What a performance indicator is
A key performance indicator (KPI) is a number you have chosen in advance because it shows whether the site is doing its job. The important words are “chosen in advance.” Plenty of numbers are available in any report. An indicator is a number that you have decided will inform a decision, such as whether to keep investing in a section of the site or change how a page works.
A good set is small, usually three to six figures. Beyond that, it becomes hard to tell which ones actually matter.
Start with the goal
Indicators come from the purpose of the site, not from whatever a tool displays. Write the goal in one plain sentence first. Examples:
- “Bring in inquiries from people in my service area.”
- “Get readers to subscribe to the newsletter.”
- “Sell products from the online shop.”
- “Help customers find answers so they contact support less.”
- “Build an audience for our articles.”
Then ask what a visitor has to do for that goal to be met. That action is the outcome indicator. Everything else is supporting evidence.
Two kinds of indicator
Outcome indicators measure the thing you actually want: inquiries sent, sign-ups completed, orders placed. They are the most meaningful, but they can be slow to change and may be low in number for a small site, which makes them noisy.
Supporting indicators measure steps along the way: visits from search, clicks on the contact button, pages viewed before enquiring. They move faster and are plentiful, but they only matter if they connect to the outcome. Many more visits do nothing for you if none of them lead to the action you want.
Pair them. For each outcome, pick one or two supporting numbers that plausibly feed it.
Example sets by site type (illustrative)
| Site goal | Outcome indicator | Supporting indicators |
|---|---|---|
| Local service inquiries | Contact form submissions and calls | Visits from search to service pages, clicks on phone link |
| Newsletter growth | Sign-ups | Visits to articles, sign-up form views |
| Online shop | Orders | Visits to product pages, add-to-basket actions |
| Information site | Return visits | Search clicks, pages per visit |
These pairings are starting points, not rules. Your own site may need different ones.
Make every indicator precise
An indicator is only useful if everyone defines it the same way. For each one, write down:
- What counts. For example, “form submissions that are not obvious spam.”
- Where the number comes from. One named source, so that you do not mix data from different tools that may disagree.
- The time period. Usually a calendar month, compared with the same month a year earlier if the topic is seasonal.
- Who looks at it and when. A number nobody reviews is decoration.
Tracking the outcomes in the first place usually requires a little setup, covered under conversion tracking, and being careful about what you gather is part of analytics setup.
Set a baseline first
A baseline is the starting value of each indicator before you make changes. Without it, you cannot say whether anything improved.
To build one:
- Record at least a few weeks of data, and a few months if the site has seasonal swings.
- Note the typical range, not only the average. If monthly inquiries have been between 8 and 15, a month with 12 is not news.
- Write down anything unusual that affected the period, such as a site outage, a promotion or a redesign.
Here is a hypothetical example. Suppose a small site recorded 9, 14, 11 and 12 inquiries over four months. The typical range is roughly 9 to 14, and the average is 46 divided by 4, which is 11.5. If the next month shows 13, that sits inside the range and is not evidence of an improvement. A run of several months above the old range would be more convincing, and even then other causes are possible.
Avoid vanity indicators
Some numbers look impressive but do not guide decisions. Total page views, social followers and raw impressions can all rise while the goal goes unmet. They can be reasonable supporting signals, but they should not be the headline. A quick test: if this number doubled and nothing else changed, would you be better off? If the answer is not clearly yes, it is not a key indicator.
Review and revise
Revisit the set a couple of times a year. Goals shift, and an indicator that once mattered may no longer fit. When you change one, note the date so earlier figures are not compared with a different definition.
It also helps to remember that indicators describe what happened, not why. A drop calls for investigation, not an instant fix, and how to read these movements is covered under metric interpretation. Our broader discussion of measuring success looks at what “good” means for different goals.