Cost of Goods Sold

There was a stretch when a company's health was reported as one number: sales. A business that sold $2.3 million last year and $4.6 million this year was twice as good as it used to be. The trouble was that a company spending $4.2 million to reach that $4.6 million sat in the same sentence as one spending $1.2 million.

Investor meetings kept stalling at the same point. A founder would present a growth rate, an investor would ask how much of that $4.6 million actually stayed in the company, and there was no answer to build if the books recorded outgoing money as a single pile. Salaries, server bills and ad spend all sat on one line, so nothing separated the money spent producing one more sale from the money that would have gone out regardless.

Cost of goods sold (COGS) is the line that appeared when spending got split into two piles. One question does the splitting: would this money have gone out if that sale had never happened? If not, it belongs here. If it would have gone out with zero sales, it belongs to operating expenses instead — rent, administrative salaries, everything it takes to keep the lights on. Once the piles are separate, subtracting only the first one from revenue leaves the amount the business truly pockets per sale.

Pick up each cost and ask the same question of it. If the cost rises when one more customer signs up, it goes into COGS. If it comes in at the same size every month no matter how many customers there are, it stays out.

For a company selling software by subscription, four items usually land in this pile.

1. Servers and storage — the more documents customers pile up and open, the higher the bill runs.
2. Payment fees — the payment processor takes 2 to 3 percent of every charge. Double the revenue and this amount doubles with it.
3. Customer support salaries — more users means more tickets, and past a certain line it means hiring another person.
4. Third-party services — what you pay per use for another company's feature running inside your product. If document translation comes from an outside vendor, each translation adds to this line.

The salaries of the engineers who built the product do not belong here. The code they already wrote runs the same whether there is one customer or ten thousand. The exception is a person assigned to one enterprise account and nothing else: that salary is COGS, because the role disappears when the contract does.

Calculating it is a matter of adding up those items for the period. If Pagedeck billed $75,000 last month and paid $9,000 in server bills, $2,000 in payment fees and $11,500 in support salaries, COGS comes to $22,500. Revenue minus that is $52,500 of gross profit, and dividing it back into revenue gives a 70 percent gross margin.

The raw figure matters less than its share of revenue, and three situations ask for that share.

The first is judging whether the business itself is strong. Subscription software treats a gross margin of 75 to 80 percent as healthy. A company at the same revenue scale reporting 40 percent is telling you that every customer needs people attached to stay served — that shape sits closer to an agency wearing software's clothes.

The second is setting prices. On a $7.50 monthly plan where servers and support cost $2.30 per person, the remaining $5.20 has to cover salaries and marketing. Teams that set discount rates without knowing that number end up with a plan that loses more money the better it sells.

The third is tracing profit step by step. Operating profit is revenue minus COGS and then minus operating expenses. Keeping the two kinds of spending apart is what lets a team tell whether a bad quarter came from how the product is sold or from how large the company's overhead grew.

One timing rule goes with it. COGS is recorded in the month the matching revenue is recorded, not the month cash left the account. A year of fees collected in January is booked as revenue month by month, so the server costs of serving that customer spread out the same way. Money still parked in deferred revenue carries no cost with it yet.

"Revenue grew 20 percent this quarter but COGS grew 35 percent."

Volume is up and the margin on each sale is shrinking, so somebody needs to find out why.

"Book that feature's server bill under COGS, not operating expenses."

The cost scales with the customer count, so it belongs where it affects gross margin.

"Enterprise deals carry a higher cost ratio, so don't judge them on the revenue number alone."

Dedicated support and custom work ride along with those contracts, leaving less behind per dollar billed.

The week Pagedeck was assembling its next funding deck, Grace sat down with Mia to re-sort the last quarter. Support salaries had been filed entirely under operating expenses, which put the gross margin at 82 percent, but most of that work was answering tickets from teams that had already paid — it belonged in COGS. After the move the margin read 71 percent.

Grace put the lower figure in the deck anyway. Eighty-two percent was a number that would collapse on the first question; 71 percent came with an explanation attached, namely that enterprise accounts had grown as a share of revenue and brought their costs with them.

The most common belief is that software has no cost of goods sold at all. An old picture lingers: build the file once, copy it to a million people, pay nothing extra. Subscription products don't work that way, because the spending happens across the whole time someone uses the product rather than at the moment of sale. Stored documents have to keep sitting somewhere, charges run every month, and tickets arrive every month too. Price a product as though the cost were zero and the bank balance drains faster the more people sign up.

Reading it as lower-is-always-better goes wrong in a subtler way. It is a cost, so cutting it does raise what's left, and halving the support team really does drop next month's cost ratio. The damage shows up later. Once replies start taking three days, cancellations climb, and a few months on the ratio looks better while revenue itself has shrunk. The useful question isn't how much cost came out but how much revenue that spending held on to.