding them double counts. Keep second majors as a note on interdisciplinary strength.
• Check whether a program exists separately from its completions. A school that launched a 30.7001 Data Science program in the last two years will show near-zero supply in the final file because its graduates have not reached the data yet. Mark it "program exists, supply lagging" and hold it at Digital-Only rather than scoring it as absent.

How do you count alumni at competitor companies by university?

Pipeline is the number of a school's graduates, zero to three years out, who now hold an on-target title at a fixed list of 8 to 15 named peer employers, recorded per cohort year. It shows that firms competing for the same hires already recruit at the school and that its graduates get through their screens.

Choose peers that hire the same role family at the same level in the same geography. Geography also changes how long the peer list needs to be. In Refolk's index of professional profiles, three entry-level segments look like this, each with a 25-profile sample read by hand:

The table shapes the method in two ways.

• Concentration differs by market. The US software sample clusters at large named employers, with Google holding 4 of 25. No UK employer appeared more than once. A UK peer list therefore needs to sit near the top of the 8-to-15 range to produce counts large enough to compare.
• Title quality differs by segment. The share of profiles with an on-target title falls from 92% for US software engineers to 60% in the UK and 44% for US analysts. A raw count of profiles tagged "entry level" overstates the pipeline, and by a different amount in each segment. Filter on title text and graduation year, never on a seniority tag.

Count per cohort year, not as one cumulative total. The closest public precedent is one consulting-prep site's list of target schools for the top consulting firms. It came from LinkedIn placement data on roughly 3,000 consultants who graduated over a seven-year span, counted cumulatively. A cumulative count hides a school whose pipeline dried up three years ago. No published standard sets a minimum sample per school or measures how much public profiles undercount by type of school. Record the count and cohort beside each score, and treat single-digit counts as low confidence.

Scoring:

• 0: no graduates found at any named peer in the window.
• 1, 2, 3: bottom, middle, or top third of the candidate list by count.

By hand, this is the slowest step: often several days of profile searching for each role family. Refolk removes most of that work. You describe the school, the graduation window, and the peer firms in one sentence, and you get back the matching profiles from LinkedIn, GitHub, and the open web, ready to filter by title and graduation year.

How do you score crowding, internal yield, and reach cost?

These three dimensions decide how expensive a talented school is to win. Crowding comes from public employer lists. Yield comes from your own hiring records, with the sample size written beside it. Reach cost comes from the per-school lines of the NACE cost-per-hire worksheet.

Employer crowding

Count the peer and competitor employers that appear on the school's career-fair exhibitor rosters, on-campus recruiting (OCR) rosters, and First-Destination Survey (FDS) top-employer lists. An FDS is the school's report of where a graduating class went. Its knowledge rate is the share of graduates whose outcome the school knows.

Crowding shows how much competition there is for students' attention. It misleads when the FDS behind it is thin.

• NACE standards say schools should aim for a knowledge rate of at least 65%. Schools can count outcomes taken from LinkedIn profiles, other online sources, fellow graduates, or parents.
• One university used LinkedIn and other sources to raise a survey response rate of roughly 19% to a knowledge rate of about 40%.
• Rates also vary inside a single school. One university's Class of 2023 report shows its School of Nursing at 59.9% and its Computing and Engineering school at 85.9%.

Score crowding in reverse: 3 for the least crowded third of the candidate list, 1 for the most crowded. Write the knowledge rate and cohort size beside every list you used.

Internal yield

Pull hires, offer acceptance, intern conversion, and first-year retention by school, and record n for each figure. Conversion is the share of interns who end up accepting a full-time offer.

National figures give context, but they cannot tier a school because NACE publishes them by employer, not by school:

Keep three things in mind when reading yield:

• Internship format moves yield. In-person internships had a 71.9% offer rate and 58.5% conversion, against 56.2% and 46.0% for hybrid. A school where you happened to run in-person cohorts will look better for reasons that have nothing to do with the school.
• Employer targets differ. Visa's early-career programs target a 70% conversion rate. Compare a school against your other schools, never against a national average.
• National samples are small and self-selected. NACE figures rest on roughly 200 to 280 responding organizations at response rates around 25%. They are fine as a rough guide and too weak to tier a single school.

Reach cost

Cost each school's reach from the NACE Cost-per-Hire Worksheet, which is based on Robert Greenberg's method. Use the lines that grow with every school you add:

• recruiter travel and accommodations, and equipment rentals
• pre-recruiting costs such as resume books and information sessions
• candidate visits and relocation
• advertising

Career-fair fees and sponsorships have no separate line in the worksheet. Add them as their own named line so they stay visible.

Do not cost schools from published benchmarks. They disagree with each other, and some are older than their labels:

The $6,275 campus figure now circulates with a recent date, but a recruiting platform was already citing it roughly seven years ago. The SHRM mean is 4.6x its median. Budgets disagree in the same way. NACE reports a median recruiting budget of $114,000, while an early-talent research firm reports a median university recruiting team budget of $291,500. The 2.6x gap comes from different samples. Only your own worksheet lines belong in the score.

Score reach cost in reverse: 3 for the cheapest third of the candidate list, 1 for the most expensive.

How do you turn five scores into a tier?

Add supply and pipeline into a talent score from 0 to 6. Add crowding and reach cost into a win score from 2 to 6. Apply the tier rule, then let internal yield move the result one tier, but only when n is large enough. The rubric below is the whole decision. Paste it into the scoring sheet so both analysts work from the same text.

