“In God we trust, all others must bring data.” – W. E. Deming
For a long time, this quote sat scribbled on a piece of paper and stuck above my work desk. Frankly, I’m more of a Zeus fan, but it was still a powerful reminder that, in every project, you need to leave your initial assumptions at the door and try to back up your decisions with numbers.
This is also a hallmark mindset of data-driven companies – but being data-driven is more complicated than that. And since the focus on data is in vogue, especially when it comes to tech companies, let’s take some time to see what being a data-driven company is all about, how you can get there, and what are the stumbling blocks along the way.
I won’t go all academic on you, so let’s look at the shortest viable definition: “Data-drivenness is about building tools, abilities, and, most crucially, a culture that acts on data.”
I like this definition as it’s deceptively simple. In reality, it packs a lot – so let’s look at the different elements:
Even if you have a great team skillset, it’ll be largely useless unless you can satiate your hunger with a good helping of metrics. We often see teams that want to focus on the numbers, but they can’t get all their data in one place. Actually, data wrangling is one of the most common pain points we identified recently while working with tech startups.
And even if you get everything in a single dashboard, stitching that data together and ensuring data consistency can be a challenge – especially in a world where users are using ad blockers and browsers are racing to offer greater cookie blocking.
So setting up the right tracking and attribution tool stack is the first hurdle to being data-driven.
Once you have data streaming in some team members will naturally go and peer at the numbers. But teams comprise different types of people and not everyone has a natural curiosity for data.
This means you need to implement a data learning program to equalize at least the basic data knowledge among key company stakeholders. At a minimum, they need to know what data is available, where to get access to standard reports, and who to talk to in order to get an ad hoc report.
Finally, making sure you enable your data champions to hone their data skills means you’ll have strong internal capabilities.
Even if you have all the data and the skill to use it, that doesn’t mean you’re data-driven. Getting this title can only happen once you embed data in your decision-making process and you consult the numbers at each crossroads in your development.
Clearly, becoming data-driven is a lot of work. So why even bother? Well, we might be biased, but we believe decisions based on data have a few key benefits to offer.
Data equalizes the decision-making process. If you don’t back up your opinion with data the rest of the team is more inclined to take it for what it is – just an opinion. This creates a lot of freedom for different team members to propose solutions without being silenced by hierarchy.
There is a tendency, even in fast-paced startups, to listen to just go with the HiPPO: the Highest Paid Person’s Opinion. But data makes it easier to spot when that happens. Including a reflection phase in your decision-making process – “Did we decide this because the numbers seem to back the hypothesis or because Bill The Boss suggested it?” – builds your organization’s critical analysis muscles.
The more focus you put on data, the higher up the analytics value chain you can get:
Even if you’re not ready for deep-dive analysis, a regular look at your key metrics will alert you of situations where you’re drifting off course and into a sea of inertia.
Take the time to define how a new initiative will move the needle and what metrics you’ll track success against. Then just check them on a weekly basis. This simple 1-2 punch is enough to help you cut your losses when an activity is underperforming. You’ll get out early, saving you budget, time, and effort.
Otherwise, you’ll be tempted to see the initiative through and see that it wasn’t worth it at a later time when you’ve already gone all-in.
A data-driven team is a team where everyone pulls their weight. If that’s not the case, it becomes very clear what’s happening and you’re ready to take action.
Having a transparent process that shows how everyone’s work contributes to the bottom line is also a powerful driver for personal accountability. In a world where employee engagement continues to be a struggle, numbers can actually help. And a process for data-driven decision making clearly demonstrates that you’re not playing favorites and everyone’s work is valued.
When you have company goals set with numbers, this not only enables agency but also can help in breaking down the company silos. If teams are aligned by common end goals, then they will quickly figure out it makes more sense to work together. It enables collaboration and makes it fun. Finally, it helps the team make decisions driven by the common company goals in a less biased way.
And in a situation where one of your key performance indicators is lagging behind, the whole team can swarm in and focus on the problem.
Of course, for all of this to happen, you need to have promote the right attitude towards data and do other changes to your organization. But focusing on data is a good first step to get there.
