One row per channel, two columns each
Each row is one acquisition channel with the money it consumed and the new customers it is credited with. The rows are the input, so ads, software, influencer fees and anything else that buys customers each get their own line rather than being pooled into a single budget figure nobody can unpack later.
Spend that cannot be tied to a named channel - a printed mailer, a platform fee, an agency retainer - goes in the other spend field. It joins the total spend but is not credited with any customers, so it raises the blended cost without pretending to have produced a sign-up. The page needs money on one side and customers on the other; with either at zero it asks for both instead of dividing.
The blended cost divides the totals
Total spend divided by total new customers is the blended cost of a customer. The default rows - $2,400 across three channels for 125 new customers - produce $19.20, which is what one customer costs when everything is counted at once.
Blended is the number a budget conversation usually needs, and it is also the number that hides things: a cheap channel and an expensive one average into the same figure, so a healthy-looking blended cost can sit on top of a channel that is failing. The hero and the last row of the table carry the same figure deliberately, so the total can be traced back to the parts that produced it.
The per-channel rows divide each channel alone
Each row also gets its own cost per customer: its spend divided by its customers. That is where a channel spending twice as much for half the result becomes visible, instead of disappearing into an average that is carried by the other rows.
A channel with spend but no customers shows as not defined rather than as zero, because dividing by nobody has no answer and free is the wrong word for money that returned nothing. A row whose name was left blank is labeled by its position - Channel 1, Channel 2 - so an unnamed line is still identifiable and cannot be mistaken for a real brand.
The assumption the page refuses to make
The page takes your numbers as given, and attribution is the assumption it refuses to make. Which customer a channel truly earned is an estimation problem that no arithmetic can settle: the same ad spend may be reported by two platforms, and part of a campaign may never be attributed at all.
That matters more than the second decimal place. The per-channel view is only as honest as the attribution behind it, so the rows should be read as your own reporting restated rather than as a verdict on which channel works. The count is acquisition spend only - the support and delivery costs of serving the customer afterwards are not in it.
The period, and the payback test
Spend and new customers have to cover the same window, or the answer is off by the size of the mismatch: a quarter of spend divided by a quarter of customers lands in the same place as a month divided by a month, but a quarterly spend divided by a monthly customer count overstates the cost twelve times over, and the reverse understates it just as badly. Pick the window your reporting already uses - a month, a quarter - fill both columns for that window, and compare one period with the next rather than with a published industry average, which was measured on a different mix of channels and a different price list.
A one-off fee belongs to the period it serves. An influencer paid once for a campaign that keeps earning for a quarter costs less per customer when the invoice is spread over the three months it works on than when the whole of it is dropped into the month it was paid, and the same goes for an annual tool subscription. Spreading it is an estimate, so say which convention you used; a number with its convention attached can be checked, and a number without one cannot.
What the number is for
The cost of a customer only becomes a decision when it is set beside what that customer is worth: the gross profit on their first order, or the gross profit accumulated over a payback window you choose. If the default $19.20 sits under a $30 first-order margin, every new customer repays the cost of winning them on the first order. At $60 a customer against the same $30, each one carries $30 of unrecovered spend, which two orders - or two months at $15 of profit each - take to repay.
Either way the comparison is the same one: the cost of a customer against the profit that customer leaves behind before they stop buying. A cost below that figure is growth the customers pay for themselves; a cost above it is growth paid for out of the bank account, which is a legitimate choice for a while and a fatal one if it never closes. That is the question the blended number and the per-channel rows are there to answer, and it is why the page prints both instead of stopping at the average.
Danfio