The Pellucid Perspective - May 2024 - 8

TECHNOLOGY
APPLICATIONS
skeptical) to solve the classification problem. This took us all of
about 10-15 minutes for our first pass.
ChatGPT then went through our instructions and pointed
out errors and inconsistencies between our instructions and the
data (this was kind of embarrassing, some were very basic, dumb
mistakes on my part). For example, in the written instructions
we told it to " look at the Greens Fee value " and it indicated there
was no value in the data for Greens Fee value (it was right, in
the data file that column was actually labeled Highest WE GF,
easy fix in the instructions to align). After we corrected all the
logic/syntax errors, it then went through a more interesting pass
to ask questions about decision rules where there could be outliers
to the decision rules. Really? Yep, like " What do I do when
there are <3 facilities in a DMA? " (quick answer, there are no
DMAs in the US with <3 facilities, fun trivia fact) and " What if
all facilities in the DMA have the same price point? " For some
of those we clarified the language and others we chose to just
ignore because I knew there wouldn't be any incidence of the
scenario. This step took us another 15-20 minutes but the combination
of my son knowing what he was doing and me knowing
the data intimately (after 20+ years of toiling with this manually)
probably helped considerably. Then came the first pass and...it
was perfect!
No, just kidding. Here were the couple of diagnostic passes
we had to make to figure out the couple of errors we saw (full
*
In the first pass the overwhelming majority of facilities
were classified Public-Value so we basically had
Private, Learning & Practice and Public-Value. We
determined that we made the dumb, basic mistake of
not telling ChatGPT that it needed to exclude those
facilities before it started the " price-binning " exercise.
What happened was almost all the Privates were being
considered as Premium and nearly all the Learning &
Practice were being considered Price hence, everything
that was left was largely relegated to Value
* The second pass corrected the Public-Regulation
segmentation but, for some reason, it was inverting
the Premium and Price designations. It took us a few
minutes to realize that we'd never really defined what
was the highest vs. lowest value designation and we
had written the instruction on the classifications in the
order of Price, Value and Premium. We had then given
it the task of sorting, from high to low, the Greens Fee
pricing so it, logically, assumed that Price was first and
Premium was third. It took us more time to figure
out why this was happening (~ 5 mins) than it did to
transparency, we decided a more efficient way to design and
troubleshoot was to use the Chicago DMA so we started with
the ~325 dataset to train ChatGPT):
Partner with the True, All-In-One
Golf Management Solution
Club Caddie was designed by golf course owners and operators to help solve
the most common pains of operating any golf facility. Our team is passionate
about integrating the latest technology, industry trends and best practices to
save golf course operators time, money and resources. To Learn more about
how Club Caddie can help streamline your golf operation, email Jake Ayala at
Jake@ClubCaddie.com, or visit our website at ClubCaddie.com/demo to
custom fit your golf facility.
Online Tee Sheet
& Electronic
Reservation
System
Retail,
Restaurant, Bar
& Mobile POS
8 The Pellucid PersPecTive
Customer
Marketing & Full
Member
Management
FullSpecial Event
& Banquet
Management
Simulator,
Court & Activity
Booking
TM
Websites &
Mobile Apps
with GPS
Accounting
and Business
Intelligence
Reports
May 2024
https://clubcaddie.com/demo/
The Pellucid Perspective - May 2024

Table of Contents for the Digital Edition of The Pellucid Perspective - May 2024
