What does being a data-driven business look like in practice?
Being a data-driven business starts with understanding what information matters to the business and deliberately capturing it.
For example, the customer success team should know how many calls they receive, the nature of those calls, how long they take, what kinds of complaints are coming in, how frequently they occur, and whether there are particular periods of the month when certain complaints become more common. The marketing team should similarly be tracking not just what comes through the website, but how people interact with content and emails, including engagement and open rates.
These are only examples. Every department should have a clear understanding of the data points that are relevant to its work and to the wider growth of the business.
But data collection alone does not make a company data-driven. There has to be a clear understanding of what is being measured, whether it is actually being measured consistently, and, most importantly, how those measurements are being used to make decisions.
You can collect enormous amounts of data and measure everything accurately, but if your executives ultimately make decisions based on instinct, you are not really operating as a data-driven business.
The process is therefore quite simple at its foundation: identify the data that is relevant to what you want to measure, ensure that every department is capturing it, and then make business decisions based on what the numbers are showing you. The scope of what you measure can evolve as the company grows, but there should always be a clear connection between the data being collected and the outcomes the business is trying to achieve.
What data should operations teams use to understand customers?
Operations teams are not always directly customer-facing, so the metrics they track may not measure the customer in the same way that marketing or customer success does.
Revenue operation is slightly different because it sits much closer to the growth ecosystem. The focus is on understanding the relationship between functions and, ultimately, the entire lifecycle of the customer, from the moment they discover the business to the point where they become a customer and continue interacting with the company.
One of the things we measure is how long it takes a customer to move through that lifecycle. For example, how effectively are marketing-qualified leads becoming sales-qualified leads?
The gap between marketing and sales
An illustrative example from Ola. A gap this size needs to be understood, not just reported.
The next question is what happens to those leads after qualification. Do they convert into revenue at a rate that makes sense? Are the leads being generated actually aligned with the company’s ideal customer profile? Are they converting according to the benchmarks the business has established for itself or according to relevant industry benchmarks?
There is another set of questions between sales and customer success. Once a deal has been closed, how long does it take for the customer to be onboarded? Where does the handover between sales and customer success slow down? How long does it take before customer success can say that the customer is fully onboarded and activated?
These measurements reveal the efficiency gaps that may otherwise remain hidden between departments. That is one of the important roles of revenue operations: understanding not just how individual teams perform, but how effectively the entire revenue system works together.
How can data help growth teams understand changing customer behaviour?
There are different ways to approach this, but customer behaviour analysis is one of the most useful. It can involve customer segmentation as well as predictive analysis. For an early-stage startup that may not have sophisticated infrastructure yet, I would start with something much more practical.
When I was at a former company, I worked closely with the customer success team. Because it was a B2B business, customers typically paid for a year before reaching the point where they might decide whether to renew or churn. That created a challenge: a customer could have a poor experience for months without explicitly telling you that they were unhappy. They might simply discover another software that was cheaper or better suited to their needs and decide to switch.
So we started looking at whether we could predict customer behaviour from the way customers interacted with the product. One concept we used was what we called “stickiness.”
Payroll was particularly sticky because once a company had fully embedded its payroll processes into a system, it became difficult to switch. The operational cost of changing was significant, particularly for larger organisations.
If a company was deeply embedded in payroll and then began using HR assistance, disciplinary management and other performance-management features, that increased its stickiness even further.
We ranked different product features and assigned scores to them. Those scores helped us identify how deeply embedded a customer was in the platform.
We also looked at product usage. Even if a company had already paid for a full year, if they were only logging in once or twice a month, that was a signal they might not be getting enough value.
That gave the customer success team an opportunity to intervene before the problem became churn. Instead of waiting until renewal to discover that a customer was disengaged, the team could proactively reach out and say, “We’ve noticed that you haven’t used the platform recently. You haven’t run payroll. You haven’t used these features. Is there something we can help you with? What would make the experience better?”
We could then explore other interventions, including helping customers understand features they were not using or offering incentives for longer-term usage.
You can use behavioural data to identify changes in the customer’s relationship with your product before the customer explicitly tells you that something is wrong.
There are other approaches as well. You can continuously collect feedback through surveys and Google Forms, monitor NPS and CSAT scores, understand which features customers value most, and analyse behavioural patterns and drop-off points using tools such as Mixpanel. All of these inputs can help you understand whether customer behaviour is becoming healthier or deteriorating, while also informing decisions about which features the product should develop next.
