Skip to content
  1. Home
  2. How KO works
  3. What to compare a number with
How KO works

What to compare a number with

In a KO report, a number is compared with four things, and each comparison answers its own question. All four are set up so that they do not mislead you at the start of the month or on small data.

In short

  • Previous period: the same stretch of days 1–12 June is compared with 1–12 May, not with the end of May.
  • An unfinished day: the same everywhere Today is removed from the total, the baseline and the chart at once, or from none of them.
  • The plan forecast is a range It takes weekdays and holidays into account. While the outcome is unclear, the plan is not coloured.
  • Highlighting compares a metric with itself Across months or across segments. Zeros with no data and approximate numbers are not coloured.
  • The chart sits next to the table It shows how the metric changed, which highlighting cannot show.

A number without a comparison tells you nothing. In a KO report it is compared with four things, and each comparison answers its own question.

Previous period

Better or worse?

118 leads ↑ 12%

against 1–12 May

Plan

On track for the goal?

62% of plan

mark: 40% of the month has passed

Highlighting

Where is it higher or lower?

314463557892

leads by month, one row

Chart

Where is it going?

the same row, the shape of the change

Below: how each one works and where we chose not to follow the usual approach

Previous period: the same stretch of days

If the report shows the current month and today is 12 June, KO compares 1–12 June with 1–12 May. Not with the last twelve days before 1 June, and not with the whole of May.

Today is 12 June, the report shows this month

June
1–12 June still ahead
Baseline
1–12 May 13–31 May
Old rule
1–19 May 20–31 May
The old baseline took the end of May, when campaigns were already at full speed, so the start of every month looked like a failure

If the previous period is shorter, the stretch is cut to fit it: 31 March is compared with the whole of February. If the period has already ended, such as last week or last month, the baseline is also a full period. For the “Last 30 days” window, the baseline is the 30 days before it.

The rule is the same for a running month and a finished one, so Δ does not jump on the 1st.

An unfinished today

Today is still running, so its number is always smaller than yesterday’s. On a chart this looks like a drop at the end of the line.

The Exclude today option removes today from three places at once: the total, the comparison baseline and the chart. Otherwise the total would include today and the comparison would not. For the same reason we did not just remove the last point of the chart.

The option is set on a section, not on the report. By default it is on only in Metrics, because there an unfinished day distorts the picture the most. On the first day of a period, today is shown anyway, or the period would have no days left.

Plan: are we on track for the goal

The previous period tells you whether things got better. But the team did not promise “better than in May”. It promised a specific number. So KO has a plan: goals for metrics for a period. Both Plan vs actual and the Metrics cards read it.

A metric card with its name, value for the period, change against the previous period, plan line and trend chart
The card shows both answers side by side: the change against the previous period and the plan line

The percentage of plan and the forecast are calculated in one place for all sections. Otherwise one page could show two different answers to “are we on track”.

The forecast is more than the actual result stretched over the whole month. KO looks at how the project’s results usually spread over the days of the week, and it treats holidays as days off. We chose not to count only working days: if a fifth of the project’s leads come on Saturdays, such a forecast would miss them.

The forecast is shown as a range that holds eight cases out of ten. We chose not to use 95%, the usual choice in statistics: on daily data such a range is too wide to decide anything. The lower edge never goes below the actual result already reached.

If the plan falls inside the range, KO colours it neither green nor red: the outcome is not clear yet. In the first days of a period there is no forecast, because there is too little data. Rates and costs such as CR and CPL have no forecast either: they do not add up over a month, so there is nothing to stretch.

A running period is not coloured as falling behind: next to the actual result you see what part of the period has passed.

Without a plan, Plan vs actual works as a table of actual results by period: the plan column and the forecast disappear, and everything else stays. A Metrics card shows the percentage of plan only when its period matches the period of the goal. If the goal is for a month and the card shows a week, the percentage would mean nothing.

Highlighting: where it is higher

Highlighting colours table cells by size: you see which cell is higher than its neighbours and which is lower. The key question is which neighbours to compare with.

In Plan vs actual, highlighting goes along the row: one metric across periods. Comparing spend with conversion rate by colour makes no sense, because they have different units.

The Plan vs actual section by month: the toolbar above the table, metrics as rows, the Subtotal and plan goal columns on the left, then highlighted months, the chart panel on the right
Each row has its own scale and its own colour: months are compared within one metric

In Breakdown, it goes down the column: one metric across segments. In a nested Breakdown, campaigns of one channel are compared with each other, not with all campaigns at once. The Total and Other rows are left out of the scale, or they would always be the brightest.

Each metric has its own direction of “better”. For revenue, more is better. For CPL, less is better, and the colour is reversed for such metrics. The histogram is not reversed: the length of a bar is just the size.

A zero with nothing to divide by is not a record. The CPL of a campaign with no leads is zero not because the lead is cheap, but because there was nothing to divide by. Without this rule, such a row would get the strongest colour as the best one. So zeros of calculated metrics are left out of the scale.

In Plan vs actual, zeros of base metrics are not coloured either. Zero leads for a month there almost always means there is no data for that month, not that there were no leads. If such a zero were coloured, your eye would go to metrics with nothing to read.

An approximate value is not coloured either. A cell with the “~” sign shows the edge of the possible value, not a measurement, and it must not become a record. More on this below, in “When it is too early to compare”.

The fill colour is the metric’s colour. When the table coloured everything with one accent colour, the whole report looked the same and you could not tell rows apart. Now a table row and its line on the chart have the same colour.

Chart: where it is going

Highlighting shows which cell is higher, but not the shape of the change. Growth of 8% a month and a one-time jump in July look the same in a highlighted row. That is why the chart sits next to the table, not in its place.

To put a metric on the chart, click the dot at the start of its row. The line has the same colour as the row. Calculated metrics are drawn as dashed lines: a number calculated from other numbers should not look measured. The running period is hidden by default, because its line always looks like a drop.

When the chart has metrics of different scale, such as spend in thousands and leads in tens, KO does not draw the values themselves. It draws the change against the first period: the first month is 100%. This way no line lies flat at zero, and the tooltip on a point shows both the percentage and the value. The percentages go on the right axis.

When it is too early to compare

A 50% conversion rate from two visits and 50% from two thousand are different things. When there is little data, KO puts a dot next to the number. If there were no events at all, you see an edge value with the “” sign instead of a zero. In the demo project, one campaign spent €387 in 30 days and brought no leads. Its CPL is not zero and not a dash, it is “€387”: the first lead will cost exactly that much, and if it does not come, more.

For rates such as conversion rate, the edge comes from the 95% Wilson interval. We chose it because it gives a sensible answer even when there are no conversions, or when almost all visits convert. For costs such as CPL, the spend is known exactly and the number of leads is the random part, so the interval is built on the number of leads. The interval itself is shown in the tooltip on the number.

Approximate values do not take part in highlighting and do not become records. You can compare them, but keep in mind that they are estimates.

In practice

  • Δ against the previous period compares equal stretches of days, so the start of a month does not look like a failure.
  • The plan answers “are we on track”: the forecast takes weekdays into account, and while the outcome is unclear, the plan is not coloured.
  • Highlighting shows where a metric is higher, and the chart shows how it changed.
  • A number with the “~” sign is an estimate on small data: you can look at it, but it is too early to draw conclusions.

Updated

Demo

Imagine your project is already in KO

The demo is the project of a made-up company. Its ads, website and CRM are already connected. Reports, decisions and data checks work on its numbers just as they would on yours.

See it live

No sign-up, no payment