Skip to content

Systems and records

Agriculture maintenance episode data: the history behind a failure

By SourceX Editorial · Reviewed by Noah Loul ·

10 sections

Short answer

A farm equipment failure is explained by the history around the work order: the diagnostic codes that appeared and were cleared, engine hours and the field work the machine was doing, earlier service, the parts and fluid lots installed and any substitutions, the warranty decision, and how many hours the machine ran afterward. Agriculture maintenance episode data joins all of these into one record with a verified outcome.

Record timeline

Agriculture maintenance

This page covers agriculture maintenance episode data: the history behind a failure.Illustrative

Key takeaways

  • The repair order sits in the middle of a failure's timeline; the codes before it and the hours after it are what make it explainable.
  • Use engine hours as the episode clock, because farm machines sit idle for months and calendar gaps mislead.
  • Record parts and fluid lots, and any substitute part with who approved it, so a repeat failure can be traced to the part rather than the machine.
  • A diagnostic code that was cleared is not a code that was resolved; the episode needs the reappearance history.
  • An episode passes only when its outcome is measured in operating hours after the repair, not by the repair order's closed status.
  • Telematics, dealer and warranty records often sit with different parties, so the right to supply each piece is checked before linking.

Why is the repair order only the middle of a failure's story?

A repair order says what a technician found and replaced on a given day. The failure itself usually started earlier and its result shows up later. A bearing begins to run hot weeks before it seizes, a fault code appears, gets cleared in the cab and comes back, the machine works through a wet harvest, and after the repair it either runs the rest of the season or returns to the shop. An AI system trained on repair orders alone sees only the day in the shop.

Most of that history is already recorded somewhere. Heavy-duty and agricultural machines built on SAE J1939 broadcast active diagnostic trouble codes, and the J1939-73 diagnostics layer includes reading and clearing them. Manufacturer telematics platforms keep those alerts: Deere's developer documentation for its Machine APIs, for example, describes the alerts captured on the telematics terminal and notes they are also called DTCs. Deere has said newer Equipment APIs are replacing those Machine APIs, so field names can change. The dealer management system holds the repair order and parts, and the warranty system holds the claim.

Agriculture maintenance episode data is what you get when those pieces are joined around one failure, in time order, with a measured result at the end.

Why should engine hours, not calendar dates, be the episode clock?

Farm machinery works in bursts. A combine may run long days for a few weeks and then sit until next year, so two events a month apart can be minutes of operation apart, and two events a week apart can span hundreds of hours. Order and measure the episode in engine hours, and keep calendar timestamps alongside them for joining to weather, field records and the dealer system.

Telematics platforms report engine hours with a timestamp; Deere's engine hours resource, for instance, returns readings with their report time and can return the last known reading. Where a machine has no telematics, take hours from the repair order and service records, and flag the episode as having manual hour readings.

Field context belongs on the same clock. ISO 11783-10 defines how the ISOBUS task controller exchanges task data with farm management software, which is where crop, field, operation type and implement often live. Linking the operation the machine was doing when the codes appeared tells a model whether a failure followed heavy tillage in wet clay or light work in dry conditions.

What goes in the maintenance history and outcome linkage schema?

The schema has twelve fields grouped by question: which machine, what symptom, under what conditions, what was done with which parts, and what happened next. Field names are suggestions. Following ISO 14224's practice of keeping equipment, failure and maintenance data apart, each group is stored separately and linked by the episode key rather than written into one note.

Keep owners, operators and technicians as role codes and pseudonymous keys from the start, and keep the machine serial number out of anything that leaves the business, replacing it with a stable key whose mapping stays with you.

Give the schema a version and a date, and store that version on every episode. When a field is added later, such as a fluid sample result or a second operator report, older episodes keep the version they were built under, and anyone using the data can tell a missing value from a field that did not exist yet.

What goes in the maintenance history and outcome linkage schema?
FieldTypical sourceWhat it recordsValidation rule
episode_id, machine_key, model_familyDealer or fleet asset registerOne failure episode on one machine, with model family and build year bandmachine_key stable across all linked records
code_historyTelematics alerts or service tool logEach diagnostic code occurrence with engine hours, active and cleared timesAt least one occurrence or an operator-reported symptom; cleared codes kept
operator_reportService request or call noteThe symptom as reported, in reviewed plain wordsFree text reviewed for names and locations
operating_contextTask data, field records, telematicsOperation type, crop, implement and conditions in the hours before failureContext window ends at or before the repair start hours
prior_serviceRepair orders and service plan recordsEarlier service on the same system with engine hoursLinked by machine_key; earlier hours than this episode
repair_orderDealer management systemComplaint, cause and correction, labor operations, hours at repairRepair start hours later than the first code occurrence
parts_lotsParts invoice and bin or lot recordsPart numbers, supplier lot or batch, fluids and filters installedEvery installed part has a lot or a recorded reason it has none
substitutionRepair order notes or parts approvalAny substitute or alternate part and the role that approved itApproval recorded before installation
warranty_decisionWarranty claim systemClaim filed or not, decision and failure code assignedDecision dated after the repair; failure code from one named list
outcomeTelematics hours and later repair ordersEngine hours run after repair and any repeat on the same systemWindow in engine hours set at close; repeat links to this episode

What does a passing episode look like? A combine feederhouse bearing at harvest (illustrative)

The following episode is invented to show structure. It does not describe a real machine, dealer or farm, and every figure is illustrative.

