You might spend more time fetching a car clock from the scrapyard than
if you just sat there with a watch and an ammeter. The mfgr
recommendation is probably to protect you from the surge at start up,
which might surprise you, if you expect say 10 amp draw, but have a one
second surge of 30 amps.

After noting a single cycle of start, run, stop, you then have the
onerous task of sitting beside the fridge all day with a beer or two,
counting, averaging the cycles per hour.

From a warm case of beer start, to keep score, or for all day, opening
the door for 20 seconds, closing it every hour, to work up a "typical"
duty cycle.

Alternatively, charge the battery check the specific gravity, let it
run for say 12 hours, then measure the sg of the battery, and
interpolate the number of amp hours from your battery specs.

How critical is all of this?

Your time is say, 10 bucks an hour. The test equipment to do this
properly starts out ridiculously expensive, then proceeds toward the
incredulous, depending on the degree of accuracy desired.

A laptop data port, some software, an interface, or multimeter with
data output would do for centibucks what you could accomplish by rule
of thumb for a little effort.

Actually, I wonder if a digital audio recorder like windat might record
DC voltages across a current shunt in the DC supply. You would need to
test and calibrate the data.

Terry K