In a 2011 talk and interview, Amazon principal engineer John Rauser described a data scientist as someone who combines applied mathematics, engineering, communication, skepticism, and curiosity. The model is less a job title checklist than a way to explain how useful data work happens: obtain and manage evidence, analyze it rigorously, test conclusions, understand the domain, and explain the result clearly.
What is a data scientist?
Rauser’s definition came from a 2011 presentation at the O’Reilly Strata Conference in New York and a contemporaneous account by Dan Woods published by Forbes on October 7, 2011. It treats data science as a combination of technical and human capabilities rather than a single academic specialty.
His first two pillars are applied mathematics and engineering. Mathematics supplies statistical reasoning for turning observations into insight. Engineering—including programming—makes it possible to acquire, clean, store, manage, and investigate data directly. Rauser’s ideal practitioner combines an engineer’s ability to work with large datasets and a statistician’s ability to extract value and present it to an audience.
Three further traits complete the model: communication, skepticism, and curiosity. Together, the five dimensions describe what a data scientist needs to do, not just what the person is called. Later institutional discussion noted that the term had spread across fields and sectors with varying interpretations, so Rauser’s framework should be read as his perspective, not a universal occupational standard.
Free tools Windows power users keep installed
One-click scans. No signup required.
#1 Best Overall
What does a data scientist really do?
Use mathematics to turn data into evidence
A data scientist chooses quantitative methods that fit the question, accounts for uncertainty, and distinguishes a pattern from a convincing explanation. In Rauser’s account, statistical knowledge is the means of extracting value from observations rather than an end in itself.
Engineer access to the data
Analysis cannot begin with data that nobody can obtain or trust. Engineering skills help a practitioner write programs, work with large datasets, organize inputs, and investigate a question hands-on. This is why Rauser placed engineering alongside mathematics instead of treating it as a separate support function.
Explain the result in writing
Communication determines whether an analysis can influence anyone outside the person who performed it. Rauser gave particular weight to writing, including documentation that a reader may encounter later. His reported maxim was: “If it is not written down, it never happened.” That quotation appears in Woods’s 2011 report and should be understood as an attributed statement, not a verified transcript of the original recording.
Try to disprove the conclusion
Skepticism means actively searching for evidence that could invalidate a thesis, not merely collecting confirming examples. Rauser also recommended checking surprising findings through more than one approach. As Woods reported his words: “If you have a healthy skepticism, you will look as hard for evidence that refutes your thesis as you will for evidence that confirms it.”
Learn enough about the domain to ask better questions
Curiosity pushes the practitioner beyond the supplied table or dataset. It includes learning how the subject works and finding the query, measurement, or experiment that can clarify the real problem. Without that interest, technically correct analysis can answer the wrong question.
Rauser’s five dimensions at a glance
| Dimension | Role in the work |
|---|---|
| Applied mathematics | Uses statistical and mathematical reasoning to convert observations into insight. |
| Engineering | Acquires, manages, programs, and investigates data, including large datasets. |
| Communication | Explains findings clearly, especially in durable written form. |
| Skepticism | Looks for disconfirming evidence and validates unintuitive results with multiple approaches. |
| Curiosity | Builds domain understanding and identifies questions that make analysis useful. |
Why Tobias Mayer appears in the definition
Rauser used the eighteenth-century German astronomer Tobias Mayer to illustrate how mathematical reasoning and practical familiarity with observations can work together. In Woods’s account, Mayer tracked the apparent motion of the lunar crater Manilius as evidence about the Moon’s libration, its apparent wobble.
The account says Mayer had 27 observations for a problem involving three unknowns, arranged in three groups of nine. Rauser regarded this as an early quantitative argument for collecting more observations and called Mayer “the first data scientist in my mind.” That is Rauser’s historical interpretation, not an established consensus that the modern occupation began with Mayer.
The numerical lesson also needs care. Woods reported that Mayer said nine times as many observations made the result nine times as accurate, but explained that the improvement would be three times at best under the square-root relationship being discussed. That historical example should not be repeated as a universal statistical rule: the relationship between sample size and accuracy depends on the estimator, noise, design, and assumptions.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →What skills does a data scientist need?
Rauser’s answer is deliberately cross-disciplinary. A strong profile balances the following capabilities:
- Statistical and mathematical reasoning sufficient to model uncertainty and evaluate evidence.
- Programming and data-engineering ability sufficient to obtain, prepare, manage, and inspect data.
- Clear writing and presentation for technical and nontechnical readers.
- A habit of testing alternative explanations and seeking evidence against a favored result.
- Curiosity about the application domain and the questions hidden behind the initial request.
The balance can vary by role. One team may need deeper modeling expertise; another may need stronger production engineering or domain knowledge. Rauser’s point is that dependable data work suffers when any one of these dimensions is ignored.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How did Rauser suggest people learn the work?
Woods reported that Rauser had studied aerospace engineering and computer science, worked as a software engineer, and later taught himself analytical techniques such as statistical modeling. The advice presented in 2011 was to supplement computer-science training with machine-learning study and to recognize that promising engineers or statisticians might need to grow into broader data-science responsibilities.
That is historical advice, not a current hiring rule for every organization. It does, however, illustrate the framework’s central idea: data science can be developed by combining existing strengths rather than waiting for a candidate who already matches every discipline perfectly.
How this 2011 definition fits the modern title
“Data scientist” now covers jobs with different mixes of experimentation, statistical modeling, machine learning, analytics, software, and domain work. A person using the title in one industry may spend most of the day building data pipelines; another may design experiments or communicate findings to executives. The term’s breadth is consistent with the interpretive variation documented by a 2012 Microsoft Research event page.
Rauser’s five-part model remains useful as a diagnostic. When evaluating a role or a candidate, ask whether the work requires mathematical depth, engineering and programming, written communication, skeptical validation, and domain curiosity—and which of those areas the team already supplies. It is a way to clarify expectations without pretending that one job description defines the entire profession.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
