Insight

What does an insight data scientist do?

The insight data science team use data science to help businesses understand their world better, and to support lower-frequency, higher-impact decisions. An insight data scientist uses analysis and visualizations to create insight containing actionable recommendations, enabling customers to make more informed decisions. 

Insight data scientists typically work on shorter term projects (e.g. one month) that involve the in-depth analysis of specific trends and behaviours to answer questions that customers have. Day-to-day tasks might include:

• Scoping what insight is required by the customer
• Exploring and analysing data in R or Python to find trends and patterns
• Creating charts and visualisations to share insights in the data
• Presenting findings back to customers, focussing on actionable recommendations

What type of projects does an insight data scientist work on?

If a retailer wants to understand their performance in key trading periods, such as Black Friday, to inform their strategy, then they may engage an Insight data scientist on a project basis. This could include understanding which customers are more likely to buy and which products are more likely to sell. This insight would then be leveraged by the marketing team to plan campaigns that will appeal to customers, and by the merchandising team to plan which products they should have in stock.

What skills does an insight data scientist need?

Insight is a particularly commercial area of data science, where industry, domain knowledge and communication skills are key. The most important skills are:

• Being able to communicate well with customers; understanding the questions they have, how their business works and how insight can help
• Visualization and storytelling with data; a high attention to detail when creating charts and presentations, as well as an ability to create a compelling narrative
• Coding; ability to analyze data and create charts using R, Python or SQL

Research & development (R&D)

What does an R&D data scientist do?

R&D data scientists research cutting-edge methods and best practises to upskill the data science team, produce research outputs, and develop software tools that increase team efficiency and improve customer commercial outcomes.

In addition to fostering best practises, R&D data scientists should be subject matter experts in some specialized areas of data science, ML engineering or business domain knowledge. They often work on long-term research projects, and their daily work may include:

• Developing software packages, tools, standard apps and templates
• Researching cutting-edge algorithms and methods in their field of specialism
• Researching and benchmarking available tools for data science
• Cross-team consulting and training in best practices and areas of expertise
• Producing research outputs such as academic papers, training materials and conference talks

What type of projects does an R&D data scientist work on?

An R&D data scientist might work with the product team to turn a set of bespoke solutions that have been built for our customers into a standardized product offering, or developing a cutting-edge application that solves a business problem in a way that provides a competitive advantage to the teams that use it.

What skills does an R&D data scientist need?

Within the R&D team there are different specializations, with ML engineering at one end, research data science at the other, and a range of hybrid roles in between. This means R&D data scientists have a diverse mix of skills across a range of areas, including:

• Expertise in software and data engineering
• Mastery of machine learning and applied data science
• Python, R & SQL languages
• Deep subject-matter expertise in one or more applied fields
• Strong research, academic writing and presentation skills
• Excellent communication skills for internal consulting, teaching and conference talks

Data engineering

What does a data engineer do?

Data engineers work alongside data scientists and business teams, as part of the operations team, to ensure the reliable processing of organizational data and ensure seamless integration of outputs back into end user systems. 

Based on data scientists’ data requirements, data engineers are responsible for working collaboratively with the customer to establish where data is located and the best mechanism for transferring (or accessing) that to Peak, for both historical data transfer and then ongoing updates. Once the best approach is determined, data engineers will either implement data transfer using Peak platform functionality, or advise the customer how best to transfer or access the data.  

What type of projects does a data engineer work on?

Data engineers will liaise with other technical teams within a business to work out where data is stored, building connections or feeds to pull that data into a platform. Depending on the use case, this will be done with built-in functionality on the platform or writing bespoke tools, for example, API consumers. Once ingested, the data engineer needs to get the data into a usable state for the data science team, by exploring the data and applying any necessary transformations.

What skills does a data engineer need?

Data engineers need an understanding of how data pipelines are created technically, as well as how to ensure these are resilient through error handling and retry mechanisms. Technologies and languages they need include:

• SQL
• Python 
• APIs and HTTP response codes 

Operations

What does an operational data scientist do?

Operational data scientists build machine learning and optimization models to enable business users to make better decisions and drive favorable business outcomes – such as increasing revenues or decreasing costs. An operational data science team use data science to support businesses through higher-frequency, lower-impact decisions. 

Operational data scientists typically work on long-term projects (e.g. six months) that involve building a full solution and working closely with end users to determine how a solution will be built. Day-to-day tasks include:

• Speaking with customers to understand their processes, explaining data science outputs and gathering feedback on the project
• Attending internal project meetings to plan out work and collaborate with the customer success team
• Writing code (usually Python, R or SQL) and putting that code into production using docker, or building webapps using languages like Shiny
• Knowledge sharing with the rest of the data science team via subject-specific working groups or training sessions

What type of projects does an operational data scientist work on?

A retailer may want to know at what level they should apply reductions to their products as they come to the end of their lifecycle. This would involve an operational data scientist exploring how price impacts demand for this retailer and applying optimization algorithms to choose the best price to reduce a product while maximizing the profit made for the business. This kind of project would involve data exploration, building both machine learning and optimization models and putting models into production.

