DEV Community

Serguey Asael Shinder
Serguey Asael Shinder

Posted on

You Reviewed Four Excellent Pull Requests in a Row and Were Hard on the Fifth

Friday afternoon.

Five pull requests
waiting for you.

The first four were excellent.

Small.

Tested.

Clear descriptions.

You approved them
with a smile.

The fifth was fine.

A bit long.

One test missing.

You left eleven comments
and asked for changes.

On Monday
you read it again
and could not find
what had bothered you.

Here is the mechanism.

You did not judge
the fifth change
on its own.

You judged it
against the four before it.

Next to excellent,
fine looked weak.

Uri Simonsohn and Francesca Gino
studied about nine thousand
MBA admission interviews
across a decade.

When an interviewer
had already scored
several strong applicants that day,
the next applicant
tended to get a lower score.

Same applicant quality.

Lower mark.

The applicants never knew.

They only knew
the decision.

Their reading:
interviewers behaved
as if each day
should come out balanced.

A few high scores
used up the day's supply,
so somebody had to come lower.

They called it
narrow bracketing.

Judging within the day,
not against everyone.

Code review
is built for it.

A queue.

One reviewer.

One sitting.

Each change scored
against the last few
you happened to see.

The person whose work
lands after a run of great ones
pays for it.

The person whose work
lands after a mess
gets waved through.

Neither of them chose the order.

Neither did you.

The tool sorted them
by the time they were opened.

So take the order
out of the judgement.

Before you start,
write down what "ready to merge"
means for this codebase.

Review against that,
not against the previous tab.

When you notice
you are harsher than usual,
stop
and ask
what exactly is wrong
with this change,
on its own.

If you cannot name it,
it is the queue talking.

Read the last one
again on Monday.

Some comments
will not survive
the weekend.

– Serguey Asael Shinder

Top comments (0)