What an hour of the owner's time is actually worth

The usual calculation, revenue divided by hours worked, is easy to run and wrong in a way that quietly costs real money.

Every founder eventually runs the same calculation. Take annual revenue, or profit if you are being careful, divide by the hours you work, and you get an hourly number. Twelve hundred thousand dollars over three thousand hours is four hundred dollars an hour. The conclusion follows immediately: never personally do anything worth less than four hundred dollars an hour.

It is a satisfying number and it is close to useless. Three assumptions are buried inside it, and all three are false in a small company.

The three things that calculation assumes

That the freed hour converts. The four hundred dollar figure is an average of every hour you worked, including the hours that produced nothing.

An hour you free up is only worth four hundred dollars if something waiting behind it produces four hundred dollars, and if you actually move into that thing rather than into more of what you were already doing. Most freed hours get absorbed by the same category of work they came from.

That is not a discipline failure. It is what happens when the queue behind you is full of the same kind of item.

That hours are interchangeable. Nine in the morning on Tuesday and nine at night on Thursday are not the same instrument, and averaging them hides the entire problem. The nine at night hour is not competing with a partnership call.

It is competing with sleep and with the people in the next room, and that comparison is real but it is not denominated in dollars.

That the price of a task is set by who does it. It is not. It is set by what waited while it was being done. This is the part the standard calculation cannot see at all, and it is usually the largest number in the whole picture.

Three separate prices for the same hour

An owner's hour has at least three prices, and they answer different questions. Confusing them is where the reasoning goes wrong.

The replacement price

What it costs to have somebody else do this competently. Not their wage. The loaded cost, which for a US hire includes employer taxes, benefits, equipment, recruiting and the eventual cost of ending it.

The true cost of a hire calculator runs that arithmetic in about a minute, and the loaded figure is usually a long way above the salary anyone was planning to post.

There is a second component of the replacement price that people consistently forget: the documentation cost. The first time you hand a task over, you pay for writing down how it works, and writing a handover typically takes two to four times as long as performing the task once.

This is why delegation looks like a bad trade in week one and an obvious one by week six, and why owners who evaluate delegation after a fortnight keep concluding, wrongly, that it is faster to do it themselves. It is faster to do it themselves, once. It is not faster the fortieth time.

The alternative price

The value of the best other use of that specific hour. This is where honesty is more useful than optimism. For a lot of the hours that get eaten by operations, the alternative price is close to zero in cash terms.

An owner correcting listings at half past nine at night is not forgoing a revenue producing activity, because nobody is available to sell to. The alternative price of that hour is not commercial at all. It is measured in things this article is not qualified to price.

Stating that plainly matters, because the standard four hundred dollars an hour argument sneaks a false urgency into evenings and weekends. Those hours are not expensive because of what they could have earned. They are expensive for a different reason entirely.

The queue price

The cost of what waited because you were busy. This is the number nobody computes, and it is nearly always the biggest one.

The owner is a single server, and queues do not fail gently

This part is arithmetic, not metaphor, and you can check it rather than trust us.

In the simplest model of a queue with one server and variable arrivals, the average time an item spends waiting is proportional to u divided by (1 minus u), where u is how heavily loaded the server is. Put numbers in it:

• At 50 percent loaded, the average wait is about 1 time the length of the task itself.
• At 80 percent, about 4 times.
• At 90 percent, about 9 times.
• At 95 percent, about 19 times.

The curve does not bend gently. It goes vertical. And a founder who is the only person permitted to approve, decide, price or answer is exactly this system: one server, variable arrivals, no overflow capacity.

This explains something owners describe constantly and rarely name. The business did not get slower because the work got harder. It got slower because owner load moved from about 75 percent to about 95 percent, and everything requiring the owner now waits many times longer than the decision itself takes to make.

A worked example

Nothing here is a case study. It is arithmetic to rerun with your own numbers.

A supplier sends a quote that needs a yes or a no. The decision itself takes four minutes: read it, compare it to the last one, reply.

At 95 percent owner load, that four minute item waits roughly 76 minutes of working time in the queue. In practice, because the owner's day is fragmented across a store, an inbox, a phone and a family, 76 minutes of working queue time lands a day and a half later on the calendar.

The supplier's price window closes. The reorder goes on the later container. Six weeks of stock timing shifts, and a season's worth of availability moves with it.

The decision cost four minutes. The wait cost the season. Nothing in that chain shows up as an hourly rate problem, which is why the hourly rate framing cannot find it.

So what should actually be given away first

The rule that follows is not delegate anything below your hourly rate. It is:

Give away whatever consumes the most owner capacity per unit of judgement it requires.

That gives you a ranking you can compute on one page. For each recurring task, write three things:

1. Hours per week it consumes. Measured, not estimated.

2.

How badly it fragments the day. A task that arrives in eleven pieces costs more than a task of the same length that arrives in one, because of what the interruption research shows: in the CHI 2008 study, interrupted people finished on time and paid in stress, frustration and effort instead (The Cost of Interrupted Work).

Fragmentation is not free just because the output still arrives.

3. How much judgement it genuinely needs. Be strict. Most tasks that feel like judgement are actually a rule that has never been written down, which makes them a documentation job, not a decision.

Multiply hours by fragmentation, divide by judgement, sort descending. The top of that list is what to hand over first, and it is frequently not the task with the highest hour count.

Meetings, and the number nobody should quote

Meeting arithmetic is the simplest arithmetic in this whole article. Attendees multiplied by duration. A weekly one hour meeting with five people is 260 person hours a year. Whether that is worth it is a real question, and multiplying at least makes it askable.

One figure worth refusing is the widely circulated cost of meetings to American business per year. Every trail behind those numbers ends at a vendor survey with an undisclosed panel or a calculator built by a company selling meeting software.

There is no trustworthy public number for this, so inventing confidence here would be worse than leaving the gap. The multiplication above needs no source at all. It is five people times one hour times fifty two weeks, and you can do it in your head.

A meeting cost tool is one of the things currently being built for the free library. It will do the same multiplication and add the fragmentation cost, and it will run in your browser like the rest of them, with no signup.

The number to put in the founder score

The Founder Score asks what you would pay someone per hour to replace your operational hours, and it asks specifically for loaded cost rather than take home pay.

That is the replacement price from earlier in this article, and it is the only one of the three prices that has a defensible market number attached to it. The queue price is yours to compute, because only you know what waited.

What to do this week

1. Estimate your own load, roughly. Of your working hours, what share is already committed before the week starts? If the answer is above 90 percent, the queue arithmetic above is the dominant fact in your business and everything else in this list is secondary to it.

2. Find one decision that is currently waiting on you and has been for more than three days. Write down what it costs per day of delay. Not precisely. To the nearest hundred dollars is enough to make the point.

3. Take the three recurring tasks that eat the most hours and score each on the three columns: hours, fragmentation, judgement. Sort them. The winner is your first handover, not the one that annoys you most.

4. Write down the rule for the top item so it stops being a decision. A number, a threshold, a named person, a reply window. If you cannot write the rule, that task genuinely needs you and it belongs in a different column.

5. Count one recurring meeting. Attendees times hours times fifty two. Then decide whether it earns that, or whether the same information fits on one page once a week.