A demand drop is a fall in inbound inquiry volume that has not yet reached reported revenue. In a services business it arrives about one sales cycle ahead of the revenue line, which makes it the earliest commercial signal a company can act on and one of the easiest to misread: a company-level inquiry number can fall while nothing in any single site or job type has moved enough to act on, and it can hold steady while one site loses half its demand.
This guide starts after the attribution rather than before it. A revenue miss decomposes into inquiry volume, conversion, job size and capacity, and deciding which of the four moved belongs to the monthly commercial operating review. What follows is the question that guide hands off: the miss is a volume problem, so what happened to demand? Whether the comparison is valid at all in a business whose year is not flat belongs to running a seasonal services business under PE ownership, and that base has to be straight first. The opposite problem, demand the business cannot deliver against, belongs to selling into a capacity ceiling.
The first move after a volume miss is not a hypothesis. It is a grain. A company-level inquiry count in a multi-site services business is the sum of several unrelated local events, and the sum is the one number that cannot say which event happened.
Take a platform of eight sites reporting inquiries down nine percent. That figure is consistent with two completely different situations. Either all eight are down between six and twelve percent, which points at something affecting the whole market or something the platform did to itself, or one site is down forty percent and the other seven are flat, which points at something local to one town. The first is a strategy conversation. The second is a phone call to one general manager. The nine percent is the only version of the number that supports neither.
The reverse case is worse, because nobody goes looking for it. A platform count that holds steady can hide a site that has lost half its demand, offset by a second site ramping or by an acquisition that joined the reporting mid-period. Growth hides decline for as long as the aggregate is the only thing anyone reads.
Three grains do almost all the work. Site, because demand is local and so are its causes. Job type, because different jobs have different buyers and different triggers, and a shift in mix reads as a volume fall at the revenue line while inquiry counts are intact. And source, because a change in one channel is the most common single cause and the only grain that identifies it. Read each as a count per period rather than as a share of the total, because a share chart in a falling market shows nothing falling.
The source grain is also the one most often missing. In one engagement, 38 percent of revenue turned out to arrive through word of mouth that no system recorded, so more than a third of the demand had no source to read. Where that is the position the source grain is not weak evidence, it is unavailable, and the honest diagnosis says so rather than drawing conclusions from the tracked remainder. A single source of revenue truth is the fix, and a project rather than something to start mid-drop.
A sudden fall in reported inquiries is often a change in the counting rather than in the demand. It is also the cheapest cause to rule out, which is why it goes first.
A measurement break has a signature. It starts on a date rather than along a curve, because code changes on a day and demand does not. It usually appears in one channel, form or phone number while the others hold. It follows something somebody did: a site release, a new tracking number, a tag change, a redirect, a switched phone system. And it disagrees with any count the marketing stack did not produce.
That last point is the actual test, and it needs one number rather than an audit. Pick a count the business holds independently, such as quotes issued, booked jobs, or calls logged at the carrier, and compare its trend against reported inquiries over the same periods. If the independent count fell too, the demand change is real and the rest of this guide applies. If it held, the problem is instrumentation, and the remedy belongs to fixing marketing attribution: one branch handed off, not the subject of this piece.
Once the count is trusted and the grain is known, a demand drop resolves into four causes. Each has an observable signature, and the signatures differ enough that the right evidence usually settles it without a research project.
More than one is usually true at once. The question worth answering is not which causes exist but which carries most of the lost volume, because that is the only one whose fix changes the number. A diagnosis that lists four contributing factors and sizes none of them describes the drop rather than explaining it.
This is where a drop becomes expensive, and not because of the lost revenue. The reflex response to falling inquiries is to buy more of them, and that answers exactly one of the four.
A channel supply change is answerable with money, and substitution is a known piece of work at a known cost: find the demand elsewhere, and move some of it onto sources the business owns rather than rents. That is the better outcome of a bad event and worth doing deliberately rather than in a panic. Owned demand is also cheaper once it exists, which is why work of this kind has produced a 60 percent reduction in cost per acquisition alongside a threefold increase in owned-channel share in our own engagements.
A local market change is not answerable with more spend into the same market, because bidding harder inside a smaller pool only raises the price of what remains. The real options differ in kind: cover a wider area from the same site, shift the mix toward job types that still have demand, or accept a lower run rate there and stop funding it as though the old number were recoverable.
