A KPI is a quantifiable metric tied to a specific goal, with a target value and a date by which to reach it. The acronym stands for key performance indicator.
The quickest way to tell whether what you are looking at is really a KPI: imagine the number gets 20% worse tomorrow morning. If you know who deals with it and what changes within the week, it is a KPI. If the answer is that you will flag it in the end-of-month report, you are looking at a metric.
What separates a KPI from a metric
A metric describes, a KPI forces a decision. The site's organic sessions are a metric: they tell you how many people arrived from Google. "Take non-brand organic sessions from 12,000 to 18,000 by 30 June" is a KPI: it has a measure, a starting point, a target, a deadline and a scope.
Outside marketing this difference is written down in a technical standard. ISO 22400-2:2014 defines KPIs for manufacturing operations management and describes each one through its formula, the elements that make it up, its trend over time and its unit of measure, with reference to the work units of IEC 62264. In that context an indicator with no formula and no unit is not a KPI: it is an opinion. In digital marketing nobody signs a document like that, and it shows. The same indicator called "conversions" can mean three different numbers in three different dashboards of the same company.
OKRs sit close to KPIs but are something else: a method for writing goals, not an indicator. A KPI can quite easily become one of the key results of an OKR.
| Metric | KPI | OKR | |
|---|---|---|---|
| What it does | Describes an activity | Measures progress towards a goal | Contains the goal and the results that measure it |
| Nature | Quantitative | Always quantifiable | Objective can be qualitative, results are measurable |
| Has a target | No | Yes, with a date | Yes, within the key results |
| Example | 12,000 organic sessions in May | 18,000 non-brand organic sessions by 30 June | Objective: reduce dependence on brand searches. KR: 18,000 non-brand sessions; KR: 40% of leads from non-brand pages |
The four parts of a KPI
A fully written KPI has four pieces. Removing one is the quickest way to make it useless.
- A measure. A number or a percentage, not an adjective. "Improve our reputation" cannot be measured, "take the Net Promoter Score from 18 to 30" can.
- A target with a date. Without a deadline the KPI cannot fail, and an indicator that cannot fail is of no use to anyone.
- A stated data source. Which tool, which view, which filter. This is the piece most often left out, and it is the one that produces meetings where two people defend two different numbers for the same thing.
- A reading frequency. It has to fit within the time needed to change course. A KPI read once a year on an activity that is adjusted every week is an obituary.
On the third point Google has made changes precisely to end the arguments. In Google Analytics conversions are now called key events: any collected event can be marked as one, and from the same key event you can create a conversion in Google Ads, so the two dashboards report the same count instead of disagreeing.
Types of KPI
The most common classification splits KPIs by area: financial (net margin, cash flow), operational (fulfilment time, error rate), sales (deals closed, average order value), marketing (qualified leads, conversion rate, share of voice) and customer (Net Promoter Score, acquisition cost, churn rate). It helps organise the dashboard, and little else.
A second split matters more: lagging KPIs and leading KPIs. Revenue for the quarter just closed is lagging and tells you where you have been. Qualified leads still in the pipeline and impressions on non-brand queries are leading and tell you where you are heading. A dashboard made only of lagging numbers is a very well-kept rear-view mirror.
When the definition changes, the KPI changes
With digital KPIs the formula is often written by whoever owns the tool, and it changes. Anyone comparing two periods without knowing which definition was in force reads changes that never happened.
Bounce rate is the clearest example. In the Google Analytics documentation it is defined as the inverse of engagement rate: a session counts as engaged if it lasts longer than 10 seconds, or contains a key event, or records at least two page views. A 40-second visit to a single page, which was a bounce in the old Universal Analytics, is not one today. Anyone who put that number in a quarterly report spanning the migration reported an improvement that did not exist.
The same goes for average position in Search Console. Google's documentation explains that the value is the topmost position occupied by a link to the site, averaged across all the queries in which it appeared, and that an AI Overview occupies a single position: all the links inside it inherit that same position. Ads, on the other hand, do not occupy any search position. Average position is not the number you see when you run the search by hand, and treating it as such is the most common way to get an SEO report wrong.
| KPI | How it is calculated | Where you find it | Where people go wrong |
|---|---|---|---|
| Conversion rate | Key events divided by sessions | Google Analytics | The denominator can be sessions or users: two different numbers for the same site |
| Organic CTR | Clicks divided by impressions | Search Console | Impressions include links inside an AI Overview, clicks only count if the user leaves Google |
| Average position | Average of the topmost positions per query | Search Console | It is an average across queries, devices and countries, not the position you see when you search |
| Bounce rate | 100% minus engagement rate | Google Analytics | It cannot be compared with the Universal Analytics figure |
| Average CPC | Cost divided by clicks | Google Ads | On its own it says nothing: read it together with the conversion rate further down the funnel |
| Customer acquisition cost | Total cost divided by new customers | Spreadsheet | Almost nobody includes the cost of in-house staff |
| ROI | Revenue minus cost, divided by cost | Spreadsheet | For SEO it has to be measured over a long window, not month by month |
| Customer lifetime value | Average margin multiplied by the average length of the relationship | CRM | It is an estimate with a margin of error, not a measurement |
How many UK businesses have the data to use them
A KPI assumes that someone collects and analyses data. In the UK most businesses do the first part and far fewer do the second. According to the UK Business Data Survey 2026 from the Department for Science, Innovation and Technology (fieldwork October 2025 to January 2026), 86% of UK businesses handle digitised data, up from 77% in the previous wave.
Only 25% of those businesses say they analyse data to generate new insights or knowledge, and the gap by size is wide: 69% of large businesses, 49% of medium-sized ones, 35% of small businesses, 26% of micro-businesses and 24% of sole traders.
In plain terms: in most UK small businesses today, the problem is not choosing the right KPI. It is that the data is collected but nobody looks at it, and the first useful meeting is the one where you decide who opens the tool and how often.
How many to keep
Few. Three or four per goal, not thirty in a dashboard nobody opens. The test is simple: if in a monthly meeting you cannot discuss each number for at least two minutes, you have too many. An indicator nobody ever comments on is not a KPI, it is decoration.
In search the same rules apply, but with their own metrics: organic traffic, rankings, organic conversions. We have collected them in our piece on SEO KPIs, and they are the starting point of every SEO project we take on.
KPI FAQs
A metric describes an activity; a KPI measures progress towards a goal and comes with a target value and a deadline. Every KPI is a metric, but not every metric is a KPI: the difference is whether someone acts when the number moves.
Three or four per goal. If in a monthly meeting you cannot comment on each indicator for at least a couple of minutes, the list is too long and nobody is using it to make decisions.
By area, there are financial, operational, sales, marketing and customer KPIs. A second, more useful split separates lagging KPIs, which describe a period that has already closed, from leading KPIs, which anticipate the result.
OKRs are a method for writing goals: an objective, which can be qualitative, together with measurable key results. The KPI is the indicator. A KPI can become one of the key results of an OKR.
No. An indicator that falls flags a problem in time to act, which is exactly its job. The wrong KPI is the one that goes up whatever happens, because it measures nothing the business can control.
One last thing that the industry rarely says. In 2026 several traffic KPIs are changing their denominator without warning: AI Overviews generate impressions in Search Console even when the user never leaves the results page, so a falling CTR with a stable average position can describe a site that is doing better than before, not worse. Anyone who keeps the two numbers separate notices. Anyone who looks only at the percentage ends up fixing something that was not broken.