What should teams ask before acting on a pattern in their data?
There is no universal checklist because the questions you should ask depend heavily on the situation.
If, for example, you have a hypothesis that savings activity will increase during salary periods and you suddenly observe a significant increase in spending instead, you should not immediately conclude that the original hypothesis was wrong or that the business should change direction.
You can investigate further. You might run a poll to understand what customers are using the product for, observe whether the behaviour continues over the next one, two or three months, and then determine whether the pattern is persistent or simply an anomaly.
If the behaviour continues, you can start asking more strategic questions. Should the business focus more heavily on savings? Is there a way to make savings more attractive? Is the current product encouraging customers to spend when the business actually has a stronger opportunity in savings?
The important thing is to recognise that data is only one facet of the situation. If the data shows that something happened, the next question should often be why it happened. You can use the behavioural data to identify the change and then speak directly to customers to understand the underlying reason. That could mean a focus group discussion with the customer success team or one-on-one interviews with your most engaged customers.
The data tells you what happened. Customer conversations can help you understand why.
What are the most common mistakes businesses make when interpreting customer data?
One of the biggest mistakes is allowing your hypothesis to become a bias and then looking for data that confirms it. It is perfectly reasonable to begin with a hypothesis. The problem is when you stop allowing the data to challenge it.
You need to give the data enough independence to either support or disprove your original assumption. Otherwise, you are not really using data to inform your thinking; you are using data to validate what you already wanted to believe.
There is another important consideration, which is context. Data can be completely accurate and still lead you to the wrong conclusion if you apply it in the wrong context.
What US data says adults between 20 and 25 earn a year. An accurate statistic.
Does that context apply to a business operating there? Not necessarily.
You have to understand both the data and the environment in which that data was generated.
Does this number mean the same thing in my market, for my customers and for my business?
Good data interpretation is therefore not simply about asking, “What does the number say?” It is also about asking that.
What customer behaviour indicates increasing or declining trust?
There is a concept I teach called “Trust as a Metric.” Trust can operate at different levels. There is trust in how a company handles your personal data, and there is trust in the product itself: the belief that the product will actually deliver what it promises.
The behavioural signal for product trust will differ depending on the product.
So there is no single universal metric for product trust. The business has to determine what behaviour demonstrates that customers believe the product will deliver.
Data trust is another dimension. One possible signal is whether customers are comfortable giving a company permission to use their data or track their interactions. Apple is an interesting example because it has built a strong proposition around transparency and control over personal data. Users are given information about what data is being collected and how different services work, while Apple also places restrictions on how other applications can access user data.
For businesses operating in a relatively low-trust environment like Nigeria, repeated financial behaviour can also be a powerful trust signal. If customers are willing to consistently put money into a product from January through December because they believe they will receive their money and interest as promised, that behaviour tells you something meaningful about the level of trust they have developed.
On the other hand, you can directly test declining trust through customer feedback. You can ask whether customers would recommend the product to their friends and family, and, more importantly, ask why they would or would not recommend it.
Define trust in relation to the specific product and the behaviour that demonstrates confidence in it.
How can poor data collection lead to wrong conclusions?
On paper, it is easy for an organisation to define what its data should look like and how it should be collected. In reality, businesses are made up of people working across different functions, often under pressure and making decisions on the fly.
A lot of businesses in Nigeria operate this way. They are implementing systems while the business is already moving, rather than establishing the underlying data structure first and then building more sophisticated layers on top of it.
One way to solve this is through clear data operations procedures that define how data should be collected across departments. You also need what I would call a single source of truth: a central data warehouse where information from different parts of the organisation can be connected and trusted.
That creates a single source of truth that the organisation can rely on. It also requires a defined refresh schedule. Depending on the company’s infrastructure and resources, data may need to be refreshed hourly, daily or weekly.
For very early-stage startups, the reality may simply be 10, 15 or even 100 Google Sheets. That is not necessarily a problem at the beginning. But as the organisation grows, data operations need to evolve with it. Integrations need to be established. Data warehousing needs to be introduced. Information should be extracted from source systems as automatically as possible rather than relying on people to manually move data from one place to another.
Where human input is unavoidable, there should be clear data-quality and data-assurance processes. This is where data governance becomes important. The organisation needs consistent rules for how information is entered, stored and interpreted.
Even something as simple as date formatting matters.