To answer the question “How can we become data-driven,” we need to look at two parameters: what activities do we need to adopt and how do we get better at them. The former can be split into four categories:
The basic level of developing your data culture means you already have tracking set up and you’re collecting data you can rely on. On the one hand side, this means you’re sure whatever metrics you get are trustworthy. There are no collection errors or blind spots. On the other hand, you can combine and correctly attribute metrics between different stages of your funnel.
This enables you to get the full picture of how your marketing activities, sales campaigns, product features, or customer support work fit together to attract, activate, and retain your customers.
You may very well be sitting on a treasure trove of data – but this will be of little help unless your whole team can get to it.
A strong focus on data access means everyone is empowered to consult the numbers in their day-to-day work. It’s not just about the technical side of providing users with access. It’s also about helping them learn some basic data manipulation techniques (be it SQL querying or something else) that will enable them to reach the data they need. There are also easy open-source business intelligence tools you can use for that, like Metabase. If you want to invest more to really make your data sing, you can rely on sophisticated solutions like Looker or Microsoft’s Power BI.
Of course, having the data to look at and extracting insights out of it are two completely separate things. That’s where robust analytical skills need to play out.
There are many ways you can approach data analysis depending on the size and type of your organization. Some companies rely on a team of dedicated data analysts and data scientists. Others try to build capabilities throughout the organization.
In any case, helping your team learn at least the basic tenets of data analysis is important. It will help you move quicker, as people will have access and the skills to make informed decisions. Even if you have a dedicated data team, this will free up their capacity for deep-dive analysis and getting to more complex insights.
It’s all about getting actionable insights. And to do that, you need to run some experiments and test your hypotheses. This is why we included experimentation under data-driven culture activities.
Here, we need to take into account proper experimental design, as well as knowing when simple A/B or A/B/n tests won’t cut it – especially if you have less traffic and arriving at the necessary sample size takes a lot of time.
The topic of experimentation is so broad that we won’t even try to cover it here. But what we’re getting at is that you really can’t just look at numbers and make decisions in a calm sequential way. Sometimes there’ll be more than one way you think you might improve results – and that’s where experimentation comes in.
It’s important to note here that you’ll be focusing on a couple of these elements at a time, but in the long run, you’ll be working on all of them simultaneously. This adds a certain complexity to the process – but that’s also what makes it fun, right?
We already covered what we’ll be doing, but is there a fundamental difference in how these activities are performed? They will look completely different in newbie vs mature data-driven companies.
Generally, there are four types of data analytics, and the more sophisticated your data-driven organization, the higher up the analytics ladder it can go.
Let’s look at each one in short:
Getting to a higher level of maturity will help you make better decisions and can be a serious competitive advantage. But the road there ain’t easy. Let’s look at the challenges along the way.
Historically, companies had a ton of data they could access. This is quickly coming to an end. Regulations like GDPR and CCPA prove that there’s already a governmental understanding of the perils free-flowing data can pose. And even when governments are slower to react, individuals take matters into their own hands – 47% of internet users now use an ad blocker. You still need to take a proactive approach and install an ad blocker, though. But in the next couple of years, the major internet browsers will completely be blocking third-party cookies.
All of this is to say we’ll need to learn to either operate with less information or take actions to gather it internally. When marketing data is concerned, for example, building high-quality content can prove to be the only way.
A very common case especially for big companies is that the data is already there – but it’s not accessible by everyone in the organizations. A mix of different tech solutions and a general lack of communication means your marketing team has one piece of the puzzle, sales has another, and product is even building a separate picture on their own.
Getting out of that situation is often painful but it can also be fixed by an umbrella business intelligence tool that will mix and align the data each department already collects. The communication problem is more difficult to fix. Setting up cross-functional work groups is a great first step to bridging different viewpoints.
Not everyone has an intuitive understanding of data. Giving your team members access to data without the right analytical preparation can spell disaster for your data initiative. So getting everyone to a common level of understanding is paramount.
