Data is easy to collect; the hard part is knowing what to do with it. This is where business analytics comes in. The field combines quantitative thinking, business judgement, technology and communication, enabling organisations to understand what is happening and decide what to do next. For anyone considering this career path, the appeal lies in learning how to transform unstructured information into recommendations that can influence strategy, operations, customer relations and financial results.
What an analyst does beyond reports
Business analysis is often reduced to charts, reports and dashboards, which are merely ways of presenting information. The work begins with a question. An analyst might investigate why customer retention is falling, which products are becoming less profitable, where a process is slowing down, or how demand might change. The task involves finding the relevant data, checking its quality, identifying patterns, testing explanations and deciding what the results mean.
What a Master’s degree can offer
A master’s degree is one of the paths to analytics, and structured study helps those seeking deeper technical training or a closer link between data and management. A Master’s in Business Analytics involves advanced study, focused on applying quantitative methods to organisational problems. The value of the programme depends on what you need from it, so the curriculum matters more than the title. Look for opportunities to work with realistic datasets, to solve open-ended problems, to communicate results and to understand how analytical methods support business decisions.
Programmes with the same content have different names, ranging from a Master of Business Analytics to an MS in Business Analytics or a Master in Data Analytics. The list of courses and projects carried out using real data says more than the name. The best question to ask the admissions office is what job titles the most recent graduates were offered and how many of them responded to the survey from which the figure is taken. The programme catalogue also shows which languages and tools are used in the courses, a detail that the programme’s website often omits.
Technical skills need a business context
Analysts need sufficient technical expertise to work with data without hesitation: querying, cleaning data sets, statistical methods, models and visualisations. Technical expertise alone, however, does not indicate which problem deserves attention.
Good analysts understand how an organisation works. They consider revenue, costs, customer behaviour, risk, efficiency and conflicting priorities. If sales fall, the analyst first looks at what has changed in the data set, then investigates whether price, demand, customer experience, marketing or some other factor explains the trend. A useful overview of the subject, ‘Skills for a Data Analytics Career’, places communication, critical thinking and collaboration alongside technical expertise.
Analysis exists to solve costly problems
Organisations invest in analytics because uncertainty and inefficient decisions end up costing money. A poor forecast leaves stock unused, whilst a superficial analysis of customers wastes marketing budgets. Inaccurate performance indicators push teams to optimise the wrong things, whilst delayed reporting means managers only spot a problem once it has become difficult to fix. Business analytics can support decisions on pricing, staffing, forecasting, customer retention, operations, risk and resource allocation.
Some costly problems never make it onto a dashboard. Incorrect opening hours in a public listing send customers to a closed door, and the loss remains unmeasured until someone asks where the calls are coming from that are no longer turning into visits. It is strictly a data issue, as the public information about the firm has fallen behind reality.
Job titles tell only part of the story
Careers in analytics come with overlapping job titles, which makes job hunting confusing. Two roles with similar names may have completely different responsibilities. One role might involve reporting and performance indicators. Another might involve experiments, forecasting, process improvement or meetings with stakeholders. Some roles involve a lot of programming, whilst others focus on business requirements and communication.
Look beyond the job title. Pay attention to the problems the person hired will solve, the tools mentioned, the teams involved and how success is measured. Rather than learning every trendy tool, focus on the skills that recur in the roles you’re after.
Why searching by job title isn’t enough
Job titles overlap partly because official classifications lag behind the reality of the professions. The US government’s occupational classification system only assigned a separate code for data scientists – 15-2051 – in the 2018 revision, even though the role had been appearing in job adverts for years. The Bureau of Labour Statistics forecasts 34 per cent growth for this occupation between 2024 and 2034, with approximately 23,400 vacancies per year.
Anyone searching for a single job title will only see the adverts that use it. The same role may be listed as business analyst, data analyst, analytics consultant or insights manager, depending on each company’s practice. A search box responds to the specific term entered, whereas a classification also shows related roles: similar occupations, with their tasks and skills, as described by the US Department of Labour’s O*NET database. A candidate who starts with tasks and works their way up to job titles will find roles they wouldn’t have known to look for.
The candidate starts with three or four things they want to do every day – for example, forecasting or A/B testing – and reads up on which occupations these involve. The job titles appear at the end, as labels for the same work. The classification system also has its limitations: the categories are rarely updated, and new occupations remain under the old label for a while – as was the case with ‘data scientist’ until 2018. O*NET lists, for each occupation, the alternative job titles reported by employers, so the same page shows the other names under which the vacancies are advertised.
Companies face the same problem when looking for suppliers. A company needing someone to clean up its customer database does not know whether the service is called ‘data quality’, ‘CRM consulting’ or ‘business intelligence’. A business directory that classifies by category addresses this situation, as it shows what kinds of suppliers exist before the person searching even knows what they are called. In the Jasmine Directory, companies in the recruitment and employment sector are listed under the ‘employment’ category, and an editor reviews each listing before publication.
Good analysts learn to ask better questions
One of the most useful skills in analysis is recognising when the initial question is too vague. A request such as ‘find insights’ gives the analyst almost no direction. A better discussion brings to light the decision behind the request: is the aim to increase retention, reduce costs, improve conversion, forecast demand, or understand why performance has changed?
Before building a model, it helps to ask who needs the answer, what they can do with it, and how much uncertainty is acceptable. The best method is the one that reliably answers the right question, however simple it may be.
