Sell Dataset & Data Analysis Reviews From Your Bio

A dataset or data-analysis review is best sold as a priced custom request: the follower sends a dataset, notebook, or write-up, and the reviewer returns written or recorded feedback on methodology, data cleanliness, and whether the conclusions actually follow from the data, within a set turnaround — no live call or portfolio marketplace required. Custom Work lets you offer clearly scoped commissions and services: you set the price, the scope and the delivery time, and followers submit a brief and pay before you start.

No monthly fee, you only pay a small platform fee when you get paid. 8% platform fee when a follower pays for work, 5% platform fee on tips and goals. Card processing is separate. pricing

A familiar DM is: "Can you sanity-check my dataset before I present this?" Instead of squeezing that request into a free reply or pushing the person toward a booked call, a data analyst or researcher can turn it into a clearly scoped, priced offer on the link already in their bio.

The stakes of an uncaught error are not hypothetical. The European Spreadsheet Risks Interest Group, which audits real organizational spreadsheets, publishes this finding on its research and best-practice page:

"Human Error – To err is human, hence the majority (>90%) of spreadsheets contain errors." — European Spreadsheet Risks Interest Group, "Research and Best Practice"

A second pair of eyes before a dataset goes into a report or a deck is a real service, and delivering that service does not require a live session.

What counts as a dataset or analysis review?

A dataset or analysis review is written or recorded feedback on data or an analysis someone else has already produced — not a rebuild, not a new analysis from scratch, and not ongoing consulting. The follower sends the dataset, notebook, or report; the reviewer checks it and returns findings within a set delivery time, entirely in writing or on recording.

Scope discipline matters because review work drifts into open-ended work easily. A review answers questions like "is this sampling method sound," "did you clean this correctly," and "does the conclusion follow from the numbers." A review does not mean rerunning the whole analysis, building a new pipeline, or taking on the project as an ongoing engagement. If a request needs a working deliverable built or fixed rather than feedback on existing work, that request is a different kind of offer.

Checking someone else's work catches real errors without rebuilding anything. Nuijten and colleagues re-checked more than 250,000 reported p-values across eight major psychology journals and found that half of all published psychology papers using null-hypothesis significance testing contained at least one p-value inconsistent with its own test statistic, and that one in eight papers contained a grossly inconsistent p-value (Nuijten et al., "The prevalence of statistical reporting errors in psychology (1985-2013)," Behavior Research Methods, 2016).

Who typically sends this kind of request?

Dataset-review buyers are people who already hold data or a finished analysis and want an outside check before they ship it: graduate students preparing a thesis chapter, junior analysts before a stakeholder presentation, bootcamp graduates building a portfolio project, independent researchers, and small-business owners who ran their own numbers and want a gut check.

The pool of people who might send a dataset out for review is large and still growing. Kaggle's 2022 Machine Learning & Data Science Survey closed with 23,997 cleaned responses from data practitioners worldwide, according to Kaggle's own survey page. The U.S. Bureau of Labor Statistics projects that employment of data scientists will grow 35 percent from 2025 to 2035, with about 24,800 openings projected each year (BLS Occupational Outlook Handbook, "Data Scientists"). The Kaggle respondent count and the BLS projection both measure practitioner supply rather than demand for paid reviews, but a growing pool of students and early-career analysts is exactly the pool that asks working practitioners for second opinions.

Should this be a Paid Question or a custom request?

Custom Work fits requests that need a file: followers can attach a dataset, notebook, or report, and you can deliver text plus files and/or a link (how it works).

Both formats are enabled from the same Brillay page: Paid Questions handle one private question and one written reply, and Custom Work handles anything involving an attached dataset, notebook, or report.

Most dataset-review requests belong under Custom Work because followers can attach a dataset, notebook, or written analysis for review. Both Brillay offer types let you deliver on your own time within the delivery time you set — no live call, no scheduling.

What should a review actually check?

A useful dataset review covers three layers: the data itself, the method applied to the data, and whether the stated conclusion is genuinely supported by the result. Skipping any one of the three layers leaves the buyer with feedback that looks thorough on delivery but misses the layer most likely to contain the error.

The data layer carries measurable financial stakes. Gartner's own data-quality page states that poor data quality costs organizations at least $12.9 million a year on average, citing Gartner research from 2020 (Gartner, "Data Quality: Best Practices for Accurate Insights").

The conclusion layer — whether a stated finding actually follows from the numbers — breaks down often, even among working professionals. Nature's 2016 survey of 1,576 researchers reported the following headline finding (Baker, "1,500 scientists lift the lid on reproducibility," Nature, 2016):

"More than 70% of researchers have tried and failed to reproduce another scientist's experiments, and more than half have failed to reproduce their own experiments." — Monya Baker, Nature, "1,500 scientists lift the lid on reproducibility," 2016

Retractions of already-published research are climbing too: Nature reported that more than 10,000 research papers were retracted in 2023, a single-year record (Van Noorden, Nature, December 2023). A reviewer who checks whether a stated conclusion actually follows from the underlying data is doing the layer of work that a clean-looking chart hides.

Common things a reviewer flags: sampling that does not match the population being described, cleaning steps that silently drop or alter rows, mismatched units or date ranges, a chart type that misrepresents the distribution, and a conclusion phrased more strongly than the data supports (correlation stated as causation, a small sample generalized too broadly).

What should a dataset-review listing include?

A dataset-review listing should state six things before a follower pays: the input accepted, a size or scope limit, the deliverable format, the depth of the check, the delivery time, and the exclusions. Naming all six up front is what keeps a bounded review from drifting into an unpaid rebuild of the follower's analysis.

