Walk to wherever your unfinished work actually sits and count it by stage. How many quotes are written but not sent. How many jobs are sold but not scheduled. How many are finished but not invoiced. Three numbers, five minutes, nothing to install.
Whichever number is biggest is where your business is stuck right now. It is probably not the stage your customers complain about, and it is almost certainly not the stage you were about to spend money on.
What “we need to scale” usually buys
When demand climbs, the reflex is to add capacity broadly. Another pair of hands. A better tool. A bigger marketing push. A weekend spent writing procedures. Each of those is defensible on its own. Bought together, on the theory that the business is generally stretched, they mostly buy nothing.
The reason is structural rather than a matter of discipline. The Theory of Constraints Institute, which teaches the method Eliyahu Goldratt introduced in his 1984 book The Goal, sets it out through a chain: “every chain will have ONE (and only one) weakest link. Strengthening the other links can never increase the overall strength of the chain, because they are not the weakest.” Then it goes one step further, and the further step is the expensive one: “strengthening a non-weakest link will probably DECREASE the overall strength of the chain due to the added weight.” (tocinstitute.org)
The worked example on that same page is one small businesses live through regularly: “if sales orders are the constraint, optimizing production output can result in overproduction.” That overproduction becomes inventory, the inventory becomes carrying cost, the carrying cost becomes a cash squeeze, the squeeze delays payments to vendors, and the vendor problems eventually damage the sales the whole effort was meant to grow. (tocinstitute.org)
Out of manufacturing language: your business has exactly one stage setting its output today. Money spent on any other stage buys half-finished work, and half-finished work costs something to hold.
Why the loudest complaint points at the wrong stage
This is the part worth slowing down on, because it is what makes counting beat intuition.
Pain is felt downstream of a constraint, never at it. Work banks up in front of the slow stage, which starves every stage after it. Customers only ever experience the starved end: the late delivery, the quote that never came, the invoice that arrived three weeks after the job was done. So the complaint reliably describes the last stage in the chain, which is the stage with the least to do.
Take the shape this commonly takes in a two-van service company, and treat it as a shape rather than a reported case, because that is what it is. Customers complain about scheduling. Scheduling is visibly chaotic. But if every job has to be priced by the owner before it can be scheduled, and the owner prices jobs in the evening after a full day on site, then pricing sets the pace of the entire company and scheduling is merely the first place a customer can see the shortage. Hire a scheduler and you have added salary to a stage that was already waiting.
The counting exercise catches this because a queue cannot lie about where it is. The complaint tells you where the damage surfaces. The pile tells you where it starts.
Five moves, in this order
1. Count what is waiting, by stage, on one ordinary day. Write your stages across a page in the order work moves through them, then count the items sitting in each one right now. Not what came in this month, not what you completed, just what is parked mid-process at this moment. Pick a normal day rather than your worst one. The output is a row of numbers, and you keep it.
2. Confirm the candidate before you believe it. A big pile can just mean a busy week, so run two checks. First, look at the stage immediately after your candidate and ask whether it is ever sitting idle waiting for work to arrive. A genuine constraint starves whatever follows it. Second, count the same stages again a week later. A real constraint’s queue holds or grows, while a temporary spike drains on its own. If the next stage goes idle and the pile does not shrink, you have found it.
3. Relieve it before you buy anything. Three things are free here and they are usually enough to start. Take non-constraint work off the constraint: if the owner is the pricing stage, the owner should not also be chasing materials during pricing hours. Stop letting it idle: give the constraint a ready queue of prepared inputs so it never sits waiting on missing information. Then move the approvals off it by writing the decision rule down, which is the move that actually compounds, and it is worth writing rules concrete enough to be wrong rather than a page of values nobody could argue with, a distinction we worked through in a leadership philosophy you could lose to. Capacity you did not know you had comes from the constraint doing only its own work.
4. Then buy or hire, but only at the constraint. Once the free moves are exhausted, spending is the right answer and it should be pointed at one place. Hiring at the constraint is a legitimate and often the correct move, and hiring anywhere else is the mistake this whole method exists to prevent. The same test applies to software: a tool that speeds up a stage that is already waiting has bought you a faster way to wait. One stage gets the money.
5. Count again in a month, because it has moved. This is the step almost every scaling plan omits. Relieve a constraint and it stops being the constraint, and something else immediately becomes one, because a system with no limiting factor would have infinite output and no business has that. A ranked list of ten fixes written in January is therefore wrong from item two onward. What you want is not a plan but a loop, and the loop is only three questions: where is the line, what frees it, what happened.
Where software helps, and what it costs
You do not need a tool for move one. A page and a pen are genuinely better for the first count, because the friction of writing the numbers by hand is what makes you look at the real work instead of at a dashboard.
Where software earns its place is move five, the repeat count, because doing it by hand every month is exactly the kind of chore that quietly stops happening. Anything that shows work as items sitting in named columns will do it, and the cheap tiers are sufficient, since the feature you need is the ability to see how many cards are parked in each list.
Trello lists its Free plan at $0 and free for up to 10 collaborators per Workspace, with Standard at $5 per user per month billed annually or $6 billed monthly, and Premium at $10 billed annually or $12.50 monthly. For counting queues, the free tier is not a trial, it is the product. Prices read live from trello.com/en/pricing today.
ClickUp runs a Free Forever tier that includes Kanban boards, with Unlimited at $7 per user per month billed yearly or $10 billed monthly, and Business at $12 yearly or $19 monthly. Prices read live from clickup.com/pricing today.
