I'm new here. I'm about to have a final interview for a data scientist position for a company (it's in the e-commerce field) that is new for data science. It's a pretty new position for the company, and from the interviews I had so far, I noticed that they don't fully understand what they want from a data scientist. They barely know what data science is. I explained them the "standard" data science workflow (Ask A Question, Get the Data, Explore the Data, Model the data and Communicate the data). But I don't think they got it. I want to set the proper expectations, so the company and I can agree on my job description, and what they can expect from me in one month, tree months, six and twelve months. So, how do you handle the expectations(specially the CEO) from a company that's new to data science?
I tell you what has worked for me: practical examples. They have probably already read about what data science is in general and which are the standard procedures. What they have not seen is someone in front of them explaining how innovative (and useful for business!!!) the data science really is. Follow the "Say what you would say to your granny to tell her how cool your work is" advice.
People want to know what you have actually done, indipendently from the algorithms or procedures you have used.
I have not worked as a data scientist. But managing expectations is something we've all done. Considering you're referring to the company CEO, tact is especially required. There are three practices that have helped me a lot over the years:
- Never say "it can't be done" right away. Many, many, many times I have said that (as a programmer), only to think of a solution later on. When I'm skeptical that something can't be done, I've learned to say, "Can I think about it and get back to you?" That approach has kept egg off my face many-a-time. It also gives me time to do some research.
- Even when something can't be done, focus on what you can provide. If the requested solution is unrealistic for some reason, try to think of what is realistic and offer that.
- When discussing what can and cannot be realistically done, clients often don't care about the technical details (even though that's what excites you, and what you want to talk about). Be ready to go into the technical details if asked, but start off with a very high level explanation. Rather than focus on the negative details of why something cannot be done, describe the positive details of what it would require to accomplish the task. That shows you have thought it through, and that you're trying to be team player rather than a road block.
Wow, you must be taking the job I just left behind!
Our CEO (and full exec team) was used to Wired article headlines and viewed any test with <30% conversion boost as "failure". My team was expected to do needlessly complex analyses because one of their friends had forwarded a click-bait article about "growth hacking".
First, find a baseline.
Ask during the interview process what they hope you'll accomplish in the first 3-6 months. If they have a solid list of deliverables, that's a good sign. If that list is something you're comfortable with even better.
Ask "why?" a lot. In the interview/early days you're understandably naive, so you can say
oh, why is that important? and it's interpreted as
I'm trying to learn about our business. If their assumptions show some thought and understanding, even if you don't fully agree, that's a great sign.
Ask how those projects will help with their overall goal. What is their overall goal?
We want to be a $1B company is a terrible goal.
But if your manager-to-be has no good answers to those questions: walk. There are plenty of places hiring.
Second, build a plan.
You have their list of ideas and some reasons they're important. Put together a rough draft plan of how you'd attack them. Don't get too in the weeds, just list the strategic goal, the data you'll need, the method you'll start with, what you hope to find and a rough time estimate.
Ok, you're disappointed with mobile conversion. I'll want to implement on-site tracking (eg Segment.io) to start logging a variety of parameters (eg browser, IP locale) as well as getting views on visitor flow. I'll use a basic SQL client (eg Mode) to segment users and see if the problem is confined to any one group (eg Android users). From there, I'll work with the product team to size/prioritize possible split tests, which we'll analyze manually at first. It'll take a few days to get Segment up and verified, ~1-2 weeks for data collection, and a full week to do a deep dive. The earliest we could start experimenting would be two weeks, but realistically after the better part of a month.
You're just riffing to see their reaction. Are they nodding reverently like you're speaking Latin, it sounds impressive but they don't follow? Are they pushing for faster turnaround and guaranteed results? Are they asking insightful questions (eg Can you really get an integration up in 2 days?)
If they're engaged and interested, then you've found a good partner. Otherwise consider how much you're willing to invest on education/management.
Finally, assess how aligned others are
Talk to the engineering team and ask what they've worked on in the past year. Push for how they source and prioritize projects. You'll likely be in a very similar boat.
Consistency or thrashing? Evidence-based or HiPPO decision making?
TL;DR: Go with genuine interest
In general, I've been most successful when I screen employers for curiosity over performance.
Everyone wants to do well in business, but the people I've loved doing data work for are ones who are eager to learn what's going on with their business and how to please customers better. The ones I've dreaded working for want to brag about KPIs and stock prices. The former will be patient as you get your sea-legs in a new org.
Good question! I've faced this problem before, where the CEO had unrealistic expectations. Unfortunately I can't tell you how we worked it out, because it didn't work out. The good news is that you're not working for them yet, so you can just walk away. At the end of the day, it comes down to trust, and they have to convince you they trust you. Talk to the other employees to find out they are treated; that might tell you something.