A price position problem is answerable only by a pricing decision. Spending into it buys more conversations the business goes on to lose, at a cost per booking that rises while the marketing report shows rising volume and the revenue line does not move.
A contraction is answerable by nobody, and this is the claim worth carrying out of this guide. Spending into a contraction buys roughly the same volume at a worse price, and the business almost never files that correctly. It files it as a cost per acquisition problem, sending the next round of effort at the marketing function instead of the plan. The honest response is to reset the number and decide what to hold, not to defend the old number with a larger budget. Where spend does have to come down, the order matters more than the amount, and that has its own method in cutting a marketing budget without cutting pipeline.
Most of the damage done in response to a demand drop comes from acting on a number that was always going to move. Weekly inquiry counts at a single site are small numbers, and small numbers bounce. Without the normal range in front of you, every reading is an event.
The fix is cheap and almost nobody has it. For each grain that gets reported, take eight to twelve trailing periods, write down the normal high and low, and agree the threshold that counts as a move before anybody needs it. Doing that once, in a calm week, is what keeps the conversation during a drop about the cause rather than about whether the drop is real. Thresholds argued over mid-fall get set wherever the argument ends.
Breadth is then stronger evidence than depth, and this is the part that gets read backwards. One site down twenty percent in a week may sit comfortably inside its own band and mean nothing. Every site and every job type down seven percent in the same week is almost certainly real, because independent grains do not move together by accident. The instinct is to chase the big local number and ignore the small broad one. The small broad one is the signal.
Where the move sits inside the band, the wait is measured in sales cycles at the grain that moved rather than in calendar time, because the cycle decides when the drop reaches revenue and so when waiting stops being free. A business selling same-week work has already seen the consequence. One with a long quote-to-decision cycle is looking at demand that lands two quarters out, which buys time to diagnose and removes the excuse for guessing.
A demand drop reaches the sponsor before the cause does, and the pressure that follows is what produces the premature plan. Four statements hold the line without claiming more certainty than exists. Which grain moved, by how much, against what normal range. Which causes are live and which the evidence currently favors. What would settle it, and when. And what is deliberately not being changed yet, with the reason stated, because an unexplained absence of action reads as drift while an explained one reads as judgment.
What not to bring is a drop with a plan already attached. It commits money to a cause nobody has established, and it makes the diagnosis harder to revise, because revising it then means conceding the spending was wrong.
Three failures account for most of the cost. Diagnosing at company level, so the response is sized to an average and lands on no site in particular. Treating the first bad week as a trend and the second as a crisis, which moves spend twice in opposite directions and makes both moves unreadable. And buying demand by reflex, which answers one cause of four.
Done in order this is short work: trust the count, find the grain, name the cause that carries the volume, then decide whether money is the answer. Sponsors are paying closer attention to work of this kind than they once did. In S&P Global's 2026 survey, 71 percent of GPs and 53 percent of LPs named operational value creation as a priority. How we work sets out the method, and results covers what it has produced.
A demand drop is a fall in inbound inquiry volume that has not yet reached reported revenue. In a services business it arrives about one sales cycle ahead of the revenue line, which makes it the earliest commercial signal a company can act on and one of the easiest to misread: a company-level inquiry number can fall while nothing in any single site or job type has moved enough to act on, and it can hold steady while one site loses half its demand.
Reconcile the inquiry count against one number the marketing stack does not produce, such as booked jobs, quotes issued or the dispatch log. A measurement break usually starts on a single date rather than curving down, shows in one channel or one form while independent counts hold, and follows a change somebody made to a website, phone number or tag. Real demand bends; tracking falls off a cliff.
At site, job type and source before company level. A company-level fall of nine percent can be eight sites each down a little, which points at the market, or one site down forty percent with the rest flat, which points at something local. The two have different causes, different owners and different costs, and the company-level number cannot tell them apart.
When the move is outside the normal range for that grain, or when several independent grains move together in the same period. Breadth is stronger evidence than depth, because separate sites and job types rarely fall at once by accident. A single noisy week at one site is worth watching for one sales cycle at that grain rather than acting on.
Only when the cause is one that money answers. A supply change at a channel can be substituted and a service area change can sometimes be covered differently, but a price position problem is a pricing decision and a genuine market contraction is answerable by nobody. Spending into a contraction buys the same volume at a worse price, and the business files the result as a cost per acquisition problem.
Have a revenue problem the board is asking about? Start a conversation.