Distinguish total rewards from recruiting
Recruiting work helps an organization find and select people. Compensation, benefits and job-analysis work concerns how roles and rewards are described, compared and administered. BLS identifies duties such as evaluating job descriptions, analyzing compensation data and comparing benefit programs. An internship may support a limited part of this work rather than design the organization's pay policy.
Search compensation analyst intern, total rewards intern, benefits intern and job analysis intern. Read the project description to see whether it focuses on market data, role documentation, reporting, employee communication or program administration. A general HR internship could include some of these tasks, but the title alone does not establish an analytical compensation assignment.
The work can suit a student who likes careful data comparison and writing as well as spreadsheets. A number can be mathematically correct while answering the wrong question. Your sample should show that you can identify the role being compared, preserve the pay basis and communicate what the result means without turning a limited dataset into a broad claim about fairness or market value.
Sources[1]
Match responsibilities before matching salaries
Consider two fictional postings titled Analyst. One maintains an operating dashboard under supervision. The other leads a client workstream and supervises two colleagues. A title match is not enough to treat their salaries as observations of the same job. For a practice project, build a role-comparison table using responsibilities, scope, required skills, decision authority and experience expectations.
Add location, working arrangement and the data's observation period. Keep distinctions that could affect comparability rather than deleting them because the dataset is small. A remote posting with a location-dependent range is not one universal salary number. A range for several levels is not automatically the midpoint for an entry-level role.
Create a documented match disposition for each record: reasonably comparable for this exercise, excluded for a stated mismatch, or needs review. Preserve the reason. The objective is not to accumulate the largest possible sample; it is to make the comparison interpretable. In a real organization, use its approved job-matching process and review rather than treating a student rubric as a compensation policy.
Preserve base pay, incentives and the time unit
Choose a single defined measure for the first analysis, such as annual base pay for comparable full-time roles. Keep signing bonuses, target incentives, commissions, equity and benefits in separate fields. Adding a one-time payment to recurring base salary can distort a comparison when the next record excludes that payment.
An hourly rate requires assumptions to convert to annual pay. Do not silently assume paid hours, overtime or weeks worked. In an internship search, a disclosed annualized salary may not represent the money paid over the internship's actual duration. Preserve the source unit and give any conversion its own assumptions, especially when comparing roles of different lengths.
For a synthetic project, include a deliberately inconsistent record and explain how you handle it. One record might state $25 per hour while the others show annual base salary. Hold it out of an annual-only summary until an explicit comparison basis is established. A missing amount or an open-ended range should remain visibly different from zero.
Work the range-position calculation without overclaiming
Suppose a fictional salary range has a midpoint of $80,000. One employee's annual base pay is $72,000 and another's is $88,000. Pay divided by the range midpoint is 0.90 for the first and 1.10 for the second. In this exercise, call that the midpoint ratio and state its formula. It describes position relative to the chosen midpoint, not the cause of the difference.
Those ratios do not establish that either employee is incorrectly paid. They do not account for whether the jobs were matched correctly, how the range was set or the organization's relevant decisions and records. A sensible internship work sample distinguishes the calculation from a recommendation. Identify what the result would prompt you to investigate rather than announcing an unsupported correction.
Now test the data definition. If the $88,000 record includes an $8,000 one-time payment while the $80,000 midpoint refers to base pay, the original 1.10 comparison mixes measures. On an $80,000 base-only figure, the ratio would be 1.00. The example illustrates why parsing the compensation definition can matter more than writing a more elaborate formula.
Explain what a small market sample can support
Assume five synthetic, comparable annual base-pay observations of $60,000, $68,000, $72,000, $84,000 and $96,000. Their median is $72,000 and their mean is $76,000. The difference reflects the values in this small constructed sample; neither statistic is an independently established market rate for a real job.
A source may report advertised ranges, self-reported salaries or employer survey data. Those are different evidence types. Record which one you have and avoid treating the midpoint of an advertised range as a salary actually received by an employee. Where the source's population and methodology are unclear, the uncertainty belongs beside the result.
The practice memo should name the sample size, inclusion rules and observed period. Explain which exclusions you made and how the result changes under a reasonable alternative set. If removing one observation changes your conclusion, show that sensitivity rather than hiding it. A small dataset can demonstrate analytical judgment without supporting a sweeping compensation recommendation.
Use benefits data to test communication and denominators
A benefits internship may involve explaining a program or maintaining enrollment information rather than comparing salaries. Build a fictional example with 200 eligible employees, 150 enrolled participants and 40 inquiries during a period. Enrollment is 75% of the eligible population. The 40 inquiries are not automatically 40 different people, and they do not establish dissatisfaction with the program.
If you want a distinct-participant inquiry rate, deduplicate the inquiry records first and define the denominator. If 25 enrolled participants made the 40 inquiries, the observed participant inquiry rate is about 16.7%. That can guide further investigation into common questions, but it does not prove the benefit is poor or the communication ineffective.
Create a one-page program explanation from a fictional approved brief. State eligibility, key dates, action required and where to obtain authoritative help. Leave legal, tax, medical and individual financial recommendations to the appropriate specialists. The portfolio task is to communicate a defined program accurately and surface ambiguous source material for review, not advise real employees on their personal choices.
Show care with sensitive information and review
Use entirely synthetic employee records in an application portfolio. Removing names from a small real team may not prevent individuals from being identifiable through salary, title and location. A realistic invented dataset is sufficient to show formulas, joins, quality checks and documentation. Do not treat private access as a credential to display publicly.
A strong package includes a role-matching table, a pay-definition dictionary, a small analysis, an exception log and a short decision note. Include a record that you refused to compare and explain why. The reviewer should see that you understand how mistakes enter the analysis, not just how to produce a polished summary of its output.
In interviews, ask who reviews job matches and external survey use, which data you would handle, and what a successful intern deliverable looks like. Describe your sample as a fictional benchmarking exercise. A precise explanation of why two similar titles are not comparable can be a stronger signal than claiming to have found an entire organization's optimal compensation structure.