The Decision Bottleneck: How to Find It and What to Write Down

"I know I'm the bottleneck."

You have said it, probably with a small laugh, because admitting it first takes the sting out. At a peer forum, to your auditor, or in an exit conversation with someone too polite to say it to you directly.

And the sentence that follows is always the same. I need to let go more.

Now try a different question. Name the decisions waiting for you right now. Not categories. The actual items. The extra discount on the Coimbatore account. Whether to replace the transport vendor. Whether the second shift starts in September or October.

Most promoters get to four or five and stop. The real number, on any given Tuesday, is past thirty.

That gap is the entire problem. Every promoter in this country knows he is the decision bottleneck. Almost none know which decisions are queued behind him, in what volume, for how long, or who is waiting. The bottleneck gets discussed as a personality issue — a failure of trust, a control habit, something to be corrected by resolve. It is an inventory problem, and inventory problems have written solutions. You cannot distribute authority you have never enumerated.

This is the second rung of the structure in employees not taking ownership — authority, the ability to decide without asking you. It is where most promoter-led Indian businesses break, and the rung founders are most confident they have already fixed.

Why "I'll delegate more" fails every time

Because it is a resolution, not a mechanism.

A resolution depends on your state of mind at the moment of decision. It survives a calm week. It does not survive the week when a large customer is unhappy, a consignment is stuck, and somebody senior has handed you a proposal with an obvious error in it. On that day you take the decision back, because taking it back is faster, and because you are right — you would have decided better.

Then it is gone. Not the authority in theory; the authority in practice. Everyone saw that under pressure, decisions return to you. They now hold theirs during the calm weeks too.

There is a second reason nobody says out loud. "Let go more" has no object. Let go of what? A resolution with no named subject can never be checked, so it can never be failed, so it changes nothing. It is a feeling about yourself, not an instruction to anyone.

The founder's half of this — why the taking-back feels necessary each time — is in why founders micromanage. The team's half, how a company learns to stop deciding, is in why your team waits to be told.

The queue cost nobody counts

What makes the bottleneck expensive is not your time.

A distributor asks your regional head for an extra three percent on a ₹40 lakh annual commitment. Your policy allows him five percent; this is beyond it. He emails you. You are travelling, then you are in a plant review, then it is the weekend. On the ninth day you read it properly, think for four minutes, and approve it.

Four minutes of your time. Nine days of the company's.

In those nine days the distributor did not sit still. He asked your competitor for a quote, not because he wanted to switch but because he had stocking to plan and you had gone quiet. And he learned something durable about your firm: things here take a while, and the man who visits him cannot agree anything.

Or the vendor switch. Your operations head wants to change transport partners because damage claims are rising, and has an alternative at a comparable rate. He raises it, you ask him to compare once more, and it waits three weeks. Every damaged consignment in that period was avoidable: the decision had been made by the man closest to it and was not allowed to take effect.

A decision that takes you four minutes but sat for nine days cost the company nine days, not four minutes. Multiply that by a queue you have never counted and you have the price of the bottleneck. It appears in no P&L line. It appears as slowness, which everyone attributes to the market, the team, or the difficulty of doing business in India.

Your calendar is not the constraint. The wait time in front of your calendar is, and it is a hundred times larger.

The two-week decision log

Before you fix anything you have to see the inventory. Ten working days, no meetings.

Every time somebody asks you to decide something — in a review, on WhatsApp, in the corridor outside your cabin, at the site — write one line. Four fields:

• What was asked. In plain words, not a category.
• Who asked. The name, and the level.
• What you decided. Including "I asked for more information", which is a decision.
• How long it had been waiting before it reached you. Not since it reached you. Since the person first knew it needed a decision.

That fourth field is the one everybody skips and the only one that produces surprise. You have to ask: how long has this been sitting? People answer honestly, because it is not a trap and they have been waiting to be asked.

Two rules. No judgement — do not comment on whether something should have come to you. And no change of behaviour — decide exactly as you would have anyway. Correct the flow and you stop measuring it.

Most founders end up with forty to eighty entries in ten days. The number is not the shock. The composition is. Nobody expects the largest group to be entirely unstrategic, or to find items they believed somebody else was already deciding.

If much of your log arrives from review meetings rather than from work, that is a separate failure — see why meetings don't produce decisions.

Sorting the log into four classes

Take the log on a Saturday morning, alone, and put every line into one of four classes. That is the diagnostic, and it takes about ninety minutes.

Decisions that were never yours

These exist because of a rule set at a smaller size. A signature limit fixed when the company was doing ₹8 crore. A vendor approval process designed when there were nine vendors. A rule that all overseas courier goes through you, from the year one courier went missing.

Each was correct when made. None was revisited, because nobody has the standing to revisit a rule the promoter installed.

