HR

Would you pass the high end of our Finance interview questions @Acellere?

We typically give a few of funny exercises to our potential hires that are focusing on their skill level as we see it. Out of many weird questions (up to 20), a few (up to 5) typically should be prepared for the interview. And there is the top-end for every candidate. We recently looked for math majors with CS background and received some pretty solid applications. And this is what we sent out. Would you ace the high end?

  1. Decently hard, but still trivial. The customer journey model.

Map out (a) customer journey and (b) employee and use model to estimate CSF (Critical Success factors => Factor loadings on KPIs) for conversion optimization using Tensorflow.

Or in other words: How do you use Tensorflow to derive strategies on how to optimize our success in sales and/or talent acquisition compared to building the models on a priori assumptions. This also requires a solid understanding of model selection and specification in tensorflow AND general statistics.

Hints: This might require you to:

a) register a google analytics account and understand the data model,

b) Identify a generic form of online journey, understand how you can measure user interaction on that journey. E.g. you could try to use URL shorteners and campaign codes included in the links there to identify who is interacting with what content in that journey, or you could do some hidden markov magic on inferring the source of any signal on our website from where people are coming from. Or you surprise us on your skills on journey analysis.

c) then you might need to research conversion optimization problems how companies describe them qualitatively, e.g. look for metrics that are normally tracked, how they look like, etc. (20% Unique visitor to sign up?)

d) You might need to transform that into a model that you can replicate in tensor flow and use it to extract information that is useful to optimize work to be done and ultimately business KPIs.

2. Straightforward. ROI-weighted Positioning in content using K-Means Clustering

Build an optimization model for choosing the best SEO strategy based on the total set of content of (a) our competitors, AND [inclusive] (b) comparables.

a) Understand what a comparable is and find them

b) Write scrapers to get competitor and comparable whitepapers and content

c) Use a text mining library to get key text information (rapidminder, R, etc.) [the goal is to identify word-stem and word-stem-cluster information and their occurances to derive segment/differentiating information that can be  used to create factors that drive revenue performance of competitors or comparables)

d) Find out how you can measure ROI-on-words (e.g. which company has highest growth and does this relate to the clusters you found)?

e) And how much are campaigns and SEO based on those words and Keywords

f) considering your budget and competitor actions. . . What strategy is best?

3. Also easy. But good practice of theory in applied research. Dimensionality, Identification and Optimization Problem.

Build Entity-Relationship-Model for Freemium model based on what you see in our registration process and discuss key issues in measurement and system identification when estimating causality chains on conversion flows.

Design a high level architecture of a working system.

a) The problem domain is about finding the dimensionality of a space in which you want to estimate causality (e.g. using simult. Equations models or Vector Auto Regressions or some other model) And obtaining that dimensionality from the ERM model. (e.g. entities have attributes, attributes can be from discrete space and are finite, every such discrete attribute could be a dimension, etc. Then you can try to reduce dimensions by looking at how the predictability works. Also looks like you are in discrete choice model space / probits or not? )

b) Once you have dimensionality, you will probably run into estimation problems and realize the ERM model discrete variables aren’t sufficient. So how would you detect it (you can solve using philosophy of statistics and error terms (If you find the book that discussed this, let us know. We forgot about it.) or by finding structure in the error term of the model.

c) And so forth. Your call on how far you run.

4. Potentially tough. Real maths/proof-writing required : Applications of Optimal Stopping theory to our business domain

How do you apply optimal stopping theory to core business problems in our business. And can you build a model for applying it ? (This is a trick question)

HINT: (a) Makes sense to know what optimal stopping theory is . . ., used in American put option pricing, (b) Optimal stopping theory neglects cost of finding an alternative, so need to solve broad concept for applying optimal stopping theory under scarce alternative options )e.g. theory tells you you at k-times variance of a martingale to have an optimal stop, but if you have no other stock to invest in, the internal rate of return of holding and selling of k-l times variance might outperform) (c) Once you have a cool model for replacement costs, maybe talk about how to build an enhanced optimal stopping theory or an algorithm that applies to a real case problem.

Example (do not use this in your case): Abandoning a customer after a optimal stop level of customer acquisition costs might makes sense, but customers aren’t coming from an infinite set. So your cost of abandoning another customer are increasing over time as new customers deplete and put the probability of reaching target revenue and/or expected revenue down. Also interesting: is this a realistic problem? Or how big is the customer set => classic market sizing problem you see in consulting. The elegant answer would be an insight that tells our sales team to treat slipping customers or “revist” deals differently.

PS: Optimal stopping ? Read https://www.springer.com/de/book/9783540740100

5. What would be our competitive advantage if we read and understood Boyarchenkos book? Or is it bull? (If you are too lazy to read and understand the book, find a mining strategy and build a model in tensorflow to derive an answer.)

6. Explain why all of those questions relate to Finance. If you are bad with words. Build a model. If you have trouble understanding the question. Read “Outsiders” by Thorndike and answer then.

Leave a Reply