Living in SSIM City with The Token King
Three truly world-changing things happened on July 1, 1972:
- The opening ceremony took place in Reykjavík for the Cold War "Match of the Century" between chess legends Boris Spassky and Bobby Fischer.
- In preparation for the Euro, the EEC created the European Currency Snake.
- To the awestruck gasps of global aviation enthusiasts, IATA introduced the SSIM (Standard Schedules Information Manual).
Though a New York Times article describes our COO Rory as a Yemeni-born software entrepreneur, around here he is known as the "SSIM Geek." Want to know about Record Type 3s (Leg Records)? The wildly popular Record Type 4s (DEI Records)? Or—oh my gosh—the holy grail of DEI codes? Rory is the guy who knows precisely what sits in every single bit and byte of SSIM goodness.
With this in mind, we recently got into a lively discussion about code-share synchronization parameters:
"Operating disclosure is null, Service Type can be S, J, or L, Operating Airline Designator equals competitors listed, and the DEI 050 appears in positions 2 through 5..."
Excited by the prospect of summiting Mt. SSIM, I tried to put together a sequence of prompts that would determine if there were any data mismatches between the operating carrier and the marketing carrier from week to week. As every good scheduler knows, the number of records increases exponentially when you try to compare 47 weeks of operating data with multiple partners. Add to that the need to explore every moving part—flight numbers, operating times, departure and arrival terminals, meal codes—and you suddenly have a ridiculous group of records to process.
However, the prompts worked... though more in theory than in practice.
Unfortunately, Maestrow generated over 600,000 rows of data due to a profound logic error in my prompts which created a cartesian cross join. This elicited a snide, though warranted rebuke from our CTO Chris, known around the dinner table as The Token King.
Taking a step back, I re-evaluated what we were trying to accomplish: Are there any differences in the file, yes or no? I asked the system for help in producing a smarter way to look at the problem.
Maestrow devised a clever data fingerprinting technique. Instead of comparing individual days, it suggested concatenating all the relevant fields together into a single "fingerprint" based on effective date ranges. From there, it would filter and display only those occurrences where the fingerprints failed to match, wrapping it all up in an Executive Summary to highlight systemic patterns.
It wrote its own SQL in real-time to work through the current schedule. The result? It drastically reduced processing time, yielded a manageable result, and called on itself to suggest where corrections were needed—and how to prevent errors in the future.
The prompts still need a bit of refinement before they are ready for primetime, but the out-of-the-box thinking helped us power engineer a strategic solution to a notoriously difficult scheduling dilemma.
I thought this was amazing. Naturally, I called and texted Rory so he could revel in the results. To my quiet dismay, he did not respond right away. I still cannot decide if this was in any way SSIMbolic or he was too busy catching up on the latest episode of The SSIMpsons.