How to Audit What You Agreed To Last Month

Why this matters

A time study tells you where your hours went. It does not tell you why they went there. The upstream cause is almost always a set of yeses you gave in passing weeks earlier, most of them in under ten seconds, none of them written down anywhere you would find them again. Owners routinely discover that a third of a bad month was spoken for before the month started.

The second stake is quieter and worse. Commitments you drop without telling anyone are the fastest way to teach a good office lead that your word is soft, and they cost you nothing visible at the time, which is exactly why the count climbs. This audit finds them.

Step 1: Pull from where a yes actually lives, not from the calendar

A commitment is almost never booked. It is spoken at a truck door, typed into a text thread, added to the end of an email, or promised at the tail of a call after the real business was done. So the sources are: sent email for the month, sent texts, your own notes from any meeting, quotes and estimates with a promise attached, and voicemails you left.

Work backwards through one full calendar month, most recent day first. Backwards matters: forwards, you get tired and thorough at the start and skim the end, which is the recent stretch you can still act on.

If you skip this step and use the calendar instead, you will collect the commitments that were formal enough to book, which are the ones that were already going fine. The audit will come back clean and you will conclude you do not have a problem.

Step 2: Define what counts, in one line, before you start

A commitment is an obligation to a named person with an implied or stated deadline. That definition has three parts and all three have to be present.

Include: "I'll get you a number by Friday", "leave it with me", "I'll take a look", "yes we can fit that in", "let me talk to him about it", "I'll cover that one myself." That last shape is the one owners forget to count, because the person they made it to was themselves and the customer only heard "we'll handle it."

Exclude: things already inside a defined process (a quote you owe because a quote is always owed), pure information ("the part number is on the invoice"), and anything with no person on the other end. Those are work, not commitments, and mixing them in inflates the count until the audit means nothing.

If you skip this step, the definition drifts halfway through the month and your two halves are not comparable. Write it at the top of the page and do not amend it mid-audit.

Step 3: Classify each one into five buckets

Five is enough to see a pattern and few enough to code quickly. Use these, and put each commitment in exactly one:

• Customer follow-up. Something you owe a paying customer that is not the job itself.
• Internal answer. A decision or an approval someone on your team is waiting on.
• Outside favor. Anything for someone who is not a customer, a supplier or an employee.
• Supplier or admin. Paperwork, renewals, callbacks to vendors, licensing, insurance.
• I'll do it myself. Work that belongs to a role you already have someone in, that you took back.

The fifth bucket is the diagnostic one. It should be the smallest by count in a healthy month, and if it is not, the rest of the audit is mostly confirming detail.

Step 4: Price each one in hours, including the hours it costs other people

Three numbers per commitment, and estimate them fast, to the nearest half hour: your own hours, hours it pulls from someone else, and the check-back tax, which is the time you spend being asked about it and asking about it before it is done. The check-back tax is real and it is invisible in every other measurement. A ten-minute promise that generates four "any word on that?" exchanges cost more in interruption than in work.

Keep the three separate and do not add them into one figure. They are different currencies: your hours come out of the only supply that constrains the shop, other people's hours are a scheduling question, and check-back time is fragmentation rather than volume. Summing them produces a big number that tells you nothing about which lever to pull.

If you skip the pricing step, everything looks equally cheap, because at the moment of agreeing everything felt equally cheap. That feeling is what the audit exists to correct.

Step 5: Mark the honest outcome

Five outcomes, no partial credit: delivered on time, delivered late, renegotiated (you went back and changed the terms with the other person's agreement), still open, and silently dropped.

Silently dropped means the deadline passed, the thing did not happen, and you never told the person. This is the number that matters and it is the number you will be tempted to soften into "still open." The test is simple: if the other person were asked today what the status is, would they say what you would say? If not, it was dropped, not open.

Step 6: Compute three numbers, not thirty

• Commitment load. Total of your own promised hours for the month, against the hours you actually had uncommitted. Anything over about 80% of your uncommitted hours means arithmetic guaranteed a default before behaviour got a vote.
• Silent-default rate. Silently dropped, divided by the total count of commitments in the month. Under 5% of the month's commitments is normal noise. Over 10% is a standing credibility cost.
• Class concentration. Which bucket holds the largest share of your promised HOURS, which is usually not the bucket with the largest share of the COUNT.

