Over the past 10 years, I have worked in analytical roles at several companies, from a small Fintech startup in Germany to high-growth pre-IPO companies (waves) and large tech companies (Uber, Meta).
Each company has a unique data culture, and each role comes with its own challenges and a set of hard-earned lessons. Below, you will find ten of my key lessons from the past decade, many of which I have found to be true regardless of the company stage, product, or business model.
1. You need to tell a story with data.
Think about who your audience is.
If you work in a research-focused organization or primarily present to technical stakeholders (e.g., engineering), a white paper-style academic analysis may be the way to go.
But if your audience is non-technical business groups or executives, you will want to ensure that you are focusing on key insights rather than diving into technical details and connecting your work to the business decisions it is supposed to influence. If you focus too much on the technical details of the analysis, you will lose your audience; Communication in the workplace is not about what you find interesting to share, but what the audience needs to hear.
The most famous approach for this type of insight communication, top-down communication, is the pyramid principle developed by McKinsey consultant Barbara Minto. Check out this recent TDS article on how to leverage it for better communication as a DS.
2. Strong business acumen is the biggest differentiator between good and great data scientists.
If you are a senior DS at a high-bar company, you can expect all your colleagues to have strong technical skills.
You will not stand out by incrementally improving your technical skill set, but by ensuring your work is driving maximum impact for your stakeholders (e.g., product, engineering, biz team).
This is where business acumen comes into play: To maximize your impact, you need to deeply understand the business priorities and the issues your stakeholders are facing and 3) communicate your insights and recommendations in a way that your audience understands them (see number 1 above).
With business acumen, you will also be able to vet your work because you will have the business context and judgment to understand whether your analysis results or recommendations make sense.
Business acumen is not something taught in school or DS Bootcamp; so how do you develop it? Here are a few specific things you can do:
- Pay attention in all company meetings and other group settings when strategic priorities are discussed
- Practice connecting these priorities to your team's work; during the planning cycle or when new projects come up, ask yourself: How does this relate to the high-level business priorities? If you cannot make the connection, discuss this with your manager
- When you are conducting an analysis, always ask yourself so what? A data point or insight only becomes relevant and impactful when you can answer this question and articulate why people should care about it. What should they do differently based on this data?
The ultimate goal here is to move from taking requests and working on JIRA tickets in isolation to becoming a thought partner to your stakeholders, shaping the analytical roadmap in collaboration with them.
3. Be an objective truth seeker
Many cherry-pick data to fit their narrative. This makes sense: Most organizations reward people for achieving their goals, not for being the most objective.
As a data scientist, you have the luxury to push back on this. Data science teams often do not directly own business metrics and therefore face less pressure to achieve short-term goals compared to teams like sales.
Stakeholders will sometimes pressure you to find data that supports a story they have already crafted. While playing along with this may score you some points in the short term, what will help you in the long run is being a truth seeker and promoting the story that the data actually supports.
Even if it feels uncomfortable in the moment (as you may be pushing back on a story that people do not want to hear), it will help you stand out and position you as someone that executives will turn to when they need an unbiased and unvarnished view of what is really going on.
4. Data + Primary Research = ❤
Data that people often frown upon anecdotal evidence, but it is a necessary complement to rigorous quantitative analysis.
Running experiments and analyzing large datasets can provide you with statistically meaningful insights, but you often miss signals that have not reached a large enough scale to show up in your data or are not captured by structured data.
Diving into closed deal notes, talking to customers, reading support tickets, etc., is sometimes the only way to uncover certain issues (or truly understand root causes).
For example: suppose you work in a B2B SaaS business. You may see in the data that the pricing for your enterprise deals is declining, and you may even be able to narrow it down to a certain type of customer.
But to really understand what is happening, you will have to talk to sales reps, dig into their deal notes, talk to prospects, etc., that start to emerge; And the odds are, that pattern does not show up in any standard metrics you are tracking.
5. If the data looks too good to be true, it often is
When people see a strong increase in a metric, they tend to get excited and attribute this movement to something they did, such as a recent feature launch.
Unfortunately, when a metric change seems suspiciously positive, it is often due to data issues or a one-time effect. For example:
- Incomplete data for recent periods and the metric will rise once all data points are in
- There is a one-time windfall that will not sustain (e.g., you see a sales spike at the beginning of January; rather than a sustainable improvement in sales performance, it is just a backlog from the holiday period that is becoming clear)
Do not get carried away by the excitement of a metric increase. You need a healthy dose of skepticism, curiosity, and experience to avoid pitfalls and generate strong insights.
