Ratio Metric

Counts stop telling you anything at a certain point. "1,200 signups this month" says nothing without last month's figure, and if that was also the month advertising spend doubled, 1,200 may be a worse result than 900.

Pagedeck's weekly report had this shape. Marketing site visits at 80,000, signups at 1,200, both larger than the week before. Dividing the two, Ravi found visit-to-signup had gone from 1.7 percent to 1.5 percent. Traffic had grown and signups hadn't kept pace, and a table of counts side by side made that invisible.

Carrying the denominator along means the number reads on the same scale whatever the size. The cost is that the same metric name means different things depending on which value sits underneath, which is why this form always arrives with a written definition attached.

Decide what goes on top and what divides it, then check that both values come from the same period and the same population.

1. Write the numerator and denominator as sentences. For signup conversion: accounts that completed signup this week, over visitors who saw the marketing site in the same week. Those two lines belong in the metric definition document, or next month's figure won't match.
2. Align the periods. Count the numerator weekly and the denominator over 30 days and the value shrinks by a factor of five. Different periods don't produce a ratio, they produce something else.
3. Align the populations. If internal staff and automation accounts are in the denominator, they have to be out of the numerator too — and if they're excluded from only one side, the value is wrong in a direction nobody notices. Recording which exclusions were applied is what lets a data quality check verify the metric later.
4. Decide between people and events. Counting one person's three visits as three or as one changes the value threefold, and both choices are defensible as long as the definition says which one is in use.
5. Check when the denominator is finalised. Dividing "created a document within seven days of signup" by this week's signups puts people whose seven days haven't elapsed into the denominator, which drags the figure down. Metrics like that use only cohorts whose window has closed, which means the most recent week is always missing from the chart.
6. Don't read the value when the denominator is small. Two signups from 40 visits is 5 percent, and one person moving makes it 2.5 or 7.5 percent. Printing the denominator next to the value keeps numbers like that from being quoted as evidence in a meeting where nobody has the raw counts to hand.

Comparing against an industry average settles nothing. A reported 2 percent signup conversion tells you nothing without knowing its denominator: divided by marketing site visitors versus divided by people who reached the pricing screen differs by a factor of five or six.

Build the standard from three places instead.

1. Your own history. Put the last twelve weeks measured under the same definition side by side and see whether this week sits inside that range. Weekly values swing, so read a four-week moving average alongside, and draw a line at any point where the definition changed — comparing across that line is the most common way a ratio gets misread.
2. What the business requires, calculated backwards. If you need 30 new paying teams a month against 80,000 visits, then 1.5 percent signup conversion times 3 percent signup-to-paid gives 36. When that arithmetic doesn't work, it shows which of the two values has to move, and whether the gap is small enough to close with a better screen or large enough to need more traffic.
3. The same metric split by segment. Behind an average of 1.5 percent there may be 4.2 percent from search and 0.6 percent from advertising. The average alone shows nothing worth acting on; acquisition channel, plan, and company size are the usual ways to cut it.

Causes of movement live on both sides of the division. If the numerator held steady and the value fell, the denominator grew — and that usually means the population changed rather than the product got worse. Launch a new campaign, bring in visitors with thinner intent, and conversion necessarily drops even though the number of signups went up. So when the figure moves, don't read the ratio alone; lay the numerator and denominator counts out next to it.

"Signups are up and conversion is down. The denominator grew faster."

Explaining opposite movement in the count and the ratio.

"What's the denominator here — visitors, or people who saw pricing?"

The same metric name differs several-fold by definition.

"The denominator is 40, so I'm not reading that 5 percent yet."

The sample is too small for the value to hold still.

When Ravi reworked the weekly report, each ratio got its numerator and denominator in brackets beside it: 1.5 percent, then 1,200 and 80,000. After that, "conversion moved" came with the answer to which side moved on the same line. Two weeks later a second table split by acquisition channel went in, and the 0.6 percent conversion from advertising appeared for the first time.

The most common belief is that ratios are simply better than counts. Carrying a basis for comparison looks more rigorous. But a ratio erases scale. Eight percent conversion on 50 visits is four signups; 1.5 percent on 80,000 visits is 1,200. Judging the size of a business needs the counts, which is why working reports keep both side by side and why a dashboard of pure percentages is hard to act on. Neither substitutes for the other.

The other habit is describing a change in percent when the metric is already a percentage. Going from 1.5 percent to 1.8 percent gets written as "up 0.3 percent," when the gap is 0.3 percentage points and the growth is 20 percent. Mix the two and the same change reads more than twice as large in one meeting as in another. State differences in percentage points, growth as a percentage, and put the unit in the table so nobody has to ask.