Industries / Finance and business services / Legal and professional services / Consulting
Research and analysis
Research and analysis is the desk work behind advice: gathering facts about a client's market and rivals, estimating how big a market is, and running the facts through frameworks that suggest what to do.
7
Milestones achieved, from 1914 to 2025.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1914
Research sold as a service
Edwin George Booz founded a firm in Chicago in 1914 under the name The Business Research Service. In the firm's account, its first major assignment was for the Illinois State Railroad, researching how customers would react to a rate increase, and in the 1920s and 1930s it sold management consulting, organization and market research services to companies such as Goodyear, General Mills and Montgomery Ward. The firm is now Booz Allen Hamilton.
Credit
Edwin G. Booz
1970
Four boxes for products
Bruce Henderson, founder of the Boston Consulting Group (BCG), argued in "The Product Portfolio" that a company should hold products with different growth rates and market shares, because fast-growing products need cash. He sorted products by market share and market growth into "stars" (high share, high growth), "cash cows" (high share, slow growth), "pets" (low share, slow growth) and "question marks" (low share, high growth). BCG calls this the growth-share matrix.
Credit
Bruce Henderson, Boston Consulting Group
1979
Five forces on profit
Michael E. Porter, then a young associate professor at Harvard Business School, published "How Competitive Forces Shape Strategy" in the March–April 1979 Harvard Business Review. It holds that competition for profits comes not only from established rivals but also from customers, suppliers, potential entrants and substitute products, and that knowing these forces "can help a company stake out a position in its industry that is less vulnerable to attack." Porter restated the five forces in the same magazine in 2008.
Credit
Michael E. Porter
1979
Measured against rivals
Xerox found that Japanese competitors sold copiers at prices well below Xerox's own production costs, and began comparing its costs and methods with theirs. The Global Benchmarking Network, a group of national benchmarking centers, says Xerox coined the term benchmarking for this in 1979 and was running benchmarking studies in all its business areas by 1981. Robert C. Camp, a Xerox manager of business analysis, co-wrote a 1987 Harvard Business Review article on the work and a 1989 book that spread the method.
Credit
Xerox; Robert C. Camp
1984
Excellence that did not last
In Search of Excellence, by Thomas J. Peters and Robert H. Waterman Jr., who had led a McKinsey & Company task force on organizational effectiveness, held up 43 model companies and stayed on best-seller lists for more than a year. On 5 November 1984 Business Week ran the cover "OOPS!" and reported that 14 of the 43 had run into trouble.
LessonThe companies doing well today, and the traits they share, do not show which companies will do well tomorrow.
Credit
Business Week, as told by Peter Baida in American Heritage (1985)
1994
The matrix misleads in tests
J. Scott Armstrong and Roderick J. Brodie asked 1,015 people, working alone in the role of managers, to choose between an investment that would double and one that would lose half its value. Six administrators ran the experiment 27 times in six countries over five years. Of the subjects shown the BCG matrix, 64% chose the unprofitable investment; of those who used the matrix in their analysis, 87% did.
LessonA popular framework can steer people to a worse answer, so it is worth testing before it is taught.
Credit
J. Scott Armstrong and Roderick J. Brodie
2025
Models graded on consulting cases
Mercor, which recruits professionals for AI work through its own platform, published the AI Productivity Index (APEX) in September 2025 and extended it in December 2025. Experts with an average of more than seven years in their industry wrote 100 held-out management consulting cases from their day-to-day work, each with about 15 pass-or-fail criteria, and an AI model, Gemini 2.5 Flash, graded the answers. On the consulting cases the best model, Gemini 3 Pro, met 64.0% of the criteria and GPT-5 met 63.1%.
Credit
Mercor (Bertie Vidgen, Brendan Foody, Osvald Nitski and 17 co-authors)
Watch for
Due diligence for a tenth
Commercial due diligence is the study of a company's market and customers that a buyer orders before a buyout. DiligenceSquared, a start-up from Y Combinator's Fall 2025 cohort, says it can deliver this analysis for about $50,000, where TechCrunch reports that private equity firms can pay $500,000 to $1 million for McKinsey, Bain or BCG to interview dozens of a target's corporate customers. Its AI voice agents conduct the customer interviews, and senior human consultants check the final output.
Achieved whenan outside party compares its reports with those of established firms on the same deals and publishes the results.
Claimed by
DiligenceSquared, reported by TechCrunch, March 2026
Open
01
A forecast scorecard
A public table of market forecasts that consulting and research firms published for a target year that has now passed, each set against the figure that came true. Every row links the forecast, states exactly what was counted, gives the actual figure from an official or audited source, and shows the error.
The briefCredit · unclaimed02
A market-sizing answer key
A public set of market-sizing questions of the kind consultants answer, such as how many hair salons a country has or what its dental offices earn in a year, each with an answer from official business statistics. People and AI models estimate each answer without looking it up, and their errors are scored.
The briefCredit · unclaimed
Open 01 · the brief
A forecast scorecard
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public table of market forecasts that consulting and research firms published for a target year that has now passed, each set against the figure that came true. Every row links the forecast, states exactly what was counted, gives the actual figure from an official or audited source, and shows the error.
Why it is still open
Dated forecasts are easy to find. Cisco's Internet Business Solutions Group predicted in April 2011 that 50 billion devices would be connected to the internet by 2020. A self-published 2025 brief on Zenodo reports a "25% hit rate" for McKinsey, BCG, Bain and Gartner forecasts but publishes neither the forecasts nor the outcomes nor how they were scored, and the closest open project, NMAI Research's scorecard of data-center electricity forecasts, says it makes no accuracy ranking.
Built with what exists
the firms' published reports and press releases; official statistics such as US Census Bureau and Eurostat tables; a current language model to pull each forecast's number, unit, scope and target year into a table; a spreadsheet that a second person checks.
First step
Collect 20 forecasts whose target year fell between 2020 and 2025 and find the actual figure for each.
Done looks like
At least 200 forecasts from at least 10 firms, each with its outcome and sources, a second person agreeing on 95 of 100 rows chosen at random, and the median error by firm and by how many years ahead each forecast looked.
Open 02 · the brief
A market-sizing answer key
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public set of market-sizing questions of the kind consultants answer, such as how many hair salons a country has or what its dental offices earn in a year, each with an answer from official business statistics. People and AI models estimate each answer without looking it up, and their errors are scored.
Why it is still open
The closest public set, the Fermi Problems data of Ashwin Kalyan and four co-authors (2021), holds about 1,000 real-world estimation problems taken from quizzes and olympiads, not business statistics. An open-source project builds market-size estimates from public filings and census data but does not score them against held-out answers.
Built with what exists
the US Census Bureau's Economic Census and County Business Patterns, Eurostat's structural business statistics and other countries' official tables, using recent releases so the answers are less likely to be in models' training data; current AI models; a short scoring script.
First step
Write 25 questions from one country's latest business census and time how long five people take to answer them.
Done looks like
200 questions covering at least five countries, each answer linked to its official table, and published scores for at least five AI models and 20 people, given as the share of estimates within a factor of two of the official figure.