They asked whether you could start in three weeks. You said yes, and it wasn't a lie or a stretch — in the moment it genuinely seemed right. Three weeks is a while. The current projects felt like they were winding down.
Then three weeks arrive, and it turns out two of those projects had a second phase you'd stopped thinking about, one client came back with revisions, and your best person is on leave for a week you knew about and hadn't connected to this.
Now the new engagement starts late, or it starts on time and something else quietly slips. Either way somebody is being disappointed by a business that meant well.
The thing that should worry you isn't the occasional bad month. It's that you can't say, right now, without asking anyone, how much work your business can take on in the next six weeks. Not roughly. At all.
Two different failures wearing the same clothes
Every missed deadline is one of two problems, and they have completely different fixes. Founders almost always try to fix the second one.
Problem one: was there room? Did the business have the capacity to absorb this work alongside everything already committed. This is a question about aggregate load and it has nothing to do with any individual project.
Problem two: how long does this take? Given that there's room, is the duration you promised achievable. This is an estimating question, about one piece of work.
They produce identical symptoms. A late project looks the same from the client's side regardless of which one caused it, and from the inside both feel like we underestimated.
You can estimate a project perfectly and still miss the date, because a perfectly estimated project can't land on time in a business that took on more than it can run in parallel.
Which is why founders who invest in better estimating and still miss dates conclude that their estimating is still off. It usually isn't. They fixed number two while number one was the binding constraint.
Why the first number doesn't exist
The commitment and the data live in different rooms.
Commitments are made in sales conversations — live, fast, optimized for momentum. Capacity is a property of delivery, which is distributed across several people, several projects and several weeks, and is only knowable by aggregating things nobody has aggregated.
For the yes to be responsible, information would have to travel from the second room to the first, in seconds, at the moment of asking. In most founder-led businesses no channel exists for that. There's no number to look at.
So the founder does what any human does with a missing input under time pressure: substitutes an intuition, and doesn't notice a substitution has occurred.
This is the part that matters. It doesn't feel like missing information. It feels like a judgment call — and judgment calls feel like the kind of thing a founder is supposed to be good at, which is exactly why nobody goes looking for the instrument.
And the intuition is optimistic for a structural reason rather than a character one. What you can recall about the next six weeks is what's currently on your mind, and what's currently on your mind is the work in progress. Not the phase two that hasn't been scheduled. Not the revision round that always happens. Not the two days a week that go to running the business. Your mental model of your own capacity is built from the visible half.
Building the first number
The requirement is specific: at the moment of a yes, a number must already exist. Not a report you could produce, not a conversation you could have. A figure that's already there, because there's no time to make one mid-call.
Four parts.
Committed load, not available time. For each active and signed engagement, remaining effort and the weeks it spans. Sum it. Available time is the thing you're bad at estimating; committed load is a fact.
A rolling eight-week window. Three-week horizons are exactly the length that hides phase twos and booked leave. Eight weeks is far enough out that the invisible half becomes visible and near enough that the numbers are real.
The load everybody forgets. Booked leave. The recurring internal commitments — the weekly review, the sales calls, the admin — that consume delivery time and appear on no project plan. A revision allowance based on what actually happens rather than what's scoped. In most service businesses these three add up to somewhere between a fifth and a third of nominal capacity, which is precisely the margin that turns a well-planned quarter into a difficult one.
Updates as a byproduct. Remaining effort has to decrease as work moves through stages that already exist. If updating it is its own task, the number will be accurate for three weeks and fiction by week six.
For a business with five or fewer active engagements, that's a spreadsheet and an afternoon, and it works.
The threshold is the actual mechanism
Then set a line — a percentage of true available capacity, deliberately below a hundred — and decide in advance what happens on each side of it.
Below: sales can commit to a start date inside the window. Above: the only available commitments are a later start or a waitlist.
The exact percentage matters much less than the fact that it was decided calmly, in a room with no prospect in it. That's the whole point. The threshold converts a judgment made under social pressure into a rule made in daylight — and every founder who installs one describes the same relief, which is that the hard conversation with the prospect stopped being a negotiation with themselves.
Keep an override, and record it. Sometimes you take the work anyway, for a strategic client or a reason that outweighs the rule. The system's job isn't to prevent that. It's to make the decision conscious, and to make it visible a quarter later that four overrides in a row preceded a bad month.
Plan to a hundred percent of theoretical capacity and you'll fail on your first ordinary week — then read that failure as a people problem.
Then, and only then, the second number
Once there's room, estimating becomes worth improving, and it improves through one habit: recording planned versus actual effort on every completed engagement.
That's it. Not better estimating technique. A ratio, per engagement type, derived from your own history. It's usually somewhere between a fifth and a third over, it's remarkably stable, and applying it turns a plan that fails every quarter into one that holds.
There's one more rule worth enforcing: estimates come from whoever will deliver, not from whoever sold. Optimism at the sales end is structural rather than dishonest — the person in the conversation is exposed to the client's enthusiasm and not to the delivery team's queue.
What this actually buys
Not just fewer late projects, though you'll get those.
Your existing clients stop paying for your new ones. A new engagement doesn't create new hours, so it's funded by attention taken from work that's already sold. Nobody is told. The existing client experiences a business that got slightly slower and slightly less thorough, and has no idea why.
You can price. A business that can't see its capacity can't know when it's full, and a business that doesn't know it's full can't raise its rates. Scarcity is invisible to you, so you keep taking work at the old number long after demand would have supported a higher one.
Your team believes the plan again. People who are repeatedly given more than fits stop believing the schedule, and once they stop believing it they stop flagging problems early — because flagging feels pointless. That's the quiet beginning of a delivery culture where bad news travels slowly.
The constraint that isn't the team
One correction worth making before you build the number, because it's the most common way a capacity model produces a confidently wrong answer.
Aggregating capacity across a team hides single-resource bottlenecks completely. If every engagement needs a specific senior person for its first two weeks — for scoping, for the technical decision, for the client relationship — then your capacity is that person's availability, regardless of how many people you employ.
A team-level number will show room that doesn't exist, and the failure looks exactly like poor estimating: everything is fine on paper and the work still doesn't fit.
So before summing anything, ask which step in your delivery process only one or two people can do. Then model that pool separately. Most service businesses have two or three constrained pools rather than one — the senior reviewer, the general delivery capacity, and whoever does the specialist step — and the threshold has to be evaluated against the tightest one.
The aggregate number is still worth having for the sales conversation. It just can't be the thing you check, because the aggregate is never the constraint.
The lever nobody names
Capacity is one of the three levers of the Execution pillar — clarity, cadence, capacity — and it's the one that gets the least attention because it's the least visible to clients.
Clarity is what the client sees. Cadence is what the team runs on. Capacity is the one that decides whether the other two are achievable at all, and it's the one almost nobody instruments.
There's a moment every entrepreneur hits where hustle stops being heroic and starts being harmful.
An unchecked yes is that moment turned into a business process. The willingness to absorb more is the thing being exploited, and the person exploiting it is you.
Build the first number this week. It's a spreadsheet. Then find out whether the second one was ever really the problem.
If the capacity number can't be built because effort isn't tracked, or the threshold can't be enforced because sales and delivery don't share a system, that's a different constraint — and it's worth knowing which one you have before building anything.
Get The OPERATE Report →