Every company has the opportunity to collect data relevant to important decisions. Google Analytics delivers website visits, the CRM counts leads, accounting knows the revenue, and Excel spreadsheets contain budgets, targets, and headcount figures.
But which of these numbers are actually relevant? Which should support concrete decisions? Which should be monitored to gain insight into how the company is developing?
Data, KPIs, and Business Intelligence do not start with dashboards. They start with the question of which data is needed in what form so that better decisions can emerge.
A marketing team monitors website page views weekly. The number has been climbing for months. But enquiries are not increasing.
Only when the team introduces conversion rate by channel as a metric does it become visible which sources also generate enquiries. Page views were being monitored, but conversion rate was what mattered.
This article explains the role data plays in Business Intelligence, how reliable metrics emerge from raw data, and what really matters when selecting them. It forms the conceptual foundation for all further topics around data, sources, and metrics.
The role data plays in BI
Business Intelligence consists of systems and processes that evaluate data to support decisions. Different types of data come together, each with its own strengths and limitations.
Structured data
Most data in BI is structured: numbers, dates, categories. Everything that can be organized into rows and columns. Revenue per month, clicks per campaign, leads per channel. Structured data can be used directly in dashboards and reports.
Unstructured data
Customer feedback, support tickets, emails, or notes from sales conversations contain valuable information that cannot easily be arranged into rows and columns.
They supplement structured data with context and qualitative insights. Modern BI solutions can store and make such datasets accessible as well – even if they cannot be visualized directly.
Dimensions and metrics
Structured data follows a core principle that is central to BI: the distinction between dimensions and metrics.
Dimensions describe context: time period, channel, region, product category, employee. They answer the question "Where, when, who, what?"
Metrics are the measurable values: revenue, clicks, costs, orders, page views. They answer the question "How much?"
A single data point – say "€4,250 revenue" – only becomes meaningful through dimensions: €4,250 revenue, online shop, March 2026. This interplay of metrics and dimensions forms the basis for every BI analysis.
To learn more about how this data is integrated into the BI solution, see Connecting Data Sources.
From raw data to metric
Not every number a system delivers is automatically a useful metric. The path from raw data to a reliable decision basis runs through three stages.
Raw data
Raw data is the unprocessed information a source system generates, for example:
- individual page views
- transaction records
- log entries
- CRM contacts
It is detailed, but rarely decision-relevant when viewed individually. To meaningfully support business decisions, a strategy for working with the data is required.
Metrics
Through aggregation and calculation, raw data becomes metrics: total revenue in the month, number of page views per day, average session duration. Metrics summarize raw data and make it analyzable. They describe what happened. Whether that is good or bad depends on context.
"We had 12,400 website visitors in March" is a metric.
"Our visitor count is 18% above last month, but the conversion rate has fallen from 3.2% to 2.1%" is a KPI with context.
Only the second statement supports a decision – in this case, to investigate why more visitors were acquired but fewer converted.
KPIs
A metric becomes a Key Performance Indicator when it is tied to a concrete goal. Conversion rate is a metric. Conversion rate compared to a target of 3.5% is a KPI.
Only through the reference to a goal, a benchmark, or a threshold does a number become evaluable. That is what makes it decision-relevant.
A piece of information in the form of a metric is not inherently helpful. The relevant metrics are those that make action necessary:
"We need to investigate more closely why this happened."
And in the next step:
"How do we use these insights to improve the relevant metrics?"
This is how observing, analyzing, and acting on relevant metrics leads to better business outcomes.
Outcome and driver metrics
There are outcome metrics and driver metrics. Both have different explanatory power, different functions, and complement each other.
Outcome metrics
Outcome metrics measure what comes out at the end: revenue, profit, customer satisfaction, close rate. They show whether a goal was reached. But not why.
From the outcome metric "revenue" alone, you cannot derive whether it is down to traffic, conversion rate, average order value, or return rate.
Driver metrics
Driver metrics measure the factors that influence an outcome: number of qualified leads, sales response time, click rate of a campaign, average order value.
They are more operational and can be changed more quickly than outcomes. Whoever measures outcomes sees the effect. Whoever also measures drivers and relates them to the outcome understands the cause.
| Outcome metric | Driver metric | |
|---|---|---|
| Measures | What was achieved | Why it was achieved |
| Response time | Slow (outcome appears with a delay) | Fast (directly actionable) |
| Marketing example | Return on Ad Spend (ROAS) | Cost per click (CPC), conversion rate |
| Sales example | Close rate | Number of first contacts, response time |
| E-commerce example | Revenue | Average order value, return rate |
| Finance example | Contribution margin | Cost per order, margin per product |
Good dashboards, reports, and analyses contain both:
- Outcome metrics that show where the company stands
- Driver metrics that show which levers can be adjusted
Those who only look at outcomes react blindly. Those who also measure drivers navigate with data.
How to build effective dashboards from this is covered in a dedicated article (see Creating Dashboards).
Characteristics of useful metrics
These characteristics are important for defining KPIs that are reliable for your goals, rather than settling for surface-level metrics.
Measurable relevance to a business goal
Outcome metrics can directly or indirectly affect this goal. While sales in the online shop feed in directly, an advisory conversation started via the website might only result in an outcome later.
The same applies to driver metrics. An increased conversion rate on sales leads to higher revenue at the same visitor volume. More visitors do not automatically mean more revenue – but with sufficient quality, they offer potential to convert them into customers.
The underlying measures in turn can directly boost these metrics and thereby indirectly improve the outcome. For example, if a certain type of social media post brings more visitors.
