Orders over visitors, in one period
The numerator is paid orders, not views or add-to-cart clicks. The denominator is the traffic those orders came from, measured the same way for the same period - sessions or unique visitors, but not a mix of both.
Divide and multiply by a hundred and the result is the percentage of the traffic that bought: 138 orders from 3,000 visitors is 4.6%. The page prints the rate to one decimal, which is enough to see a change without implying precision the underlying numbers do not have.
If visitors are zero there is no ratio to compute, so the page asks for the traffic rather than dividing by nothing. Both counts are rounded to whole numbers and floored at zero before the arithmetic starts.
The reverse is the planning half
At a target rate, a target order count needs a predictable number of visitors: divide the orders by the rate expressed as a fraction, then round up, because half a visitor cannot walk through the door. The defaults show the shape - 200 orders at 4% need 5,000 visitors.
Rounding up matters at small volumes: 10 orders at 3% works out to 333.33 visitors, and the honest planning answer is 334. A plan built on a rounded-down figure starts a visitor short.
If the target rate is zero the reverse has no answer, and the page prints that instead of a number. A conversion plan with a zero in it is a prompt to choose a rate, not an invitation to divide by zero.
The traffic unit has to stay steady
Sessions from one analytics tool and unique visitors from another are not the same denominator, and mixing them quietly changes the meaning of the rate. Pick one unit, use it for the current period, and use the same one for the target side.
The same caution applies to the period. Thirty days of orders against a calendar month of visitors produces a plausible-looking percentage built from two different windows, and nothing on the page can detect that for you.
A rate above 100% is the visible symptom of mismatched inputs: more orders than visitors usually means one number came from a different definition or a different date range, not that every visitor bought twice.
A good overall rate can still hide a leak
The page treats conversion as one ratio, and its list is explicit that the headline can hide a checkout step where most visitors drop out. The aggregate is the start of the conversation, not the whole of it.
Read the ratio with its two inputs attached. The same 138 orders on half the traffic is a doubled rate and an unchanged business; the improvement is real only when the traffic moved for a reason you can name and repeat.
The page does not break the journey into add-to-cart and checkout steps, and it does not attribute traffic to a channel. Those are separate measurements; this page answers what the store converts at, and what the target costs in visitors.
The denominator decides the story
A rate is only comparable when the denominator is the same unit. Three thousand sessions from 2,400 people who visit twice each is a different business from 3,000 people who visit once: the 138 orders are 4.6% of sessions and 5.75% of people, and both figures are true. Pick sessions or users, write the choice down, and compare months only inside that choice.
Order counts carry noise that the decimal point hides. At 3,000 visitors and 138 orders the 95% confidence interval around 4.6% runs from roughly 3.9% to 5.4% - the standard error of a proportion is sqrt(0.046 x 0.954 / 3000), about 0.38 percentage points, and two of those either side of the estimate. A change of half a point at this sample size proves nothing; at 300 visitors a single order moves the rate by 0.33 points, so one quiet Tuesday cannot settle a test.
The headline rate is a product of the steps, not a sum. If 55% of visitors add to cart, 60% of those start checkout and 78% of those pay, the store converts at 25.7% (0.55 x 0.60 x 0.78). Because the steps multiply, five points added to the cart step (55% to 60%) take the total to 28.1%, while the same five points on the payment step take it only to 27.4% - equal effort, unequal return, because the later steps carry less of the traffic. The funnel, not the headline, decides where the next week of work goes.
Traffic that never intended to buy dilutes the ratio without saying anything about the store: a support page, a size chart, a blog post that ranks for "how to measure". Segment by landing page before you compare two months, or the marketing report will read as a merchandising failure.
Danfio