Do you want a job in analytics too? Part 1 – the skills you need


So you’ve decided that you want to go into analytics? Great. Now it’s time to identify the skills needed for the specific job you are targeting and create a plan for developing them.  I will cover this in three posts; the hard skills needed to perform the job, the soft skills needed to perform the job well, and the experience that will convince a hiring manager that you’re the best candidate for the job.


Analytics jobs break down into five categories.  Identifying what category you are targeting will help you determine the path forward.  Some roles are easier to access than others. If you are targeting an execution role you may be able to build those skills quickly on your own.  If you are looking for a data scientist / modeling type role, the necessary skills will take years to build and may require formal schooling.


There are lots of options in terms of specific programs or types of analysis in which a candidate can be skilled.  We can break it down into three main components: basic, standard, and advanced level skills.


Basic skills


Basic skills are just that: the bare minimum.


The first is Microsoft Excel.  Excel is capable of doing all sorts of analyses as well as being an excellent tool for sharing data and presenting findings in graphical formats.  I covered why Excel is necessary here.  I will continue to update Excel content and case studies here.


Knowledge of SQL is the second skill that is necessary to become an analyst.  I covered why that is the case and the easy way you can learn it here.


The final basic skill is working knowledge of at least one analysis program/language.  There are dozens of programs to choose from, some are more relevant to specific jobs than others.


I want to include statistics as a basic skill, but I’ve met plenty of analysts who lack basic statistical understanding.  I do feel that developing statistical skills will help set you apart.


Standard skills


Standard skills are more specialized than basic skills and can differ based on specific roles.  Not all jobs will require all skills, and most hiring managers will have their own ideas of how much variation from a job description is acceptable for a candidate.    However, I feel comfortable in saying that there are two skills that analysts will have more often than not.


Statistical knowledge is the bedrock of analysis and a certain level of understanding will be beneficial.  The specific level and type of knowledge will depend on the requirements of the role.  For instance, six sigma master black belts must understand the normal and exponential distributions in depth, but can typically ignore the other distributions.  Operations analysts should have strong knowledge of standard hypothesis testing techniques such as t-tests, and possibly linear programming. Despite different qualifications, there is a standard level that all analysts must know.  The ability to use test hypotheses, evaluate sample size, and calculate summary statistics are probably the average capabilities in the market.


The more robust statistical knowledge a candidate has, the more attractive they will be to most employers.  However, it is imperative that they can show how it would be relevant to the position.  This is the primary difference between an academic approach and a business approach.


I once personally failed the academic vs. business approach when I was unable to convince my business to improve the way we performed a simple test.  I suggested a more complicated, but also more accurate measure.  However, the business case for switching was weak; it would not drive action that would save money.  Few, if any, of the people receiving the reporting would understand the difference and it was likely to confuse the execution analysts updating the code.


Fluency in one analysis program, including the ability to write commands with code, is the standard level of analytical programming required.  It is an important distinction to be able to code rather than simply use a drag and drop GUI analysis.


In the future, coding knowledge may not be required as better GUIs are developed for robust analysis languages and programs.  However, with more new graduates than ever before learning applicable languages such as R in school, it could go the other way and become a basic level skill.


I have often heard people say that learning to speak other languages is easier if you already know one.  The same trend applies with programming languages, probably more so.  If you know R, learning SAS or Python will be much easier.


The ability to write code in one programming language is often enough to get a job in a role where a different language is used.  There is not enough good analytical talent in the market for companies to turn away skilled analysts simply because they know R instead of SAS.


Note:  Some recruiters, companies, or managers will insist on experience in a particular language/program.  In some instances where expert level proficiency is required it may make sense, but it is typically the sign of a poor talent strategy.  I have personally been “hurt” by this; in reality it was a blessing because I wound up in a better role with a much stronger manager. 


Advanced level skills


Multiple advanced skills are what separate typical candidates or analysts from exceptional ones.  Not every candidate or analyst will have a skill in this list, but those that have multiple skills in the list will typically be sought after.


Linear and logistic regressions are forms of explanatory or predictive modeling.  They can be complex, but their simpler versions can be created in Excel.  Large or robust models can take months to build using more complex programs.  Knowledge of how to create a simple linear regression is not much of an advanced skill, but experience with complex modeling can be sought after.


