How to Review Work You Did Not Watch

Forty steps and eleven minutes

You come back to your desk. A job you set off before lunch has finished. There are forty steps in the log and a finished document underneath them. You have eleven minutes before your next meeting.

Reading all forty steps is not a plan. It takes longer than doing the job yourself. Skipping to the output and calling it good is not a plan either, because that is how drifted work gets shipped.

There is a middle path, and it is the skill this course has been building towards. You do not review a loop by re-running it. You review it at four specific places, and those fit in the time you have.

Check one: the finish against the brief

Always first, and never skip it. Open what you originally asked for. Read it. Then read the finished work and ask one question: is this an answer to that?

Not is it good. Not is it useful. Is it an answer to the thing I asked. This is the drift check, and it takes about a minute. Do it before you get impressed by the output, because after that it is much harder to see.

If the brief had a list in it, check the list. Five questions asked means five answers found. Four excellent answers and one quietly missing is a very common shape, and it is invisible unless you count.

Check two: the turning points

Now the steps, but not all of them. You are looking for the handful where the loop changed its mind.

Skim the log for the places where something surprising happened: an error, a blocked action, a decision between two options, a moment where the plan changed. In a forty step run there will usually be a few. Those are the joints, and everything downstream of a bad joint is also bad.

The rest of the steps are mostly repetition of an approach that was already chosen. If the joint was sound, they are probably fine. If it was not, no amount of reading them will help.

STEPS 1-6    opened files, listed contents          routine, skip
STEP 7       source unavailable, used a cached copy  TURNING POINT
STEPS 8-14   extracted figures                       routine, skip
STEP 15      two figures disagreed, picked one       TURNING POINT
STEPS 16-23  built the comparison                    routine, skip
STEP 24      widened to include a fourth product     TURNING POINT
STEPS 25-38  wrote and revised sections              routine, skip
STEP 39      could not verify one claim, left it in  TURNING POINT
STEP 40      finished
Four steps to read properly. Thirty-six to skim.

Check three: verify what the conclusion rests on

Not every fact. The load-bearing ones.

Take the recommendation at the end and ask what would have to be true for it to hold. There are usually two or three things. A price, a date, a figure, a quotation. Check those against the actual source, yourself, in the original document or on the actual page.

Everything else can wait. If the load-bearing claims survive, the work is standing on something. If one of them does not, the polish above it is irrelevant.

CONCLUSION   Switch supplier B to supplier C. Saves money and ships faster.
WHAT MUST BE TRUE

Supplier C's price is genuinely lower           verify at source
Supplier C's lead time is genuinely shorter     verify at source
C can handle our volume                         verify at source
WHAT DOES NOT NEED CHECKING TODAY
the tone of the summary
the order of the sections
the wording of the recommendation
If claim 1 is wrong, nothing else in the document matters.

Check four: hunt for the silences

The last check, and the one people never think of. Ask what is not there.

Loops are much better at reporting what they did than what they could not do. A source that would not open, a figure that could not be confirmed, a question in the brief that turned out to be unanswerable: these get quietly smoothed over more often than they get flagged.

The fastest way to find them is to ask directly. It costs one message and it is often the most useful message in the whole exchange.

Before I review this, answer these four questions plainly.

Which parts of my brief did you not fully complete, and why.
What could you not verify, and what did you do instead.
Where did you have to choose between two options,
and what was the other option.
If one thing in this is wrong, which is it most likely to be.
Short answers. Do not defend the work, just tell me.

The checklist

This is the thing to keep. Copy it into your notes app, print it, put it in a document. It lives with you, not with this site, and nothing here is stored or remembered on your behalf.

LOOP REVIEW CHECKLIST


BRIEF        Re-read what I asked for. Is this an answer to it.
If my brief had a list, count the answers.


TURNING      Skim the steps. Find where it changed direction,
POINTS       hit an error, or chose between options.
Read only those properly.


FACTS        What must be true for the conclusion to hold.
Verify those two or three at the source, myself.


SILENCES     Ask: what did you not finish, not verify,
and what were you least sure about.


DECIDE       Accept, send back with one specific fix,
or do this part myself.

If I cannot answer 1 and 3, I have not reviewed it.
The rest is optional. Those two are not.

Deciding what to do next

Three outcomes, and only three. Accept it. Send it back with one specific correction, naming the step where things went wrong. Or take that part back and do it yourself.

The middle option is worth practising. Pointing at the exact turning point beats asking for a general improvement, because a general request restarts the loop and a specific one repairs it.

You have finished the course

You know what a loop is, what iteration buys, the three ways it fails, how to set stopping rules and guardrails, what each pass costs, and how to review work you did not watch. That last one is the practical skill, and passing this quiz earns you the Loop Reader badge.

Final Quiz