What skills does an operational data scientist have?

Operational data scientists are expected to have a range of skills, including:

• Coding: usually R, Python, and SQL
• Mathematical skills; knowledge of statistical methods/optimization techniques
• Presentation skills; as an operational data scientist often needs to speak to non-technical stakeholders, so needs to be able to explain data science clearly
• Problem formulation; often a data science problem starts as something a lot more broad, e.g. “we want to reduce our transportation costs.” A major skill of operational data scientists is being able to turn that kind of problem into a more specific problem that can be solved using data science techniques
• For more detailed information on this, see our blog on how to get a job in data science! 

Team leader

What does a data science team leader do?

Their role is to build and lead a team of data scientists, providing guidance and strategic expertise to ensure projects are successful and that the team are developing the skills required to be great data scientists.

Data science team leaders will manage a team of data scientists, providing them with support on the technical approach and management of their projects and helping their team develop skills to progress their careers in data science. Day-to-day tasks include:

• Attending project or customer meetings to offer technical guidance and ensure the project is on track to be successful
• Contribute to strategic initiatives that help to build and run a data science team, such as defining and implementing new processes
• Having one-to-ones with data scientists to help with their development, and creating a learning and development framework
• Recruiting new data scientists into the team by reviewing applications and taking part in interviews

What type of projects does a data science team leader work on?

These would include designing the recruitment process for data scientists to join the team. A recent example of this for the Peak data science team was around creating a process for our new graduate scheme. Graduates can have similar levels of skills and experience, so the recruitment process needs to be able to highlight which graduates best align with company and team values. This could involve adding questions to their initial application (as graduate CVs are very similar), a hands-on data challenge (to ensure they have the minimal technical skills required for the role), and an assessment day (as a way to evaluate many people at the same time across a variety of skills.)

What skills does a data science team leader need?

Most data science team leaders have been data scientists themselves. They need to possess these technical skills, as well as strategic and management skills to build a great team. The most important skills are:

• Experience in delivering high quality data science projects; being able to identify and mitigate risks, scope projects, create realistic timelines and guide the technical approach
• Mentoring and leadership skills to develop and support a team to perform at its very best; people skills are so important, so that your team are open with you about any challenges they’re facing and what they’d like to achieve in their careers
• A strategic mindset to further improve the data science team going forwards; you need plenty of ideas on what can be done to scale your team, increase efficiencies and build a great culture

1. Start collecting data (if you aren’t already)

Can you tell I’m a data scientist? Data is the best way to objectively measure what your team’s gender diversity actually  looks like. This will help you to quantify where you are now vs. where you want to be, and see if gender diversity is heading in the right direction. Here are some questions to get you started:

• What is the gender split in your current team? How has that changed over time? Are things improving, staying the same or getting worse?
• Are there any key differences in salary/seniority/types of role by gender?
• Does the attrition in your team vary by gender?
• What is the gender split of applicants by role? How does this change throughout the recruitment process?
• What proportion of interviews have a gender-diverse interview panel? How does this impact the outcome for the candidate?

2. Make diversity a team priority

Your team’s metrics of success should include a goal for gender diversity that is equally as important as your other objectives. At Peak, we have the bold target of 50% female employees by 2025. An increase from 26% to 33% of female data scientists by the end of 2021 would mean we’re on track to achieve our 2025 goal.

This has been communicated to the whole team alongside other objectives – creating the momentum we needed as we’ve recently hit our 33% goal! We also mention diversity as early on as we can in the recruitment process and in our job ads, in order to make sure that we’re attracting people who believe it’s an important cause, too.

Are you part of a team with no clear diversity goals? Lead by example. Define these goals, start a Diversity & Inclusivity committee and create a roadmap for the next year for well-defined initiatives you believe will be impactful.

3. Create more opportunities for women

By looking at the skills and experience of women in your team, you can evaluate whether you’re offering opportunities that will attract women. Analyzing which subjects your team have studied and to what level is a great place to start.

38% of women in our data science team have PhDs vs. 57% men, meaning opportunities that are relevant for students finishing BScs and MScs could have a positive impact on gender diversity. We have introduced a graduate scheme this year and four of the eight successful candidates are women and have supervised five MSc students with their dissertations, three of which have been offered full-time roles. Subject area is also important; a higher proportion of women in our team have come from a mathematics and statistics background vs. physics or computer science, so we’ve focused on building connections with and taking part in career events for mathematics departments.

Research by the Alan Turing Institute shows that women are more likely than men to be in analytics roles vs. alternatives such as data engineering, architecture or development. Creating specialist roles within teams can be a great way to scale a team anyway, so increasing diversity is an added bonus! We have an Insight team (which currently has a 50/50 gender split) that we advertise using a different job description than when we’re recruiting for data scientists.

We currently have 33% female data scientists, but this would be 26% if people in the roles mentioned above hadn’t joined the team. This goes to show that reviewing the opportunities you offer can be hugely worthwhile! We still have work to do on improving the diversity of our R&D team, which will be a huge focus for us next year.