The fix is not to approve faster. Delete the rule, not the approval. If the ₹25,000 spend limit made sense at ₹8 crore, the equivalent at ₹60 crore is a different number, and stating it is your job.

Decisions that are yours only because the standard is unwritten

This will be your largest class, usually by a wide margin.

Whether to accept a customer's request for 45-day credit. Whether this candidate's salary ask is acceptable. Whether to take the rush order that disturbs the schedule. Whether this quality deviation can ship.

You are not deciding these because they need your judgement. You are deciding them because the standard exists only in your head, and the person asking is doing the correct thing by asking the only man who has it. That is not timidity. It is accuracy.

Every decision in this class converts to somebody else's the moment the standard is written down. Not delegated — converted. That distinction separates a handover that holds from one that returns in six weeks, taken apart in how to delegate as a founder.

Decisions that are yours because of relationship or risk

The top three accounts where the owner will only speak to you. The bank. Anything involving family in the leadership layer. A statutory exposure.

These stay with you, and there is no shame in that. A register that pretends otherwise is abandoned inside a month. The only discipline needed here is honesty about size: if this class is thirty of your eighty entries, the word "relationship" is doing work it should not be doing.

Decisions nobody should be making

The last class is the interesting one. Recurring exceptions: the same customer's dispatch expedited every month, the same quality deviation waived, the same overtime approval every fortnight for the same reason.

An exception that recurs is not an exception. It is a broken process upstream, patched by your signature. Deciding these faster makes the underlying failure permanent and invisible, because the patch keeps working. This class does not enter the register. It goes onto a list of processes to fix.

The Decision Register

The output is one page. Not a policy document, not a delegation-of-authority matrix drafted by a consultant. One page that every senior person has seen.

For each recurring decision class in your log, three things:

• The named owner. A person, not a department. "Sales" does not decide anything.
• The boundary. A rupee figure, a customer tier, a credit period in days, a named class of exception. Discounts up to 8% on tier-2 and tier-3 accounts. Spend up to ₹2 lakh inside an approved budget line.
• What happens beyond the boundary. Who it goes to, and by when.

The boundary must be numeric or categorical. Never "use judgement", never "within reason", never "as appropriate". Those phrases feel generous and are the exact mechanism by which authority returns to you, because an undefined boundary is one both sides must guess at, and the safe guess is to ask. If you cannot state the number, you have not handed the decision over. You have decided to seem like you did.

A register is also worthless unless you honour it in public the first time you disagree with an outcome. That moment arrives within about three weeks. Somebody exercises their boundary correctly and you think the call was wrong. Overturn it and you have not corrected a decision, you have withdrawn a register, and everyone knows by the afternoon. Let it stand and take the reasoning up privately afterwards, and the register becomes real. This is the same public signal that hollows out a middle layer, described in why middle management fails.

What actually changes, honestly

Three things, and one of them is unpleasant.

Decision speed improves within weeks. Things that sat for nine days now take a day, because the person who always knew the answer is allowed to say it.

Quality dips briefly. It has to. People deciding for the first time inside a written boundary will make calls you would not have made, and two or three will cost money. That is the price of the transfer, paid once, not evidence the register was wrong. The founder who cannot tolerate the dip reclaims the register within a month, usually while explaining that the principle is sound but the timing is not right.

Your calendar frees up more than you expect. Not because the decisions took long — four minutes each — but because the queue in front of them consumed follow-ups, reminders, status questions and half-conversations in corridors. That whole layer of traffic disappears.

Consider what it does for the layer under you. Gallup's 2025 India data found engagement among Indian managers fell from 39% to 30% in a single year, almost twice as fast as individual contributors. A manager with responsibility and no boundary is not a manager. He is a well-paid message-forwarding service, and he knows it.

The register's second use

There is a second use, not obvious until you have built it. The register is the only honest input to the question what does my next senior hire actually own.

Promoters who feel the bottleneck reach for a person. A COO, a business head. The logic is sound: hire someone senior, hand over the running of the firm, step back. Eighteen months later that person is expensive, underused and quietly looking, and the promoter is still deciding everything — because nobody wrote down what "running the firm" consisted of. The job was defined as a title and a reporting line, not as a set of decisions with boundaries.

With a register the role description writes itself: these classes, at these limits, transfer to this person on this date. Without one you are hiring a senior man to guess at the contents of your head. The case for and against that hire is in when to hire a COO.

Where to start

Do not build the register this week. Build the log.

Ten working days, four fields, no judgement, no behaviour change. Keep it in a notebook or one note on your phone — choosing a tool is a way of not starting. On day eleven, sort it into the four classes and count each one.

Most promoters find at that point that the bottleneck was never about trust. It was a few dozen standards that lived in one head and were never written down.

To see where decisions actually get made across the whole business — as against where your org chart says they do — the Business Pulse diagnostic maps that gap. In most promoter-led firms it is the largest single explanation for why things move slowly.