A fuel ticket is the record of one uplift: the aircraft, the flight, the gallons, the date, the destination. Part 1 of the case study showed that 98 percent of four months of live tickets reached the ledger with no human touch. That raises the obvious question: how much was there to fix?
To answer it we compared every one of Part 1's tickets, as the handler's system or the fueler wrote it, with the record it finally posted as. Same tickets, same window, same 50 stations, read from the production database on November 4, 2026. Nothing was sampled. Get the complete Part 2. We'll email it to your work address.
How we count. A wrong value means the ticket named a different aircraft, flight or destination from its posted record, had its date corrected by a person or a ruling, was re-issued by the source with materially changed values (net moved by more than 50 US gallons, or a meter or truck changed), or carried no quantity. A required field left blank means no destination, flight number or registration, or no net volume on a feed that reports one. Each ticket counts once, however many faults it had.
Not counted: re-sent copies, carriers that are not customers, a missing net on the 12 feeds whose format reports gross volume only, and a registration written without its country prefix.
Why the ticket is often not what flew
Part 1 showed that the hard part of processing a fuel ticket is not reading it but deciding what it means. Part 2 shows why. A tail one letter off from the aircraft on the stand, the airline's code where the registration should be, a flight number from the previous rotation, a date keyed before midnight for a departure after it, a destination column never filled in. Each of these posts cleanly to an airline's fuel-management system if nobody checks, and each of them is wrong.
| Fault on the ticket | What it looked like | What the pipeline does |
|---|---|---|
| Destination left blank | The export has a destination column the station doesn't fill. 2,821 tickets, almost all from two source systems. | Filled from the airline-provided schedule for the leg the ticket matches. |
| A different aircraft | A neighbouring tail in the same fleet, the airline code in the tail box, or a destination code in the registration column. 603 tickets. | The aircraft on the matched leg is adopted and the written value kept beside it. A dispute with the schedule is recorded, not silently overruled. |
| Quantity or meters re-issued | A later edition of the same day's export changed the net, the gross, a meter reading or the truck. 490 tickets. | The later edition replaces the earlier one and the change is written to the ticket's history. |
| Date keyed a day early | An overnight departure written on the day the fueling began. 133 tickets. | The fuel-day rule and the schedule leg set the date. The overnight case is the one judgement Part 1 hands to a person. |
| Wrong flight or destination | The previous rotation's flight number, or a destination that wasn't the leg's. 257 tickets. | Matched to the leg that flew, from the schedule and live flight data. |
What the faults were
Three quarters of the faults were one thing: no destination. The rest were wrong values, led by the aircraft. A ticket can carry more than one fault, so the counts add up to more than the 3,814 tickets with an error.
Month by month: a floor, not a trend
A wrong value leaves a trace whenever the posted record differs from what was written, so wrong values are visible across all four months. A blank field only leaves a trace if a copy of what the ticket wrote was kept before the gap was filled. FuelDeck™ began keeping that copy on September 15. Before it, a blank destination was filled from the schedule and nothing recorded that it had been blank.
| Ticket month | Tickets | Wrong value | Wrong or missing | Written copy kept | Visibility |
|---|---|---|---|---|---|
| July 2026 | 5,996 | 3.9% | 3.9% | 1% | Blank fields not yet visible |
| August 2026 | 6,031 | 4.8% | 4.8% | 19% | Blank fields not yet visible |
| September 2026 | 5,779 | 7.4% | 22.3% | 83% | Copy kept from September 15 |
| October 2026 | 5,356 | 8.3% | 37.4% | 100% | Full visibility |
| Beta total | 23,162 | 6.0% | 16.5% | 49% |
The climb from 3.9% in July to 37.4% in October is visibility, not tickets getting worse: the written copy exposes blank destinations the earlier months cannot show. From September 15, one ticket in three arrived with no destination (32.5% of 8,467). Applying each feed's later rate to its earlier tickets adds about 4,473 tickets and puts the likely figure for the whole window at about 35.8%. The headline stays the measured 16.5%.
Where the faults sat
The faults follow the source system more than the station. The two export formats that leave the destination column empty account for most of the headline, while fuel-management system files arrive nearly clean. Two of the 50 stations had no wrong or missing field at all, and 13 of the 43 stations with 100 tickets or more were under 2 percent.
| Source | Kind of file | Tickets | Wrong value | Wrong or missing |
|---|---|---|---|---|
| System A | Hydrant dispensing export | 12,052 | 4.0% | 11.5% |
| System B | Into-plane supplier export | 3,570 | 15.7% | 49.0% |
| System C | Into-plane supplier consumption report | 1,547 | 10.0% | 19.8% |
| System D | Fuel-management system export | 1,427 | 2.6% | 2.6% |
| System E | Station report by email | 1,377 | 3.3% | 10.2% |
| System F | Fuel-management system report (spreadsheet) | 757 | 0.3% | 0.3% |
What became of them
None of the 3,814 reached an airline as written. The destination was filled from the airline-provided schedule, the aircraft and flight were matched to the leg that flew, the fuel-day rule set the date and the later edition replaced the earlier one. Thirty were posted only after a review flag and seven were still open when measured. The 348 human decisions Part 1 counts are among them.
Reading the two parts together
The two figures describe the same tickets from opposite ends. Part 1's 98.2 percent is the share that reached the airline's system with the right aircraft, flight, day, destination and quantity without anyone stepping in. Part 2's 16.5 percent is the share that would not have, had nobody checked. Between them sits the work a person used to do on every ticket.
| Measure | Part 1 · delivered | Part 2 · given |
|---|---|---|
| Headline | 98.2% no human touch | 16.5% wrong or missing |
| Tickets | 23,207 | 23,162 matched |
| Behind it | 348 needed a person | 1,393 carried a wrong value |
| What it measures | Who had to step in | What there was to fix |
An operation that measures only the first figure will believe its tickets are clean. One that measures only the second will believe it needs more people. Measured together, on the same tickets, they say that the checking is the work, and that the work can be done by rule.
What these figures don't say
They count what was wrong on the ticket as written, not what it would have cost. They come from one operation, the managed fuel-data service Logicraft runs for its airline customers, and stations and source systems are anonymised. Every definition is stated so that an airline, a handler or an airport can apply it to its own tickets, and that is a conversation worth having.
The complete Part 2
Every figure, every definition
The full report adds the three tiers in detail, all 13 source systems, the twelve largest stations, the month-by-month charts and the method behind each number.
Request the complete Part 2 PDF.
Get Part 2 of the case study