4. Create an open culture that listens to feedback

There are two approaches to this that are both important; (1) building trusting and open relationships and (2) regularly collecting widespread feedback that you can analyze by gender.

It’s crucial that managers can be allies for women in your team. Both formal and informal training for managers can be helpful. As an example, our Data Science Management team read and discussed topics from ‘Invisible Women’ by Caroline Criado Perez, relating it to experiences data scientists in the team may have. You can also set up regular catch ups for women in your team, where they can openly discuss the challenges they face with others that can directly relate. This is less about problem solving, and more about making sure women in the team feel valued, empowered and included.

Having an accurate gauge of how your team feels and a consistent mechanism for feedback is helpful for a lot of reasons. Surveys can be a very useful tool for this, especially as you can include a combination of quantitative and qualitative questions. Consider getting feedback on the following areas on a regular basis… and most importantly, make sure you look at the differences by gender:

• The recruitment process: We improved the diversity of our panels after feedback that someone found a completely male panel uncomfortable
• Team or company policies: We created a new maternity leave policy after feedback from women that it may affect their likelihood of staying at Peak
• Overall job satisfaction and stress: Women, on average, have more care responsibilities, so may feel more stressed or burnt out

5. Remember to give back

Having diversity goals as a company and team is important, not only to provide a shared goal, but also to make businesses accountable for change. You may create an amazing team that is great at hiring diverse talent, but if you’re hiring all of the diverse talent without creating more, you’re not contributing to longer term change.

A few ideas on how you can give back are:

• Set up external mentoring schemes: As well as helping people get into the data science industry, it’s a great development opportunity for people in your team
• Be involved with data science communities: There are events targeted at women too, such as Her+Data, which is a great way to support the great work other women are doing in the industry. It’s also a great chance to build your own network and profile!
• Outreach to universities and schools: Inspire the next generation of data scientists! Speak to them about all of the exciting problems you work on and be a role model.

To conclude, this blog post focuses on how collecting data, setting and communicating diversity goals, building a feedback culture, creating more opportunities and giving back can help you to build a diverse team, with a specific focus on gender. However, it’s important to remember that diversity is not limited to gender and this approach can (and should) be applied to other characteristics, too. 

Thank you for joining me on this mission to improve gender diversity within data science teams and I hope this has given you some inspiration!

Peak data scientist Amy Sharif was one of the 2,000 in attendance at the event at Novotel West. Here are her three key takeaways from this year’s conference…

Interpretability of black-box models is becoming increasingly important

With GDPR raising awareness on how data is collected and stored, Alan Rutter argued that consumers will start to question how their data is actually being used and what impact this has on their experience of a product or service.

AI and Machine Learning models can often be described as ‘black-box’; we know the inputs and outputs without any knowledge of how the outputs were calculated.

Going forward, it will be crucial for these black-box methods to be explained, so that we can understand the factors that contribute towards a certain prediction. As well as transparency, this allows us to check whether a model is working as expected and will highlight any bias within the model that needs to be considered to ensure AI is ethical. Wojciech Samek demonstrated the importance of being able to identify bias using an example of how a model which was built to predict the age of a person inferred that ‘older people do not smile’. 

Making AI quicker and smarter

We are quickly getting more and more computational power which allows us to execute complex ML and AI methodologies that we could not run 10 or 20 years ago. Speed and efficiency were touched on in talks about Auto ML, which aims to reduce the amount of manual input needed from a data scientist. Multi-Task Learning is also an exciting development in the field, which enables AI to solve more than one problem at a time. A great example of this is Newsie (created by Cloudera Fast Forward Labs), which categorises news articles across both broadsheets and tabloids.

Developments in the complexity of cutting-edge AI were also apparent, such as Long Short-Term Memory (LSTM) models, which allow models to remember sequential patterns for long periods of time in order to mimic human-like learning. These are currently employed in technology such as Google’s Speech Recognition and Amazon’s Alexa. This technique was demonstrated in practice in Olaf de Leeuw’s session on predicting the price of Bitcoin by using LSTMs to perform sentiment analysis on one million tweets.

Sharing knowledge and giving back to the community is key

The conference clearly demonstrated the importance of sharing knowledge, as this is how AI will continue to progress at a rapid pace. There were many examples, across different industries, of how AI is being used and how we should draw inspiration from work in other fields. Thomas Wiecki’s keynote presentation focused on how there should be more cross talk between statistics and machine learning in order to solve complex problems in a better way, rather than the friendly rivalry that currently exists between the two fields.

Mo Haghihi’s keynote speech, meanwhile, focused on the importance of open-source for both developers and businesses; for developers, open-source gives them the freedom to run and modify code, give back to community and obtain valuable feedback. For businesses, it is cost effective and allows them to keep up with the competition. 

The conference presented a great opportunity to learn, network with other data scientists and discuss some of the current developments in AI – from visualisation, ethics and new modelling techniques to discussing the exciting future of AI.

This future, according to keynote speaker Juergen Schmidhuber, will involve the governing forces of the future becoming intelligence, not the likes of gravity – which is definitely food for thought!