Step 7: Convert the finding into exactly one intake rule

One. An audit that produces five new rules produces zero enforced rules. Pick the bucket carrying the most hours and write a rule with a trigger, an action and a threshold, in one sentence, and put it where you will hit it at the moment of being asked rather than at the moment of reviewing.

A worked audit: one owner, one month

He pulls sent email, texts and his notebook for a month and finds 34 commitments under the Step 2 definition.

By class: 11 customer follow-ups, 8 internal answers, 6 outside favors, 5 supplier and admin, 4 in the "I'll do it myself" bucket. That last bucket is 4 of 34, about 12% by count, which looks reassuring for roughly four seconds.

By his own promised hours: 47 hours across the month. The 4 "I'll do it myself" items account for 18 of those 47 hours, so 12% of the commitments by count carried 38% of the promised hours. The 11 customer follow-ups, the biggest bucket by count, carried 9 hours between them, about 19% of the promised hours.

Commitment load: he runs calls three days a week and counts about 12 uncommitted hours in a typical week, so roughly 48 uncommitted hours in the month. Promising 47 of 48 available hours is 98% of his slack. The rule above puts the ceiling at about 80%, so about 38 hours. He was 9 hours past the ceiling before a single truck broke down. Nothing about his discipline was going to fix that month.

Outcomes: 19 delivered on time, 6 delivered late, 5 renegotiated, 4 silently dropped. That is 34, and the silent-default rate is 4 of 34, about 12% of the month's commitments, which is over the 10% line. Of those 4 drops, 3 were outside favors, so that bucket dropped 3 of its 6, half of what he agreed to there.

The class concentration says the hours problem is the "I'll do it myself" bucket at 38% of promised hours. The silent-default rate says the credibility problem is outside favors at 3 dropped of 6. Two findings, and Step 7 allows one rule, so which?

Take the hours. The favor drops are embarrassing and they cost goodwill outside the shop, but 3 dropped favors did not create the 98% load - the 18 hours did. Fixing the hours also fixes some of the favor drops, because the drops were downstream of having no room. Fixing the favors first leaves the load untouched and simply relocates the defaults to a different bucket.

His rule: any task that belongs to a role I have already filled, and that I estimate at over 2 hours of my own time, gets a named owner other than me or an explicit decline before I answer. Trigger, threshold, action, one sentence. He writes it on the inside cover of the notebook he carries to jobs, not in a planning document, because the yes happens at a truck door and that is where the rule has to be readable.

Reading the second month against the first

Re-run the audit the following month, same definition, same five buckets. Expect the numbers to move in this order, and treat a different order as a sign the rule is not actually being applied at the moment of the ask.

Count moves first. The rule fires at intake, so the total commitment count and the hours in the targeted bucket drop within the first two weeks. In the worked case, the useful check is whether the "I'll do it myself" bucket fell below about 9 hours, which is half its prior 18.

Silent defaults move second, and only if load actually fell. A dropping default rate with flat promised hours usually means you got lucky with the month rather than that anything changed.

The check-back tax moves last, because commitments made before the rule existed keep generating check-backs for weeks after intake tightens. If total promised hours fell but your days still feel exactly as chopped up, you are living in the tail of the previous month and the honest read is at week six, not week two.

One caution on the second audit: it is tempting to widen the definition once the numbers improve, because it feels rigorous. Do not. Change the definition and you have two months of data measuring different things, and the only thing you will be able to say is that something moved.

Two failures this audit is specifically prone to

It turns into a performance review of yourself. The moment the audit becomes about whether you are a person who keeps their word, it stops producing usable numbers, because the coding goes soft to protect the finding. Code the outcomes before you read the totals, and do not read the totals until every row is coded.

The month you pick is the month you remember. Choosing a "representative" month is choosing the finding. Take the last complete month, whatever it was, and if it was genuinely unusual, note the reason at the top of the page and run the audit anyway - a distorted month with the distortion labelled is worth more than a curated one.

References

• U.S. Small Business Administration (SBA), small business owner time and workload management
• Trade-standard practice, owner commitment tracking in small service shops
• See related: How to Run a Time Study on Yourself, The Commitment You Should Not Have Made, How to Get Out of a Commitment Gracefully, How to Say No to Work That Does Not Fit, The Check-Back Cadence for Delegated Work