How to Write Work Projects as Research in an SOP

How to Write Work Projects as Research in an SOP

A work project counts as research evidence in your statement of purpose only if you can state four things: the question, the method, the result, and what you could not conclude. If you cannot state all four, do not call it research. Call it what it was, analysis, engineering or delivery, and use it as preparation.

Our view is blunt: the word "research" on a delivery project is an easy overclaim for a faculty reader to spot, because the paragraph around it describes building, not finding out. Industry work is strong evidence for a master's. It just has to be the right kind of evidence, named honestly. When the paragraph is drafted, the MS statement of purpose rubric reads it for preparedness.

Every program page and definition quoted below was checked on October 31, 2026. This page owns one task: turning industry work into research evidence. If you have no projects of this kind at all, no research experience? How to write your SOP anyway covers what replaces research. If your work was in an IT services company, the SOP for MS CS after TCS, Infosys or Wipro covers tickets, bench time and bonds.

What programs ask about work outside the transcript

Some programs say outright that the statement is where job-based training goes. Harvard's School of Engineering and Applied Sciences, on its master's in data science application page, says: "If your training in computer science, math and/or statistics is not reflected in the courses listed on your transcripts, you will need to find other parts of the application to make your training clear to the Admissions Committee." It adds: "You can also describe your training at one of your jobs or through self-study in your Statement of Purpose." The same page asks you to "keep the statement to about 1,000 words."

UC Berkeley's Graduate Division statement guidance asks applicants who describe research to name who they worked with, the project's title, their own responsibilities and the outcome.

Neither page says work projects count as research. Work counts as training; research needs a named role and an outcome.

What the formal definition of research says

There is an official definition of research, and it is strict. The OECD's Frascati Manual (7th edition, 2015) is the international standard for counting research and development. The US National Center for Science and Engineering Statistics reproduces it in Definitions of Research and Development: An Annotated Compilation of Official Sources (NCSES 22-209, May 19, 2022):

"Research and experimental development (R&D) comprise creative and systematic work undertaken in order to increase the stock of knowledge—including knowledge of humankind, culture and society—and to devise new applications of available knowledge."

The manual lists five criteria. The activity must be "novel creative uncertain systematic transferable and/or reproducible", and "All five criteria are to be met, at least in principle, every time an R&D activity is undertaken".

The same compilation reproduces an accounting-standard list (FASB Statement No. 2) of activities that "typically would be excluded from research and development". It includes:

• "Quality control during commercial production including routine testing of products."
• "Trouble-shooting in connection with break-downs during commercial production."
• "Routine, ongoing efforts to refine, enrich, or otherwise improve upon the qualities of an existing product."

Published requirement vs our recommendation. These are statistical and accounting definitions. They are not admissions rules, and no admissions page we checked cites them. We use them as a test because they draw the same line a faculty reader draws: work that resolved an uncertainty is research-like; work that fixed, maintained or shipped something known is not, however hard it was.

The four-part test: question, method, result, limit

Run every project you want to call research through these four questions. Write one sentence for each.

1. Question. What did you not know at the start? It must be a question, not a task. "Why do 30-day users leave?" is a question. "Build a churn dashboard" is a task.
2. Method. How did you find out? Name the design: a controlled test, a comparison of groups, a model and how you validated it, a structured set of interviews.
3. Result. What did you learn, and how sure are you? A direction and a size, if you have one.
4. Limit. What could you not conclude, and why? This is the sentence that shows research judgment.

All four: you may call it research, or "a research-style investigation" if your role was partial. Only the first three: call it an analysis. Only what you built: it is engineering or delivery, which is still good evidence for a coursework master's.

Our view: the limit sentence is worth more than the result. A reader can believe a modest result with a clear limit. A large result with no limit reads as a sales pitch.

The countercase: if you are applying to a coursework master's that asks for no research, do not force the test at all. Describe what you built and what it taught you, as the SOP after years of work experience guide shows. Calling delivery "research" for a program that never asked for research only adds a claim the reader has to check.

Before and after: two work projects rewritten

Two constructed examples, written by us, not real applicant statements. The companies, figures and results are invented.

Constructed example: a fintech data engineer's churn analysis

Context (invented): a data engineer at a payments app in Lagos, applying to an MS in data science.

