The payment provider
failed about one checkout
in two hundred.
Everyone knew the number.
There was a dashboard for it.
There was a retry for it.
There was a support agent
who knew the workaround
by heart.
A newer provider
was cheaper,
with a better API.
The meeting went
the way these meetings go.
"At least with the current one
we know what we are dealing with."
Heads nodded.
Nobody asked
what the new provider's
failure rate was.
Nobody knew.
That was the whole argument.
Here is the mechanism.
In 1961
Daniel Ellsberg
described two urns.
The first holds
fifty red balls
and fifty black.
The second holds
a hundred balls,
red and black,
in a mix nobody tells you.
Bet on red.
Most people
choose the first urn.
Now bet on black.
Most people
choose the first urn again.
But no belief
about the second urn
makes both choices sensible.
If you think it has fewer reds,
it must have more blacks.
The preference
is not about the odds.
It is about not knowing them.
Ambiguity aversion.
We would rather face
a known bad number
than an unknown one,
even when the unknown one
is probably better.
In engineering
it has a favourite costume.
The incumbent.
The old library.
The flaky test you know how to rerun.
The server you know how to restart.
Known pain
feels like control.
Unknown pain
feels like risk.
Usually it is only
unmeasured.
The fix is not courage.
It is measurement.
Ask what it would take
to turn the second urn
into the first.
A week of traffic
at five percent.
A reference call
with two customers
who left the old one.
A test account
and a script
that fails on purpose.
Then decide.
Sometimes the incumbent
still wins.
That is fine.
It should win
on its number,
not on the absence
of the other one's.
"We know what we are dealing with"
is a sentence about you,
not about the vendor.
– Serguey Asael Shinder
Top comments (0)