1

I have recently applied to a Job by a startup that wants to higher a data analyst. I have been going through the rounds and it seems like they want to hire me for the position. I have recently graduated with a bachelor in statistics and have not yet acquired working experience. However, this company has never had a data analyst and they are asking me questions that I do not have answers to yet because of my lack of work experience. It also probably helpful to mention that the company wants me to collect data from a mobile app. The questions are as follows:

  1. How are going to collect the data from the developer?

  2. How are you going to collect data without a third party (I think he means by not using google analytics or another tool like Cognos.)?

  3. What is the best tool for collecting data from apple?

  4. What are you going to do first when you have the data?

I have not been told what the objectives at the moment are from the company. I assume that I would use SQL to collect data from their database. I'm not sure how I am supposed to get data from the developer. I looked online about tracking mobile app metrics but I would like to know if there is a daily standard procedure that data analysts do when working with mobile apps? What kind of tools would you recommend to work with mobile apps? I know this question is kind of broad but I would love an experienced data analyst to tell a newbie how it goes. Especially since I have the feeling that they don't know either what a data analyst should do.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
Nick
  • 23
  • 4

1 Answers1

1

Welcome to the site, @Nick, and good luck with your job application. I will wade in regarding your forth question.

Whenever data are collected from scratch, which seems to be the case here, you have to understand first why the data are collected. Let's say in this case the data are collected to help the company predict which customers are likely to opt out of a service offered by the company. The why will drive how the data will be analyzed.

So before you even look at any data given to you, make sure you understand the questions the company would like to answer based on these data. For example:

  • Are men more likely than women to opt out of this service, all else being equal?
  • Are older people more likely to opt out of this sevice than younger people, all else being equal?
  • Are long-term customers less likely to opt out of this service than recent customers, all else being equal?

With these questions clear in your mind, the first thing you will have to do is try to import the data in your statistical software of choice (e.g., R). You may discover that the data are not properly formatted for import (e.g., they are saved in a spreadsheet with text annotations made by the developer). So formatting the data for import could be your first task (e.g., remove all text annotations from the data). Creating a back up copy of the original dataset is also a good idea!

Presuming the data could be imported in your statistical software of choice, the first thing you will want to do is to look at each of the variables of interest for your analysis and describe their distribution numerically and visually. For example, one of your variables could be Gender (a qualitative variable) - look to see how many people of each gender you have in your data set. Another variable could be duration of being a customer with the company (expressed in months) - a quantitative variable. So you could plot the histogram of duration, report the average duration (and a standard deviation), etc.

This initial exploration of the data will enable you to determine whether the data present any issues that need addressing:

  1. Do you have all the data you need for your analyses or are data on crucial variables or cases (i.e., customers) not available yet?

  2. Are the qualitative variables (e.g., Gender) coded properly? If not, will you need to clean up those variables to make sure their categories are coded consistently?

  3. Do the quantitative variables (e.g., Duration) present non-sensical values (e.g., someone has a duration of -12 months) or values that are implausibly large or small?

  4. Do any of the needed variables include missing values and if so, what is the extent of missingnes?

  5. Do any of the needed variables need to be recoded?

Isabella Ghement
  • 18,164
  • 2
  • 22
  • 46
  • 1
    Thank you for the detailed answer. I have an interview on Monday with the head honcho so this information will come in handy. – Nick Mar 21 '20 at 16:15
  • Ideally, you would want to keep an eye on the data collection process as it unfolds - do preliminary data look ok or are there any issues that need to be addressed as the data collection proceeds? Good luck on Monday! Let us know how you did! – Isabella Ghement Mar 21 '20 at 20:45
  • Don't forget to ask questions such as: "How have you collected data from the developers in the past? What worked and what didn't work? Can the past process be improved for this current task?", "Can we focus on collecting the most important data for this project in a first phase? It is always tempting to collect too much data, but that also reduces the chance of successful project execution", etc. – Isabella Ghement Mar 21 '20 at 20:52
  • Another thing you can answer for the 4-th question is "Have a meeting with the stakeholders involved to establish what is important to address from the collected data and come up with a plan of action - including timelines and deliverables - for the data analysis". This way, you see the forest before you get lost in the trees. – Isabella Ghement Mar 21 '20 at 20:55