Light-sport operators keep a surprising amount of information in motion. Hobbs readings go from a tablet to a spreadsheet, inspection due dates feed reminders, and rental hours drive invoices. Each handover is a place where numbers can quietly go wrong. This checklist lets you audit those handovers in about an hour with a pen. Work through it, tick what you can prove, and score yourself at the end.
The structure follows ideas from the Ops Atlas page on DataOps, which treats the field as running data pipelines under engineering discipline. The Atlas calls the term emerging and its maturity scale its own scheme, not an industry standard, so use the score as a conversation starter, not as a certificate.
Before you begin
Set aside a quiet hour, ideally with the person who maintains the spreadsheet or scheduled job and one person who reads its output. The conversation between those two is often worth more than the ticks.
- Pick one flow: for example, flight hours recorded in the cockpit that end up on the maintenance schedule.
- Name the people who read the output. They are your consumers.
- Print the checklist. Tick an item only if you can show evidence in a file or screen, not memory.
Section A: Ownership and tiering
- A named person owns this flow and can describe where each field comes from.
- The flow has been assigned a tier: critical, important or exploratory.
- If it is critical, there is a stated time by which the output must be ready each day.
- Someone is reachable when that time is missed.
The Atlas explains that critical pipelines justify contracts, audits that can block publication, a service target and an on-call contact, while exploratory ones need only version control. A due-date feed for annual inspections is plainly critical, because a stale date could mean flying past a deadline unnoticed. A chart of busiest weekends is not, and a missed update there costs nothing worse than a slightly dated slide.
Section B: Change control
- Spreadsheet formulas, scripts and scheduled jobs are stored somewhere with history, not only on one laptop.
- Edits are reviewed by a second person before they go live.
- You can say what changed last week and who changed it.
- Before any big restructuring, there is a snapshot you can return to.
The Atlas also cautions against rewriting history with no snapshot to fall back on, and a shared spreadsheet makes that remarkably easy to do by accident.
Section C: Checks on every load
- A check confirms the newest record is recent enough.
- A check confirms the number of entries is plausible for the day.
- A check confirms identifiers, such as registration marks, are present and not duplicated.
- Thresholds account for weekday and weekend differences instead of using one average.
- Checks that matter can stop publication; they do not merely send a warning.
- Reruns replace a whole day’s records rather than adding a second copy.
On the fourth item, the Atlas stresses that fixed limits across mixed patterns can miss a partial load that a weekday-only baseline would catch. On the sixth, it recommends replacing partitions rather than appending, so a retry cannot double-count anything.
Section D: Late and changed data
- Records keyed in a day or two after the fact are still captured, for instance by re-reading recent days on each run.
- Counts are compared against the source system, not only against yesterday.
- Someone has written down what each status value means, such as what “deferred” covers on a defect list.
- The producer of any feed you depend on knows you depend on it.
Section E: Watching and learning
This section is about whether the operation improves over time. A flow can pass every check today and still drift, so look for evidence of regular attention, not a single good day.
- There is a log of every wrong number found in the last quarter.
- Each log entry records who found it, you or a consumer.
- Someone reads the log weekly and acts on patterns.
- Warnings that nobody reads have been fixed or removed.
- Personal details of pilots or renters do not appear in any debugging output or shared test copy.
Scoring yourself
The Atlas offers a five-level scale. Level zero is ad hoc. Level one means the work is versioned. Level two means tested and monitored, and the Atlas suggests tracking the consumer-detected share for a quarter before claiming it. Level three adds contracts and audited publishing, with no unaudited critical publication for a quarter. Level four is the optimised stage, with seasonal monitors, managed cost and a low and falling consumer-detected share.
For a quick mapping: mostly empty Section B means level zero. Sections A and B complete means level one. Sections C and E mostly complete means you are near level two. Section D plus a signed producer agreement points toward level three.
What to do with the result
- Choose the three unticked items whose failure would cost you most this month.
- Assign each an owner and a date.
- Re-run the audit in thirty days and compare.
If the result feels discouraging, remember that most small operators start at level zero. Keeping the checklist honest matters more than the score. A related read on reclaiming control over where your money goes uses the same attitude: look at what actually happens, not what you assume happens.
For anyone planning to build technical skills in this direction, a guide to building tech skills through DevOps training is a helpful companion, and the audit above makes a good practice project.