If one system records a date in one format and another system records the same information differently, inconsistencies can eventually create problems when those datasets are combined. The objective is to reduce unnecessary human error while creating a governance structure that ensures the information entering the system is accurate and consistent.
How do businesses turn data into decisions executives can act on?
This is where business objectives become important. Every department should have measurable outcomes, and the data should help the organisation understand whether those outcomes are being achieved.
For sales, that could be monthly revenue or month-on-month growth. The team might also want to understand where its sales are coming from geographically. If Lagos is generating the majority of sales, that becomes a business question: should the company double down on Lagos, or should it use what is working in Lagos to develop other markets?
For customer success, the questions might be different. Which product features generate the most complaints? Is there a particular feature that is generating enough frustration that engineering needs to investigate it? Are the biggest complaints related to logistics, product quality or something else?
These departmental measurements should ultimately connect to the organisation’s broader goals.
A sales goal for the year. Use the data to see where revenue comes from, where the strongest opportunities are and where investment would have the greatest impact.
An active-user goal. Look at users who registered but never activated. Converting them may be more efficient than acquiring entirely new users.
You can also use customer satisfaction data to identify opportunities. If customers are consistently indicating that they would recommend the product, that could provide a basis for introducing a referral programme. The important point is that the data should always be interpreted against a business objective.
It can also help the company change direction during the year. Imagine that the annual revenue target is $1 million and, by month six, the data shows that the auto-save feature is generating significantly more revenue than other features. At that point, the business has a decision to make. Why continue allocating the same level of acquisition investment to lower-performing areas if the savings product is demonstrating stronger commercial potential?
The company could increase investment in acquiring customers for that particular feature, measure the resulting change and determine whether it moves the business closer to its annual goal.
That is what a data-driven decision looks like: the data is not simply being reported. It is changing where the business puts its time, money and attention.
Why is Nigerian customer data particularly difficult to interpret?
One of the challenges is that Nigeria is not a particularly data-centric society, and many parts of everyday life remain highly informal. That extends from how people buy food and other basic goods to how different parts of the healthcare and service economy operate.
The technology sector, however, has an advantage because even in an informal environment, digital interactions create footprints that can be measured. Businesses can use tools such as Google Analytics, but they should not rely on quantitative data alone. Polls, focus groups and direct customer feedback can help fill in the gaps.
Customer feedback should become part of the operating culture of the business.
If customers understand that every time they use your service you are going to ask about their experience, and they see that the business actually uses that feedback to improve, the process becomes part of the relationship between the company and its customers.
You are not just collecting feedback because it looks good in a report. You are creating a culture in which customers expect to be heard and the business expects to respond. That consistency makes the data more useful because customers understand the feedback loop and are more likely to participate in it.
What do you wish founders and growth teams understood about data?
A lot of businesses are becoming more data-driven, which is encouraging. Even smaller businesses are increasingly using tools such as lightweight CRMs and digital platforms to track their inflows, outflows and customer activity.
But I think the next step is more important: we need to move from simply collecting and tracking data to understanding how that data can actually help businesses grow. That is where I think revenue operations remains significantly underutilised in Africa.
Revenue operations brings together marketing, sales, customer-facing functions and, in some cases, finance. Its role is to create a coherent flow of information across those functions so that the customer does not feel like they are interacting with completely separate businesses every time they move from one part of the organisation to another.
There should be continuity in how information flows into the business, how it moves between teams, how teams act on it, and ultimately how the customer experiences the product. That is the real value of revenue operations. It turns what could otherwise be a linear customer journey into a connected ecosystem, where the customer comes into the business and has reasons to remain within it.
One of the clearest examples of poor revenue operations is something many Nigerians have probably experienced with their banks.
I send a complaint saying, “There has been a wrongful deduction from my account.”
At almost the same time, I receive a promotional email from that same bank: “Invest ₦50,000 and take advantage of this promotion.”
Meanwhile, money has just been deducted and I have not even received a response to my complaint.
The problem is not necessarily that either team is doing its job badly in isolation. The problem is that the teams are operating in silos. The marketing system does not appear to know that the customer is currently dealing with a serious issue, and the customer service system does not appear to be informing the marketing system about that context.
Revenue operations is partly about solving that disconnect. When those systems are integrated properly, the business develops a more holistic understanding of the customer. It becomes easier to coordinate how information moves across the organisation, respond appropriately to customer behaviour and ultimately create a much more coherent customer experience.
Data is not useful simply because you have it. It becomes useful when it changes the decision the business makes and improves the experience it creates for the customer.