Setting up tracking is not just about getting all data in one place – it’s about creating a single source of truth everyone on your team can trust. The reason why many data initiatives die in their infancy is exactly a lack of trust. It’s what Brian Balfour aptly called “The Data Wheel of Death”:
What we’ve often seen as a milder version of this is the case where every team will have a preferred source of information and a different definition of key metrics. For example, marketing uses Google Analytics and counts an activated user as someone who created an app account. Product relies on Mixpanel and defines an activated user as someone who filled-in their profile and completed the full onboarding. You can see why this creates misalignment and tension.
Having a well maintained and properly integrated tracking tool stack is not just a task for the developers. It must be the prerogative of a senior data officer or even the CEO. You can’t move forward without it.
By now you’re probably convinced becoming data-driven is impossibly hard work. On the contrary, my dear reader! It’s a matter of persistence and getting a couple of things right at the start.
Let’s look at the most important steps you need to take right away to promote the data-driven mindset.
Every aspect of a digital transformation can happen only with buy-in at the top. Building data capabilities requires a lot of time and resources. It can’t work as a side project.
The role of the C-suite at this stage is to allocate the resources, but also to demonstrate they are willing to change with everyone else. Setting up measurable goals, conducting regular meetings focused on data analysis and data-backed decisions, solving discussions about the company’s direction with data – all this proves data is part of your organization’s DNA.
Setting goals and tying them to key performance indicators is a way to simplify and streamline the first steps of your data journey. It will help team members to not feel overwhelmed. It will also keep your eyes on the prize.
It’s important to set success metrics for every major initiative and all experiments you’ll be running. In this way, you’re less likely to deceive yourself that something worked when it didn’t.
The simplest case is an A/B test between two landing pages. If you start the test without setting a clear goal, it’s much more likely for you to say “Hey, the new variation didn’t really get us more signups but look at the time on page and the overall engagement – it’s surely a winner!” Now expand the same effect to a complex initiative that uses a lot of budget and human resources. It’s not a cool spot to be in.
We already covered this extensively, so it serves just as a reminder. Data democratization is about both granting access to the metrics and teaching your team how to manipulate said data. When people are able to consult the numbers, they tend to do it more often. And that’s what you want.
We often hear that the numbers don’t lie. That might be true – but your brain does a wonderful job at it all by itself. Even scientists have been proven to arrive at completely different conclusions from the same set of data. So it’s probable your team will get in the same situation sooner or later.
It’s important to be aware of the assumptions you hold when you start analyzing your data and how they might affect you. So getting good at data analysis is about understanding yourself just as much as it is about working with databases.
Providing regular training and increasing the understanding your team has about data analysis will help you quickly move forward. What also works well is to take the time and discuss data-driven decisions out in the open. The more examples people see, the better they’ll understand how numbers drive actions in different parts of your organization.
This piece of advice is strongly linked to the previous one. Yet, it goes a step further. You need to codify discussing decisions and find a time and place for it on your company calendar.
In practical terms, this can be as simple as adding a 10-minute data overview at the start of each company all-hands. If your team is bigger than 50 people, you’ll want to add this element to internal team meetings, as well. This overview can focus on the company-wide key metrics or the numbers related to a specific issue or question. Just make sure you help your team to build a habit of looking at the numbers.
Now, you might be reading these pieces of advice on becoming data-driven and realize you’re already covering most of them. Good job! But that doesn’t mean the work is over.
The next step is to really transform your company strategy into something informed by and set on data. Building a data strategy is more than separate teams being data-driven – it’s about aligning the whole organization to the benefit of long-term growth. This will happen if the following requirements are met:
If you’ve come so far, I guess there’s no need to persuade you in the value a data-driven mindset can bring. But it’s important to note that over-reliance on data is just as bad as not looking at metrics at all.
Today’s fast-paced and complex reality means you can’t always wait for statistically significant results – or even identify a specific causing factor in an environment that’s ever-changing.
So what you need to do to be truly successful is to look at the data, but also use your intuition – being data-informed rather than just data-driven. You need to remind yourself that data can never tell you the full story and you need to analyze it in context. You have to rely on the experts who know your product, your users, and the industry in general.
This is not a “Get out of jail free” card that you can just put on the table when looking at data is hard or not to your liking. But it’s an important notion that we can’t leave without.