ROLE FAMILY: ______   CIP CODES: ______   AWARD LEVEL: ______
PEER EMPLOYERS (8 to 15, fixed and dated before counting): ______

SUPPLY (S)    0 = no relevant program | 1/2/3 = bottom/middle/top third, 3-yr avg first-major completions
PIPELINE (P)  0 = none found at peers | 1/2/3 = bottom/middle/top third, grads 0 to 3 yrs out, on-target title
CROWDING (C)  3/2/1 = least/middle/most crowded third, peer + competitor employers on rosters and FDS
COST (R)      3/2/1 = cheapest/middle/most expensive third, NACE worksheet lines per school
YIELD (Y)     hires __ | acceptance __ | conversion __ | retention __ | n __

TALENT  T = S + P   (0 to 6)
WIN     W = C + R   (2 to 6)

CORE          T >= 5 and W >= 4
SECONDARY     T >= 4 and not Core
DIGITAL-ONLY  T = 3, or program exists with supply lagging
DROP          T <= 2

YIELD MODIFIER  only if n >= 10: top third of own schools = up one tier; bottom third = down one tier
EVIDENCE NOTES  FDS knowledge rate __ | FDS cohort size __ | pipeline by cohort __ / __ / __

The matrix below shows the shape of the rule. When a school sits near a line, the rubric thresholds decide.

A tier is a number with a source column beside it, not a preference with a slide.

What makes the tier reproducible is the blind re-score. A second analyst scores the same schools from the same evidence without seeing the first result. If the two tiers differ, find the dimension that caused the gap and fix its definition: a CIP code, a peer firm, a cohort window. Do not just argue about the score. A disagreement that you settle by debate will come back next cycle.

The tiering procedure, step by step

Eight steps take a role family from CIP codes to a defended school list, and the blind re-score is what turns the result into a standard rather than an opinion. My time estimates for one role family: about a day each for codes, supply, yield, and cost; 1 to 2 days for crowding; and 2 to 5 days for pipeline if you count by hand.

Sources disagree on the order of operations. One recruiting-software vendor recommends tiering first and then concentrating spend on the top tier. An early-talent research firm describes cutting schools as a cost-control step taken after budget cuts. I tier first. A list cut under budget pressure without scores keeps whichever schools have the loudest advocates.

How does the tier read go wrong?

Most bad tiers come from eight errors, and each one produces a specific false positive: a school that looks stronger, more crowded, cheaper, or better proven than it is. Check for all eight before the list leaves your hands.

1. CIP miscoding. Schools code the same discipline differently. Charlotte codes its Data Science doctorate as 11.0701, and Brown codes CS-Economics as 11.0101. A count based only on codes misses some programs and includes others you never meant to count. Check: read each candidate school's published CIP list.
2. Double-counted second majors. IPEDS reports first and second majors separately, so adding them together inflates supply, most at schools with popular double majors. Check: headcount from first majors only, with second majors kept as a note.
3. Supply lag read as absence. The final completions file comes out about two years after degrees are awarded, so a new 30.7001 program shows near-zero supply. Check: confirm the program exists separately, and hold it at Digital-Only.
4. A noisy "entry level" filter. Seniority tags let founders, CEOs, and CTOs into every sample in Refolk's index. A raw count overstates the pipeline. Check: filter on title text and graduation year, not on seniority tags.
5. FDS lists from schools with low knowledge rates. Below the 65% standard, a school's top-employer list leans toward the graduates who responded. Check: record the knowledge rate and cohort size next to every list.
6. Stale benchmarks. The $6,275 figure labelled with a recent year is older than its label. One vendor's own two versions of the same page compare it against $5,475 in one and $4,700 in the other. Check: cite primary sources only, and cost schools from your own worksheet.
7. Small-n yield. Two hires with 100% retention is noise. The standard error at n = 2 is about 0.35. Check: yield does not move a tier below 10 hires.
8. Mean versus median cost. Using the SHRM mean of $5,475 instead of the $1,200 median makes every school look more expensive, each by a different amount. Check: never use a national cost figure in a per-school score.

Internship format causes a quieter version of error 7. Because in-person offer rates ran 15.7 points above hybrid, part of a school's apparent yield may reflect how you ran the internship, not the school. When yield does qualify to move a tier, check whether the strong schools were also the in-person ones.

How do you defend the list and keep it current?

Every school an executive asks about gets the same five scores as every other school, and you re-score the whole list each cycle as new completions files, new rosters, and new yield come in. Defending the list means showing the number, not winning the argument.

When an executive's school does not earn a tier, there are two honest outcomes: it stays where it scored, or it goes on the list as a visible exception. Never quietly move it up a tier.

SCHOOL: ______                 REQUESTED BY: ______        DATE: ______
ROLE FAMILY: ______            PEER LIST VERSION: ______

SCORES   S __  P __  C __  R __   T __  W __   YIELD n __
TIER BY RUBRIC: ______
COMPARABLE SCHOOLS AT THIS TIER: ______, ______, ______

DECISION: [ ] Tier as scored   [ ] Exception
IF EXCEPTION
  Tier granted: ______
  Reason (business case, not affiliation): ______
  Reach cost committed: ______
  Owner: ______        Review date (next cycle): ______
  What would end the exception: ______

To keep the list current, re-score on a fixed cadence rather than when someone complains:

• Supply: each year, check the federal catalog for the newest final completions file. Each IPEDS collection covers awards from July 1 to June 30, so a new year arrives annually, still about two years behind.
• Pipeline: add the newest cohort each cycle and drop the oldest from the three-year window. If you built the counts in Refolk, re-run the same plain-English query with the graduation years moved forward one year. The definition stays identical, so year-over-year changes are real.
• Crowding: pull fresh rosters and FDS reports, and update the knowledge rate beside each one.
• Yield and cost: add the cycle's hires and worksheet lines. A school may cross the 10-hire line this year, so check.
• Definitions: keep the peer list and CIP codes fixed within a cycle. Date any change between cycles and note it on the list.