Bifalgorithm post

This is a tricky case study. I've had this on interviews and real-world projects. The marketing team wants to put up a billboard. How do you estimate the ROI or the impact? Many people will just go silent and say I need to think about the question. Absolutely do not do that. Do this instead. While you're thinking about the solution, repeat the problem back to them. Okay, so marketing is interested in putting up this billboard and they're interested in their return on investment, whether it be visitors or revenue or something else. And at that point, start thinking about the metrics. Next, describe the simplest approach to this type of problem. In this case, it would be an A/B test. That's for a general type of impact problem. Obviously, not this one specific. So you might say something like this. The gold standard, when we want to know the impact of something, it's to run a randomized experiment. But an experiment in this case would be difficult because and then go into the reasons why. This is a challenging problem and you're not going to be able to solve it in your 20 to 30 minutes of interview. The key here is to describe to them how you would think about such a problem which is impact estimation. And next, can you identify the problems like what makes experimentation challenging? What are the obstacles that you have to overcome and then just come up with something that is that is plausible, but be humble about it. You are not going to come up with the correct answer in this 20 or 30 minutes. One example might be geo-experimentation. You might use a synthetic control. Or if you had really precise geolocation data, you might do some propensity score matching. But that's almost not really the point to get it perfect. It's can you identify the problem, can you identify the challenges and at least come up with some reasonable solutions for a 20 or 30 minute case study.

This is a tricky case study. I've had this on interviews and real-world projects. The marketing team wants to put up a billboard. How do you estimate the ROI o…

Ideas in this flowchart

  • Case Study Approach
  • Initial Steps
  • Restate the Problem
  • Identify Metrics
  • Simplest Approach
  • A/B Testing
  • Randomized Experiment
  • Challenges in Experimentation
  • Time Constraint
  • Problem Identification
  • Experimentation Obstacles
  • Geo-Experimentation
  • Synthetic Control
  • Propensity Score Matching
  • Reasonable Solutions
Bifalgorithm
Sign up
Feed/

This is a tricky case study. I've had this on interviews and real-world projects. The marketing team wants to put up a billboard. How do you estimate the ROI or the impact? Many people will just go silent and say I need to think about the question. Absolutely do not do that. Do this instead. While you're thinking about the solution, repeat the problem back to them. Okay, so marketing is interested in putting up this billboard and they're interested in their return on investment, whether it be visitors or revenue or something else. And at that point, start thinking about the metrics. Next, describe the simplest approach to this type of problem. In this case, it would be an A/B test. That's for a general type of impact problem. Obviously, not this one specific. So you might say something like this. The gold standard, when we want to know the impact of something, it's to run a randomized experiment. But an experiment in this case would be difficult because and then go into the reasons why. This is a challenging problem and you're not going to be able to solve it in your 20 to 30 minutes of interview. The key here is to describe to them how you would think about such a problem which is impact estimation. And next, can you identify the problems like what makes experimentation challenging? What are the obstacles that you have to overcome and then just come up with something that is that is plausible, but be humble about it. You are not going to come up with the correct answer in this 20 or 30 minutes. One example might be geo-experimentation. You might use a synthetic control. Or if you had really precise geolocation data, you might do some propensity score matching. But that's almost not really the point to get it perfect. It's can you identify the problem, can you identify the challenges and at least come up with some reasonable solutions for a 20 or 30 minute case study.

E
Ethan Spencer
8w

"This is a tricky case study. I've had this on interviews and real-world projects."

👍
0
0 comments0 shares

Sign in to view and post comments.