6. Be open to changing your mind
If you work with data, changing your opinion regularly is natural. For example:
- You proposed a course of action for an executive, but lost confidence that it is the right path because you have additional data
- You explained a metric movement in a certain way, but you ran an additional analysis and now think something else is going on
However, most analysts are hesitant to walk back statements they made in the past for fear of looking incompetent or angering stakeholders.
It is understandable; Changing your recommendation often means extra work for stakeholders to adjust to the new reality, and as a result, there is a risk that they will be upset.
However, you should not stick to a previous recommendation simply out of fear of losing face. You will not be able to do a good job defending an opinion when you have lost confidence in it. Leaders like Jeff Bezos recognize the importance of changing your mind when faced with new information or simply when you look at an issue from a different angle. As long as you can clearly articulate why your recommendation has changed, it is a sign of strength and intellectual rigor, not weakness.
Changing your mind a lot is very important. You should never let anyone trap you with anything you said in the past. - Jeff Bezos
7. You need to be pragmatic
When working in analytics, it is easy to develop perfectionism. You have been trained in scientific methods and take pride in knowing the ideal way to approach an analysis or experiment.
Unfortunately, the reality of running a business often imposes severe constraints on our way. We need answers faster than experiments can provide statistically meaningful results, we do not have enough users to split unbiasedly properly, or our data does not go back far enough to establish the time series patterns we want to see.
It is your job to help the business operating teams (those shipping products, closing deals, etc.) get their work done. If you emphasize a perfect approach, the business is likely to just continue without you and your insights.
As with many things, done is better than perfect.
8. Do not burn out your data scientists with special requests
Hiring full-fledged data scientists to build dashboards or conduct special data investigations all day is a surefire way to burn them out and drive them out of the team.
Many companies, especially high-growth startups, hesitate to hire data analysts or BI people dedicated to data investigations and dashboard building. The tasks are limited, and managers want flexibility in what their teams can tackle, so they hire comprehensive data scientists and plan to give them dashboard investigation or data investigation tasks regularly.
However, in practice, this often exceeds the ratio, and DS spends an inordinate amount of time on these tasks. They are drowning in the churn of pulling them away from their focused work, and quickly asking for a way (never as fast as initially) to fill the day, making it difficult to advance larger strategic projects in parallel.
Fortunately, there are solutions to this:
- Implement an AI chatbot that can answer simple data questions
- Train relevant teams on basic SQL (at least 1 analyst 2 for each team) to make them more independent. With SQL AI Assistant or Gemini support in BigQuery, extensive SQL knowledge is no longer entirely required to pull data and generate insights
- Use self-service BI tools that provide users with autonomy and flexibility in getting the insights they need. There have been many advancements in recent years, and tools like Omni are bringing us closer to a world where self-service analytics is a reality
9. Not everything needs a fancy Tableau dashboard
Companies tend to see it as a sign of a mature, strong data culture when data is pulled from spreadsheets into BI solutions.
While dashboards are used by many stakeholders in the organization and serve as a basis for important, hard-to-reverse decisions, they should live in a managed BI tool like Tableau, there are many cases where Google Sheets can get you what you need and get you there much faster, without the scope and building a robust dashboard over several days or weeks.
The truth is, teams will always leverage the analytical capabilities of the software they use daily (e.g., Salesforce) as well as spreadsheets because they need to move quickly. Encouraging this type of agile, decentralized analysis instead of forcing everything through the bottleneck of a BI tool allows you to preserve the resources of data science teams (see number 8 above) and equip teams with what they need to succeed (basic SQL training, data modeling, and best practices for visualization, etc.).
10. Having perfectly standardized metrics across the company is a pipe dream
As discussed in number 9 above, teams across the company will always self-unlock by conducting hacky analyses outside of BI tools, making it difficult to enforce shared data models. Especially in fast-growing startups, it is impossible to enforce perfect governance if you want to ensure teams can still move quickly and get the job done.
While it gives many data scientists nightmares when metric definitions do not align, in practice, it is not the end of the world. More often than not, the differences between numbers are small enough that they do not change the overall narrative or recommendations significantly.
As long as important reports (anything that goes into production, to Wall Street, etc.) are handled rigorously and adhere to standardized definitions, it is okay if the data is a bit messy across the company (even if it feels uncomfortable).
Final thoughts
Some of the points above may feel uncomfortable at first (e.g., pushing back on cherry-picked stories, taking a pragmatic approach instead of pursuing perfection, etc.). But in the long run, you will find that it will help you stand out and establish yourself as a true thought partner.
For more practical analysis advice, consider following me here on Medium, on LinkedIn, or on alternative.