The agency's pitch was to double Brightline's ad budget. Its own numbers said not yet — one channel is too new to trust, and the other is already running out of room

The moment

Brightline Heating & Air runs six vans out of the same Dayton, Ohio yard as the true-cost-per-customer and price-increase stories: six technicians, a dispatcher, an office manager, two apprentices and the owner. The replacement-install line — full system replacements sold from paid leads rather than an existing relationship — is the same one true-cost-per-customer already priced out in October: $3,000 a month in ad spend, 25 systems closed, $120 a customer on the agency's report, $238 to $268 once the office manager's selling time, the $60-per-install commission, the agency's fee and the discounts given to close were all added back.

It is the second week of November. The decision that number was actually for is still open: the agency's pitch to double the budget to $6,000 a month, on the promise that $120 a customer holds at twice the volume. What the owner had never done is ask what the $3,000 is actually two things, not one — the agency runs it through Google Local Services and Facebook, and the monthly report has never once shown what each of those two is earning on its own.

What they did instead

The report arrives as a single number every month: $3,000 spent, 25 systems closed, $120 a customer. Saying yes to the agency's pitch doubles both channels at once. Saying no leaves both alone. Neither answer asks whether the next dollar would do more in one channel than the other, because nobody at Brightline had ever pulled the combined report apart into its two lines long enough to ask.

There was also a second question nobody had separated from the first: is more marketing even the right thing to be buying more of right now, or would a bigger ad budget just bring in leads the office manager does not have the hours to quote and the crew does not have the hands to install. That question and the channel question had always been treated as the same decision, answered together with one yes or no to the agency's email.

The questions a CFO asks first

1. Before any channel gets compared: is demand actually what is holding the business back right now, or is the real limit capacity, cash or staff — in which case more leads make the problem worse, not better?
2. If demand is the real constraint, what has each channel returned on average so far, backward-looking, on the spend already going into it?
3. What does the owner honestly expect the next $100 into each channel to bring — not the average, the marginal, and stated as a range because nobody can know it exactly in advance?
4. Does either channel's honest marginal guess sit well below what it has returned in the past, which is what running out of room actually looks like?
5. Is there enough history behind a channel to give an honest guess at all, or is the honest answer that the guess does not exist yet?

Question 5 is the one Brightline had not separated from question 3 before running this — a channel that is only two months old does not stop being askable, but the honest answer to "what will the next $100 bring" can legitimately be "nobody knows yet," and the tool treats that answer differently from a guess dressed up to look confident.

The arithmetic, shown

The same $3,000 and 25 installs true-cost-per-customer already put on the record, split into the two lines behind the combined report: $2,000 a month into Google Local Services, crediting $147,600 in revenue — 74x on average, backward-looking. $1,000 a month into Facebook, crediting $57,400 — 57x on average. Facebook has only been running since September; Google has been running for years.

Scenario one — the honest answer this month. Demand is genuinely the constraint — the owner wants more replacement-install leads, not fewer, and neither the crew nor the office manager's calendar is the thing stopping that. But two months of Facebook data is not enough for an honest guess about what the next $100 into it would bring, and the owner would not put a number on it. Left blank, the tool does not invent one:

Nothing ranks. The tool's own missing-numbers list, quoted exactly as it appears, holds two lines: "Facebook's low guess for the next $100" and "Facebook's high guess for the next $100". No ranking runs without them, and no ranking ran this month.

Scenario two — the guess the owner was tempted to enter anyway. Early Facebook numbers looked promising enough that it would have been easy to type in a number just to see the ranking. Run with that temptation given in honestly: a $150 to $400 low-to-high guess for Google's next $100, and a $1,500 to $4,000 guess for Facebook's, both stated as the range the missing-numbers rule asks for rather than a single invented figure. On that input the tool ranks Facebook first — its low guess of 15x beats Google's low guess of 1.5x by a wide enough margin that the ranking is not close, even though Google has earned more on average historically, 74x against Facebook's 57x. The tool's own flag on Google, quoted exactly: "Google Local Services shows signs of saturation". The reason it gives, also quoted exactly: "your own guess for what the NEXT $100 brings is well below that". Sales are not what the $100 has to pay for itself out of, though. Brightline keeps about $40 of every $100 of install sales once the equipment and the crew's hours are paid - the tool asks for that before it ranks anything - so the same guesses read differently as profit. Facebook's next $100 comes out from making $500 to making $1,500. Google's comes out from losing $40 to making $60: at the low end of the owner's own guess, the next $100 into Google loses money once the install it wins has been delivered. That is the arithmetic behind the pitch to double the budget falling apart on inspection — a uniform doubling sends half of the new money into the one channel already showing it is running out of room, on the owner's own guess.

**Scenario three — testing the first question, not the channel question at all.** What if the honest answer to "what's actually limiting the business" had been staff, not demand — the office manager's selling time, already a fifth of her week per true-cost-per-customer, genuinely out of room for more quotes rather than demand being the open door it looks like. On that answer the tool stops before any channel is ever asked about. Its own headline, quoted exactly: "The next marketing dollar isn't the fix — there aren't the people yet to deliver more work." Its own reasoning, quoted exactly: "More marketing brings in more demand, and right now there aren't enough hands to deliver on it. Extra customers without extra people to serve them just means more work sitting unstarted. Fix staffing first; ranking channels can wait." Nothing about Google or Facebook gets computed at all — the tool's own design treats that as correct, not as a missing feature.

What changed

The ad budget did not double. Google's spend stayed at $2,000 a month — even the owner's optimistic guess for its next $100 sits far under what the channel has already earned, which is what saturation looks like on this tool's own terms, and more money into a channel already showing that sign is not what the agency's pitch was actually promising. Facebook's spend also stayed at $1,000, not because the early numbers looked bad but because two months is not enough history to trust a guess on, and the tool would not rank one that was typed in just to get an answer. The agency was told to hold the combined budget at $3,000 and to revisit Facebook once it has run long enough — two more months, the owner decided, matching how long it had already run — for an honest guess to exist.

Nothing here is a big number. It is a decision not to spend $3,000 more a month on the strength of a report that had never separated two channels earning very different returns, and a decision not to invent a guess just to get past a blank field.

What the tool will not do

It will not invent a guess when the owner leaves one blank, even when a ranking would otherwise be one field away. Facebook's low and high guesses for the next $100 are both required before anything ranks, and a channel that is genuinely too new to guess about stays out of the ranking rather than being assigned a number nobody actually believes.

It will not treat "more customers" as automatically good. The first question it asks is what is actually limiting the business, and if the honest answer is capacity, cash or staff rather than demand, it stops before a single channel gets compared — more marketing spend on top of the wrong constraint does not get modeled, because modeling it would make a bad decision look like a considered one.

It will not rank channels on which one has earned the most in the past. The backward-looking average sits beside the ranking, clearly labeled as backward-looking, but the ranking itself runs on the owner's own marginal guess for the next $100 — because a channel that earned well on average can still be the wrong place for the next dollar if it is running out of room, and a channel with a shorter track record can still be exactly where the next dollar belongs, once there is enough history to trust a guess about it.

It will not decide how much total budget the business should be spending on marketing at all. It ranks where a given amount of additional spend is most likely to earn something back; it does not size that amount, and it does not weigh marketing spend against every other use of the same dollar.