Influenceability and long-term impact
A discount campaign may bring more visitors and revenue in the short term, but it is only conditionally suitable for sustainably increasing website revenue.
Increasing the conversion rate through optimization of product pages, on the other hand, would be a more sustainable method for more website revenue.
Examining the relationships between metrics
A good metric measures something the team can change through its own measures. Overall market development is an interesting number, but not a KPI for the marketing team. Cost per qualified lead, on the other hand, is directly influenceable.
This approach is recommended so that important metrics are genuinely embedded in processes and measures by employees:
Define goals and relationships:
- Business goal
- Outcome metrics
- Driver metrics
- Processes
- Measures
- Employees
First you need a clear, measurable, and influenceable goal, e.g. increase website revenue. The relevant outcome and driver metrics are then identified for this goal. The underlying processes, measures, and employees have a direct influence on these metrics.
Monitor and analyze:
To be able to determine these chains and thus direct and indirect influence, the structures and processes also need to be in place.
In this example, an analytics tool first needs to be set up to capture visitors on the website.
Whether certain blog articles, discount campaigns, or social media activities influence the metrics can in turn be achieved by capturing detailed user activity on sub-pages of the website.
This data then needs to be analyzed by employees, who draw conclusions for the success and planning of their measures.
Practical example: increasing website revenue
The following example shows how an e-commerce team puts the described approach into practice. The business goal (20% more net revenue through the online shop) is broken down into measurable outcome and driver metrics. Each metric has a current value, a target, and a responsible person. The structure can be transferred to other goals and departments.
The first table shows the metric logic (what we measure and why), the second the implementation plan (what we do concretely).
Table 1: Goal, metrics, and relationships
| Level | Metric | Current | Target (12 months) | Influenced by | Responsible |
|---|---|---|---|---|---|
| Business goal | Net revenue (after returns) | €500,000/year | €600,000/year (+20%) | All drivers | Marcus (CEO) |
| Outcome | Orders per month | 420 | 520 | Traffic, conversion rate | Tom (Performance) |
| Outcome | Revenue from newsletter channel | €2,900/month | €6,000/month | Sign-ups, email series | Sarah (Email) |
| Outcome | Repeat buyer revenue share | 18% | 28% | Repeat rate, post-purchase | Sarah (Email) |
| Driver | Website visitors per month | 32,000 | 45,000 | Content, social media, SEO | Lea (Content) |
| Driver | Conversion rate | 1.3% | 1.5% | Product pages, checkout, A/B tests | Tom (Performance) |
| Driver | Avg. order value | €78 | €88 | Cross-selling, product bundling | Tom (Performance) |
| Driver | Newsletter sign-ups/month | 350 | 800 | Blog articles, social media, pop-ups | Lea (Content) |
| Driver | Return rate | 15% | under 12% | Product information, size guidance | Marcus (CEO) |
| Driver | Repeat buyer rate (within 6 months) | 22% | 30% | Post-purchase series, incentives | Sarah (Email) |
Table 1 answers the question "What do we measure and why?" For the metrics to move, the next table presents measures with a clear link to the drivers – including a responsible person and a start date.
Table 2: Measures and implementation
| Driver metric | Measure | Current | Planned | Responsible | Start |
|---|---|---|---|---|---|
| Website visitors | 2 blog articles/week (comparisons, tips) | 2 articles/month | 8 articles/month | Lea (Content) | Month 1 |
| Website visitors | Social media: 3× reels, 2× carousels/week | irregular, 2×/week | 5×/week with editorial plan | Lea (Content) | Month 1 |
| Conversion rate | A/B tests on product pages (CTA, images, reviews) | no tests | 2 tests/month | Tom (Performance) | Month 3 |
| Order value | Cross-selling in cart and checkout | not present | automated recommendations | Tom (Performance) | Month 2 |
| Newsletter sign-ups | Optimize sign-up forms (pop-up, blog sidebar) | simple footer form | pop-up + sidebar + incentive | Lea, Sarah | Month 1 |
| Newsletter revenue | Welcome series (5 emails over 14 days) | no automation | automated series | Sarah (Email) | Month 2 |
| Repeat buyer rate | Post-purchase series (care, accessories, rebuy incentive) | no process | 3-part email series | Sarah (Email) | Month 3 |
| Return rate | Return analysis by product category + cause | no evaluation | quarterly analysis | Marcus (CEO) | Month 3 |
Such goals, metrics, and measures can vary greatly from case to case.
A structured approach – for example through tables like these – helps greatly with initial planning, tracking, and optimization.
Avoid these typical KPI selection mistakes
Too many metrics
It is less about the number of metrics than about their significance. Listing potentially relevant metrics and assigning each a priority value from 0 to 1 creates clarity about relevance.
Vanity metrics
Some numbers look good but offer no added value on their own. Typical examples are follower counts, page views, or number of emails sent. They are not universally bad, but without appropriate context they are not universally good either. They have no concrete influence on business goals.
Missing or unclear definitions
Revenue, for example, can be gross, net, before returns, after returns, etc. So that everyone in the team is clear on the meaning of metrics, they should be clearly defined.
Conclusion
The value of Business Intelligence lies not in the volume of available data, but in the numbers that actually improve decisions and outcomes.
Raw data becomes metrics through aggregation, and metrics become KPIs through a goal reference. Only these KPIs in their context are the foundation on which dashboards, reports, and data-driven decisions are built.
All Data. One system.
Contact
Paul Zehm
Founder at Sandbank