Asana prices Personal at $0 forever but caps it at 2 users, which is the detail that matters for a small team, with Starter at $10.99 per user per month billed annually or $13.49 monthly, and Advanced at $24.99 annually or $30.49 monthly. Prices read live from asana.com/pricing today.
If your stages already live somewhere, count them there. Adding a system to measure the system is its own small constraint.
Where an assistant genuinely helps
The counting is judgment and you should do it yourself. The tedious half is different: if your jobs pass through a spreadsheet or an order export, working out how long items actually sat at each stage means subtracting timestamps a few hundred times, and that is work where a machine does not get bored and a person does. Export the data, give an assistant the column names, and ask it for the median time each stage held an item and the count still open per stage. Read the output against your hand count from move one. Where the two disagree, trust the hand count and find out why the data does not match it, because the mismatch is usually a stage nobody is recording.
If you want a structured way to think through the wider planning, the Operational Scaling Plan prompt at BusinessPrompter.com is filed under Operations and Systems and summarized as planning how to scale operations as the business grows. One caution if you click through: the descriptive panels below that prompt’s title currently describe a different exercise about brand awareness and marketing channels, so take the title and the summary as your guide rather than the blocks underneath them.
What this does not fix
Finding your constraint tells you where to spend. It does not tell you whether the underlying business works.
Some stages are slow because the price is wrong, not because capacity is short, and no amount of flow analysis surfaces that. Some are slow for reasons entirely outside the process: a carrier’s published rules, a supplier’s lead time, a license that takes a fixed number of weeks. We went through one such case in detail, where the biggest available lever turned out to be a published threshold rather than anything happening inside the business, in logistics optimization starts at 1,728 cubic inches.
There is also a floor under how much you should relieve. A constraint running at full tilt with no slack is fragile, and a single sick day takes the whole company’s output with it. The goal is a constraint that is busy and known, not one that is maximized.
And the durability problem is real. A fix that lives in one person’s memory lasts until that person takes time off, which is why the third move ends in writing the rule down rather than remembering it, and why the document should be short enough that someone will actually revisit it, a case we made in a process documentation template cut to five fields.
Our read, and it is a read
A judgment rather than a reported finding, so treat it as one. We think the reason the constraint in small businesses is so often the owner is that owner-held work does not look like a queue. It looks like evenings.
Work waiting in a stage shows up as a pile someone can point at. Work waiting on one person’s attention shows up as a full calendar, a late night, and a reputation for being busy, and none of those read as an operational problem. They read as commitment. So the one stage most likely to be setting the pace of the entire company is also the only stage nobody counts, because counting it would mean writing down that the business is waiting on its founder.
That is also the version of this we find most worth doing, and it is where scaling stops being a spending question. Relieving an owner-shaped constraint almost always means somebody else on the team is now allowed to decide something they were previously only allowed to recommend. The capacity was in the building the whole time, held behind an approval step. A small team that grows into more responsibility is a considerably better outcome than the same team working faster under the same permissions, and it is the outcome that survives the founder’s next vacation.
The steady-state context is worth holding here too. Of the private sector establishments that opened in the year ended March 2020, 51.4 percent were still operating five years later, and in each of years two through five, between 88.0 and 90.0 percent of the previous year’s survivors made it through the following year, per the Bureau of Labor Statistics establishment survival table (bls.gov). Surviving the first year does not move a business onto safe ground. That is an argument for knowing which single thing currently limits you, not for buying capacity in every direction at once.
Frequently Asked Questions
How do I find the bottleneck in my business?
Count the unfinished work sitting in each stage of your process on one ordinary day, then confirm the largest pile with two checks: whether the stage immediately after it ever sits idle waiting for work, and whether the pile is still there a week later. A genuine constraint starves the stage that follows it and does not drain on its own. The stage customers complain about is usually downstream of the real one, because shortages become visible to a customer at the end of the chain rather than at the point where they start.
Should I hire before or after I find my constraint?
After, and then hire at it. Hiring at the constraint is frequently the correct move once the free steps are used up. Hiring at any other stage adds salary to a stage that was already waiting for work to reach it, which raises cost without raising output. The Theory of Constraints Institute makes the general form of this point through its chain analogy: “Strengthening the other links can never increase the overall strength of the chain, because they are not the weakest.” (tocinstitute.org)
How often should I redo this?
Monthly at first, then quarterly once the numbers stop moving much. Relieving a constraint means it stops being the constraint and another stage takes over as the limiting factor, so a fixed ranked list of improvements goes stale almost immediately. The repeat count is the part that makes the method work rather than an optional extra.
Do I need software to do this?
No for the first count, which is better done on paper. For the repeat counts, any tool that displays work as items in named columns is enough, and the free tiers cover it: Trello is $0 and free for up to 10 collaborators per Workspace (trello.com/en/pricing) and ClickUp’s Free Forever tier includes Kanban boards (clickup.com/pricing). Asana’s free Personal plan is capped at 2 users, which rules it out for most teams wanting shared visibility (asana.com/pricing).
Before you approve the next hire
The whole method reduces to one sentence: find the single stage where work waits, take everything that is not its job away from it, and count again next month because it will have moved.
The uncomfortable half of that sentence is the taking away. Every constraint in a small business is being protected by somebody’s habit, and often the habit belongs to the person reading this. The count takes five minutes. Acting on it means giving up a decision you are currently making, which is why most owners do the count, recognize the answer immediately, and buy something else instead.