• Input: Dataset, notebook, code repository link, or written analysis, plus the question to answer.
• Size/scope limit: One dataset up to a stated row or file-size limit, or one notebook up to a stated length — not an open-ended audit.
• Deliverable: Written notes, an annotated notebook, a short voice walkthrough, or a combination.
• Depth: Methodology and conclusions only, or also re-running code to verify it reproduces the stated output.
• Delivery time: A period the reviewer can meet within Brillay's service delivery cap of 5 days.
• Exclusions: Rebuilding the analysis, additional statistical modeling, or ongoing advisory.

For example: "One round of written notes on a dataset up to 50MB or one notebook up to 500 lines. Covers sampling, cleaning, and whether conclusions match the data. Does not include rerunning the analysis or building new models."

Related reading: how many revisions a service should include covers setting that boundary before a follower expects unlimited follow-up rounds.

How do you price a dataset or analysis review?

Price a dataset review by the reviewer time it consumes, not by how technical the subject sounds. A narrow check on one chart takes far less time than verifying an entire analysis end to end, so most reviewers publish two or three tiers that scale with depth and delivery time rather than a single flat rate.

Preparing data before anyone can analyze it consumes a large share of practitioner time. Respondents to Anaconda's 2020 State of Data Science survey reported spending an average of 45% of their time loading and cleansing data before they could use it to develop models (Anaconda, "2020 State of Data Science"). A reviewer who spots cleaning problems early saves the buyer from repeating a large share of that data-preparation work. For a rough sense of what an hour of an experienced reviewer's attention is worth, the U.S. Bureau of Labor Statistics reports that the median annual wage for data scientists was $120,230 in May 2025 (BLS Occupational Outlook Handbook), which works out to roughly $58 an hour at 2,080 hours a year. Every price and delivery time example on this page is illustrative, not a guarantee.

How is a dataset review different from funding a new dataset?

A dataset review is paid feedback on data a follower already holds, delivered once within a set delivery time. Funding the creation of a brand-new public dataset is different work with a different payment shape: it collects money toward a goal instead of returning a deliverable. Both are data-adjacent, but the matching Brillay offer type differs.

A follower-funded open dataset joins a very large existing public corpus: Data.gov, the United States government's open data catalog, lists 556,482 datasets. On Brillay, funding a new open dataset is not an all-or-nothing campaign: every contribution is yours even if the project falls short of its target, with no deadline unless you set one and no rewards to ship. Keeping the three offer types distinct in the listing descriptions prevents a follower from paying for the wrong thing.

What else can you sell alongside dataset reviews?

A dataset review rarely needs to be the only paid option on a Brillay page. Tips, Goals, Personalized Videos, and Brand Collaboration Inquiries each match a different intent — thanks, funding, celebration, and partnership — from followers who already trust the reviewer's judgment about data.

• Tips: A way for someone who used the reviewer's free tutorials or open-source tools to say thanks without buying a review.
• Goals: Funding toward a stated goal, like building and releasing a new open dataset.
• Personalized Videos: A recorded congratulations for someone who just published a thesis or portfolio project.
• Brand Collaboration Inquiries: A separate intake for data-tooling companies or research organizations discussing a partnership rather than a one-off review.

These five offer types — Tips, Goals, Personalized Videos, Paid Questions, and Custom Work — run from a single Brillay page, priced the same way as described above (see how it works and pricing).

See dashboard reviews for a closely related offer when the follower's work is a BI dashboard rather than a raw dataset — the skills overlap, but the listings should stay separate so followers know which one to book.

How do you get a dataset-review offer live?

Getting a dataset-review offer live takes four decisions: whether incoming requests fit one private question with one written reply or need Custom Work for attached files, what the deliverable is, what size and scope limit applies, and what delivery time the reviewer can actually meet. Publish the listing once those four are written down.

Creators can enable Paid Questions and Custom Work regardless of degree or certification. That is a platform fact, not a claim about professional or academic norms in any specific field — describe actual background and experience level accurately in the listing itself. The same no-live-call format works for developers and technical reviewers publishing other Custom Work on their Brillay page.

Frequently asked questions

The questions data reviewers ask most are whether a credential is required, which offer type fits a file-based request, what to do when a submission exceeds the stated scope, how a dataset review differs from a dashboard review, and what selling one costs. Short answers: no credential, Custom Work, decline and refund, different artifact, and the platform fee described below.

Do I need a PhD or research background to sell dataset reviews?

You can enable Paid Questions or Custom Work regardless of degree or credential. Represent your actual experience accurately in your listing so followers know what kind of review they are paying for.

Should I use a Paid Question or a custom request for this?

Use a Paid Question when one written reply answers it without a file, such as whether a single test was the right choice. Use Custom Work whenever the follower needs to send a dataset, notebook, or write-up; Custom Work accepts attached files from followers and lets you deliver text plus files and/or a link.

What if the dataset or analysis is bigger than what I scoped?

Decline and refund a request that falls outside your stated size or scope limit, or point the buyer to a larger tier before starting the work.

How is this different from a dashboard review?

A dataset or analysis review checks methodology and whether conclusions follow from the numbers, typically in a notebook or written report. A dashboard review checks a built BI dashboard's layout and chart choices instead — keep the two listings separate.

What does Brillay charge?

Brillay's platform fee is the one stated near the top of this page and on the pricing page. There is no follower minimum, and payouts run through Stripe; Stripe's published US pricing for card processing is 2.9% + 30¢ per successful card charge (Stripe pricing).

Create your free Brillay page and turn the next "can you check my dataset?" DM into a clearly scoped paid review.