The Data Analyst – Interview Prep & Skills

Data Analyst;
Interview Preperation;
Skills;

Thoughts

As a biostatistician role, we also play Data Analysts. With comprehensive and deep knowledge of statistics, we apply available methods and choose the proper method or, if necessary, modify/create a new approach to drive the data into a story that fits.

In general, Steve Olson’s Talk (CCASA Presentation) could tell us about essential requirements to prepare skills and interview for the data analyst.

Outline

1- Goal of the Data Analyst

2- Hard Skills

• Understanding the Data
  
• Approach to Analytics
  
• Creating Valuable Analytic Output
  
• Test and Learn Principles
  
• Skills & Tools

3- Soft Skills

• Valuable Analyst Attributes
  
• Questions to Ask

Goal of the Data Analyst

The Goal of the Analyst is to generate insights from available data in service of improving the company or brand Key Performance Indicators (KPIs), goals, metrics

• Understanding a brand or company’s KPIs and/or goals is foundational as all analytics are conducted in service of KPIs and identifying opportunities to improve the metrics
  • Commerce: Monthly sales, revenue, margin or profit, steal share from specific competitors
    
  • Marketing: Engagement (CRM) such as email open rates or click through rates, revenue, new names generated
• Examples of questions to be answered through analytics:
  • How are the KPIs trending?
    
  • Where can we increase sales?
    
  • Where is engagement strong or lacking?

How to understand the Data

All analytics start with a full understanding of the underlying data

• How to “break down” a dataset:

  1- Know the exhaustive set of variables and data with which you are working

  • Data dictionary is sometimes a good starting point – often will just be high level description and format
    
  • How are different data sets linked? What are the primary keys? (individual id, phone number, store id)

  2- Calculate basic metrics to fully understand the data – up to 90% of analytic work – Discovery or EDA

  • Calculate means and medians for each variable
    
  • Create distribution metrics of each variable: standard deviations, interquartile ranges etc
    
  • Data ranges
    
  • Survey data – know the possible answers to each question, the scale
    
  • Uncover questionable data: missing data, miscoded or incorrect data, outliers
    
  • Create cross tabs

  3- Visualize the distribution of data in the variables

  • Create graphs of distribution frequencies, histograms

Approach to Analytics

A single number or metric or datapoint is not an insight

Metrics require CONTEXT to know it it is “good or bad”

Analytic approaches to create context:

1- Benchmarking against past similar timeframes or events

• Compare to last year (YoY), Past promotion comparison, Previous month average

2- Trending over time – is the metric increasing or decreasing?

• line graphs of : past 12 months, year over year
  

• cohort analysis – Q1 engaged, did they engage in Q2, Q3 etc
  3- Segmenting the data: how does the metric look for different segments?

• Age groups
  

• Purchase segments
  

• High, medium, low engagement tiers

4- Industry benchmarks or comparisons

5- Create maps to determine if there are geographical differences

6- Cumulative response distributions - linear growth or plateau

7- Test results – how did group A perform vs group B?

Creating Valuable Analytic Output

• Think about the numbers in terms of telling a story
  • “The metric appears to be trending in this direction, this might be why, and here is what I think we should do”
    
• Make recommendations – what recommendations do you have because of the story in the data and how might we alter the KPIs ore metrics favorably?
  • Because the metrics are doing this, we should consider doing this
    
• Generate hypotheses – if information is lacking to make a firm recommendation – why might things be trending the way they are – be provocative
  • The metrics may be trending this way because
    
  • Is there other data that could help explain what you are seeing?
    
• Propose testing – where appropriate to evaluate hypotheses – can a test be set up to confirm or disprove the hypothesis?

Test and Learn Principles

• Identity A) the metric to change and B) the feature(s) which, if changed, may improve the metric
  • Increase open rates: modify subject lines/preview text
    
  • Increase online conversion/purchase: add purchase message pop ups or abandon cart messaging
    
  • Increase sales: increase discount amounts
    
• Generate a hypothesis – if we change this, this will happen
  
• Understand the engagement or purchase funnel
  • Is there enough sample to get a read on results?
    
• Statistical testing
  • Understand statistical vs. practical significant – effect size vs. P values
    
  • Is the result sizeable enough to warrrant making the change?
    
  • Random sampling is vital

Skills & Tools

Speak the “language” of the different tools you might use – may not need to be an expert

• Know the basic functions within SQL ana what they do
  • select, from, where
    
  • different joins and what they do – frequent interview question
• Familiarize yourself with different analytic tools
  • Visualization tools: Excel, Tableau, PowerBI
    
  • Web analysis: Adobe analytics, Google analytics
    
  • Statistical languages: R, Python
• Prepare an example for when you learned something similar quickly to be successful

Soft skills

• Attributes of the Valuable Data Analyst
  • Curiosity - about the business and the metrics
    • You are hungry to understand what is the data telling you
      
    • You have an appetite for knowing what you can learn from the data
      
  • Creativity – with data and output
    • Excited to explore different ways to visualize data
      
    • Ability to explain the findings – you may need to present the data to non-data experts
      
  • Attention to detail
    • Data errors are easy to make – always double check the work and get others’ feedback
      
    • Code running without error is not enough – validate the output
      
  • Ability to have “Big Picture” view
    • Step back and ask yourself if the numbers you are presenting make sense
      
    • Gui check – refer to other work, as others for input

Questions to ask during the interview

• What kind of data will I be working with?
  • CRM engagement, sales data, media spend, what level (household, individual, store)
    
• How is the data structured?
  • Relational tables?
  • What are the different datasets?
    
  • How are different tables linked? (primary keys)
    
• What are the Key Performance Indicators we are trying to improve?
  
• Will I be presenting the data and results and who is the audience?
  • Will I be working with others to tell the story?
    
• What analytic tools are used most often in daily work?

Sum up

Source: dasca