Linear programming is foundational to operations research and is common in scheduling, manufacturing, and supply chain problems.  Linear programs can be incredibly mathematically complex, or require so much processing power they are unsolvable.  The travelling salesman problem is the best known example of a problem too hard to solve.  In order to get around this issue heuristics are often used.


Note: You should be able to tell the difference between an algorithm and a heuristic


The ability to automate analyses or scripts to update or run is an attractive skill for an analyst to have.  In any organization an analyst’s workload can spiral out of control. The question is can you automate as many recurring tasks as possible so that you spend most of your time on analyses that add new value?


Automating analyses is also important in any business that depends on multiple inputs to make decisions.  Banks often have “production” processes which score customers and determine what offers they receive.  Supply chains will often have automated and detailed routing of products that will say exactly where in a warehouse or truck they are to be loaded.  This routing updates frequently based on actual and forecasted sales and deliveries.  Those processes need to be automated.


You don’t need to be able to write automated scripts to be useful as an analyst.  You will deliver huge value if you understand the impacts that operational changes will have on the automation, or vice versa, and can stop problems before they start.  On the flip side it is also valuable if you can track down the root causes when something falls through the cracks.


Machine learning is a new-ish and exciting area of analytics.   Neural networks, random forests, and gain boosting models are interesting ways of using powerful computing to improve predictions.  With the increasing quantity of data becoming available, the importance of having tools that can quickly process huge amounts of data will increase.


Data mining is another new-ish area of analytics that can be attractive to employers.  SAS Enterprise Miner is a relatively common program which can rapidly produce insights, you might see it listed in job postings.


Note: I started writing a description of how data mining and machine learning are different, but I quickly realized it was too big for this article.  Stand by for future content describing the differences. 


Hadoop can handle unstructured data so it is desirable to companies which seek to gain insight from things like photos and media, e.g., a company trying to gauge social media reactions.  Hadoop analysts and developers are in short supply and well paid.


Un-orthodox data pulls, such as website scrapes, can also deliver huge value.  The amount of captured data is exponentially increasing, but the ability to effectively gather non standard information will always deliver value.


Beyond regression, there are other advanced statistical analysis methods used in business.  Analyses such as conjoint, ANOVA, cluster, factor, etc., are all valuable and will help get you interviews.  If you can demonstrate having used advanced statistics to add value it will make you attractive, whether the position requires it or not.


What is better, more skills or deeper knowledge of a skill?


It depends.  That’s the consultant’s answer for everything, right?


For me, the answer depends on the type of role.  For a practitioner role in an organization which uses SAS, which would you rather hire?


Programmer A who has three years of in depth experience in R but has never used SAS




Programmer B who has done work with SAS, R, and Python but never in significant detail


I would take Programmer A.  I would rather see someone have deep expertise in a relevant skill than to simply be knowledgeable in the exact skill I am looking for.


However, if I were looking for someone to manage a team of analysts I would rather have someone with a broader skill set.  Unless I need the team to be lead by a technical expert to work on difficult issues, I would rather have a manager who thinks about issues broadly and can recommend new approaches.  I also think that soft skills are more important for a manager, as well as management skills of course.


In my experience, every company places far too great of an emphasis in recruiting on subject matter knowledge rather than talent, integrity, and drive.




My lists of skills were not intended to be exhaustive, but to give you an idea of what skills people who currently work have and what skills employers like to see.  Obviously having more skills is better, especially if they are advanced skills.  However, if you are a job seeker and lack many of the advanced skills, do not worry, your competition probably does as well.


Hard skills can tell a manager your ability to perform the job.  However, soft skills are what tell them whether or not you can do the job well.  Hard skills get you into the interview, but soft skills get you the job.  Stay tuned for future posts about what those soft skills are and how you can demonstrate them.

What do you think?  If you have a different take please leave a comment below.  If you agree with me I would love to hear about it also.






3 thoughts on “Do you want a job in analytics too? Part 1 – the skills you need

  1. I would love to read an example of the automation you touched on brings all of the parts together. Sorry if this is naive but maybe an example would be using SAS data miner to build a scatter plot of demand for an industry, using excel to find a line of best fit, using the line to build a optimization model, and using python and SQL to automatically order and deliver products. The have this process repeat weekly. Am I off the mark here? Maybe this will be a bit more clear when I read through some case studies.

  2. That is an interesting idea for a future post, let me think of something and write out a story.

Comments are closed.