The third kind of error: the right answer to the wrong question
In 1957, Allyn Kimball, a statistician at Oak Ridge National Laboratory, published the article ‘Errors of the Third Kind in Statistical Consulting’ in the Journal of the American Statistical Association. Alongside the two classic errors of hypothesis testing, he identified a third: the error of giving the correct answer to the wrong question. Kimball attributed this to poor communication between the statistician and the person requesting the analysis, and called for future consultants to be taught to ask questions, in addition to learning computational methods.
In one of his examples, a consultant flawlessly applies a test that answers a different question from the one posed by the experiment, and both parties leave the meeting satisfied. In business analysis, the same error manifests as a precise model predicting an indicator that nobody uses in decision-making, or as a report on falling sales delivered in a month when management wanted to know whether to close a shop. Its everyday form appears in any search: whoever types in a specific word receives a precise answer to that word, even if their problem is called something else. A simple check is carried out before the first query: if no possible result would change the decision, the analysis can wait until someone formulates a question on which something depends.
The concept does not have a single definition. Frederick Mosteller had used the term in 1948 to mean something else—the correct rejection of a hypothesis for the wrong reason—whilst Mitroff and Featheringham broadened it in 1974 to include the incorrect choice of how the problem is formulated. Kimball’s text is an observation from a practitioner’s perspective and does not provide a method for measuring how often the error occurs; it therefore serves as a test question.
Clear communication drives analysis towards action
A conclusion is of little value if the decision-makers do not understand it. Analysts often have to explain technical results to people who do not work with data on a daily basis, which means deciding what information matters, stripping out unnecessary detail, and showing why the result changes a real business decision.
A piece on data storytelling for strategic decisions helps to link the analysis to clearer recommendations and actions. Good communication gives people enough context to understand the result, its limitations and the decision it supports. A simple explanation, backed by sound analysis, is often more useful than a complicated presentation that nobody can interpret with confidence.
Judgement matters when data is imperfect
Real-world data sets are rarely clean, complete or neutral. Missing information, measurement issues, biased samples and weak assumptions can alter the conclusion. Before trusting a result, ask where the data comes from, what might be missing, whether the measurement reflects what you’re interested in, and how sensitive the conclusion is to your assumptions.
Ethical judgement also matters, as analytical decisions can affect customers, employees, prices, access and opportunities. A model that looks good statistically can lead to poor results if the underlying data or objective is flawed. Responsible analysts explain the uncertainty, however precise the figures may seem.
How to check the source of a list of companies
Lists of companies are a handy example when considering the question of where data comes from. A list of companies that has been purchased or extracted automatically usually contains the same company multiple times, under slightly different names, alongside defunct companies and out-of-date addresses. An analyst counting potential customers on such a list obtains an exact figure for the wrong question – that is, Kimball’s error. If one-tenth of the entries are duplicates, the market appears one-tenth larger than it actually is, and the error goes unnoticed throughout all subsequent calculations.
The source indicates in advance just how much cleaning will be required. A database automatically compiled from other databases inherits the errors of each one, and matching records by name, address and telephone number becomes the first stage of any analysis. A list where a person has visited each company’s website before including it is smaller and grows more slowly, but every entry has been viewed at least once by someone. The Jasmine Directory blog features an article on the difficulties of data collection in digital markets, which is useful for anyone working with such sources.
An edited directory confirms that the company exists, that it operates in the category in which it appears, and that it can be found again. Turnover, the number of employees and the quality of services are not included there, and an analyst who needs this information obtains it from the commercial register of the country in which the company is registered or directly from the company itself. For a market analysis, the procedure remains the same regardless of the source: first, you establish what constitutes a distinct company, then you count them.
How to check a company before an interview or signing a contract
A novice analyst also has some checking to do: the companies they are applying to. Anyone can publish a job advert, and the job title says little about who is paying the salary. The national company register shows whether the company exists and for how long; the firm’s careers page shows whether the vacancy is listed there too; and the name, address and email address in the advert must be the same across all these sources.
Independent consultants also use these same sources before entering into their first contract. A client who appears in the register under a different name to the one on the invoice, or whose address varies from one website to another, warrants a follow-up question before work begins.
Gather evidence that you can do the job
Practical projects demonstrate how you think. Choose a real-world problem, work with a credible dataset, document your reasoning and explain what decision could support the results. The project doesn’t have to be huge to be useful. Employers want proof that you can define a problem, handle unstructured information, recognise limitations and communicate a conclusion.
A portfolio faces the same problem as a small business: it exists, but it needs to be found and believed in by someone who doesn’t know you. The recruiter searching for your name compares what they find across three or four websites, and projects described differently from one profile to another give them cause for doubt. A project with public data, open-source code and a conclusion summarised in a single paragraph can be verified in a matter of minutes – something that’s impossible for a list of tools on a CV.
The name under which you appear matters too. A consultant listed in one place under their own name, in another under their company name, and in a third under a pseudonym from a coding platform appears, to the recruiter, as three different people. A single form of the name, the same brief description and the same contact details link the profiles together without the recruiter realising.
For an analyst working independently or for a small consultancy firm, a third-party verification serves the same purpose as a degree does for a graduate. A listing in a published directory, under the appropriate category, tells a client that the firm has been vetted by a human being before publication. The directory’s guide on the effectiveness of listings for business owners shows how to measure whether such a presence delivers results. The listing shows that the firm exists and where it can be found, whilst the quality of the analysis is demonstrated through projects.
Business analysis lies between information and action. Tools and job titles will change, and methods will become more sophisticated. What remains is the skill of asking better questions, weighing up the evidence carefully, and turning your findings into decisions that people can use.