Before: At my company, I conducted extensive research on customer churn using big data technologies, and my research helped the business reduce churn significantly.

After: We did not know why new users stopped paying through our app within a month. I compared users who left with users who stayed, on 18 months of transaction logs, and found that the drop was concentrated among users whose first transfer failed. A retry change based on that finding reduced early exits in the next quarter, but I could not separate its effect from a marketing campaign that ran at the same time. I want the causal-inference coursework that would have let me design that test properly.

The "before" version (23 words) uses "research" twice and gives none of the four parts. The "after" version (88 words) has a question (why users left), a method (a comparison of two groups), a result (failed first transfers), and a limit (the overlapping campaign). It never uses the word "research", and it is stronger for it. The limit also points straight at the program.

Constructed example: a process engineer's yield investigation

Context (invented): a process engineer at a pharmaceutical tablet plant in Hyderabad, applying to an MS in chemical engineering with a thesis option.

Before: I was responsible for research and development activities to improve the yield of our tablet line.

After: Our coating step lost more tablets in the monsoon months than in the dry season, and no one knew whether humidity or the operators' adjustments to it were the cause. With my supervisor's approval, I ran a two-month trial in which we held the spray rate fixed on alternate batches and logged humidity for each one. Losses tracked humidity even at the fixed rate, so we changed the drying schedule. I could test only one line and one product, so I do not know whether the effect holds for other coatings. That is the question I would like to take into a thesis.

The "before" version (16 words) is a job description. The "after" version (103 words) passes all four parts, and it states the role honestly ("with my supervisor's approval"). It would still not be research if the engineer had only adjusted the line until the losses stopped. That is troubleshooting, and the exclusion list above says so.

NDAs: describe the method, not the client

You can almost always describe the question type, the method and the kind of data, even when you cannot name the client or give figures. "Transaction logs for a consumer payments product" is enough. A result can be a direction ("losses tracked humidity") rather than a number. Do not name a client or give a figure your employer has not made public. If the confidential work is the core of a PhD application, the PhD SOP guide to confidential research work covers the permission check.

When to call it research, and when not to

For UK applicants deciding between an MSc and an MRes as a route to a PhD, the UK MSc, MRes and PhD preparation guide compares the routes. And if the application form has a separate "research experience" field, put there only the projects that pass the test. Industry work that does not pass belongs under employment.

Questions about work projects in an SOP

Can industry experience count as research experience?

Yes, when the work answered an open question with a planned method and you can state what it could not show. Development and delivery work is valuable evidence of preparation, but it is not research, so do not label it as research.

My project was a team effort. What can I claim?

Claim the part you did, in the first person, and name the rest as the team's. "I designed the comparison; a colleague built the pipeline" is more convincing than "we researched". The guide to explaining a co-authored paper in your SOP covers contribution wording.

How long should a work-project paragraph be?

About 90 to 120 words is enough for the four parts, as in the two rewrites above. Harvard SEAS asks for the whole statement in "about 1,000 words", so one project paragraph should not take a third of it.

Get a free review of your statement of purpose

Get a free Quick Review of your statement of purpose against GradPilot's general MS statement of purpose rubric. It reads how well your academic, research or professional experience shows you are ready for graduate study, and how clearly your goals and program fit come through. For a single-issue check, the preparation-first lens reads whether every claim about your readiness is named or bounded. You can also start from the statement of purpose review page.

More guides on turning work into evidence:

• SOP after 10 years of work experience: choosing three decisions from a long career.
• SOP after a failed startup: writing about a company that closed.
• Graduate school essays hub: every master's statement guide.

Sources

Retrieved October 31, 2026, except where noted.

• Harvard John A. Paulson School of Engineering and Applied Sciences, Master's in Data Science: How to Apply (no year stated): training at jobs in the statement; about 1,000 words.
• UC Berkeley Graduate Division, Writing Your Statements (no year stated): what to say about research you conducted.
• National Center for Science and Engineering Statistics, Definitions of Research and Development: An Annotated Compilation of Official Sources (NCSES 22-209, May 19, 2022): the OECD Frascati Manual 2015 definition and five criteria, and the FASB Statement No. 2 exclusion list.