A combine logs a feederhouse drive speed code early in harvest. The operator clears it twice in the cab over the next few working days and keeps cutting. The third occurrence comes with a reported grinding noise, and the machine is brought in. The technician finds a failing bearing, replaces it with a part from a named supplier lot, and notes that the original part was out of stock and an alternate was approved by the service manager before installation. A warranty claim is filed and approved with a bearing failure code. The outcome window was set at 150 engine hours, and the machine finishes harvest past that mark without a repeat.

What does a passing episode look like? A combine feederhouse bearing at harvest (illustrative)
Engine hoursEventSource recordSchema field
1,212Feederhouse drive speed code first appears; cleared by operatorTelematics alertcode_history
1,229 and 1,241Same code reappears and is cleared; field operation is corn harvest in wet conditionsTelematics alert and task datacode_history and operating_context
1,247Third occurrence with operator report of grinding noiseService requestoperator_report
1,248Bearing replaced with approved alternate part from a recorded lotRepair order, parts invoice, approval noterepair_order, parts_lots, substitution
1,248Warranty claim filed and later approved with a bearing failure codeWarranty systemwarranty_decision
1,248 to 1,410Machine completes harvest past the 150-hour window with no repeat codeTelematics hours and alertsoutcome

Which checks catch the same episode recorded badly?

Now keep the same machine but change what was recorded. The cleared codes were never exported, so the code history starts at the shop visit. The alternate bearing has no lot number and no approval. The repair order is closed, and 40 engine hours later the same symptom opens a new repair order that nobody links to the first. A model would read this as a clean, successful bearing replacement and a separate new failure.

At a glance

Which checks catch the same episode recorded badly

Run the checks below on every episode. A record that fails one of the first four is repaired from source if the source exists; if it does not, the episode is excluded with a reason code so the gap is visible in the dataset description.

  • Sequence: first code or reported symptom, then repair start, then repair close, all in increasing engine hours.
  • Cleared-code coverage: the code history includes cleared occurrences, not only codes active when the machine arrived.
  • Parts traceability: every installed part has a lot or batch, or a recorded reason that none exists.
  • Substitution approval: any alternate part has an approving role and an approval time earlier than installation.
  • Outcome source: outcome hours come from telematics or later service records, never from the repair order status.
  • Repeat linkage: a later repair on the same system inside the window is linked and marks this outcome failed.
  • Code list: failure and cause codes come from one named list and version across the dataset.

Can the episode be replayed through parts, lots and decisions?

A useful test is whether someone outside the shop could replay the episode: start from the first code, follow each record forward, and reach the same conclusion about cause and result without asking anyone. If replay depends on a technician remembering why an alternate part was used or which bin it came from, the record is not finished.

Replay also shows when the failure belongs to a part population rather than a machine. If several episodes across different machines share a supplier lot and fail inside their outcome windows, the lot is the common factor. That pattern only becomes visible when parts lots are stored as structured fields rather than lines on an invoice, and it is one of the reasons linked episodes are worth more than a pile of repair orders.

How do you join dealer, telematics and warranty records without guessing?

Join on keys the systems already share before falling back to time windows. The machine serial or product identification number usually appears in the dealer management system, the telematics platform and the warranty claim, so it becomes the anchor for machine_key. The repair order number usually appears on the warranty claim and the parts invoice, which ties those three together without any matching by date.

Codes are the hard part, because a telematics alert never names the repair order it led to. Link code occurrences to a repair order when they fall on the same machine and the same system within a set number of engine hours before repair start, and record that the link was rule-based. Where the technician's cause and correction text names the code, mark the link as cited. A field showing cited, rule-based or reviewer-confirmed lets a buyer choose how strict to be.

Prior service is linked the same way, by machine and system, looking back a fixed number of hours. Keep the look-back and the code window the same for every episode in a dataset and write them in the dataset notes; changing them episode by episode makes the outcomes impossible to compare.

What should be removed before machine histories leave the business?

Linked machine history can identify a farm even when no name appears. GPS tracks and field boundaries show where a family farms; the combination of model, build year and dealer area can point to a single customer; operator notes may name people, neighbors or gate codes. Replace locations with coarse regions, convert field boundaries into crop and soil categories, keep owners and operators as pseudonymous keys, and review free text line by line.

Some records describe other parties' work. Manufacturer service bulletins, diagnostic procedures and warranty policy text quoted in repair notes belong to the manufacturer, and supplier part drawings belong to the supplier. Keep references to them by document number and leave the text out. The episode still explains the failure, and the record no longer carries content you do not control.

Who can supply which part of the history?

The pieces of an episode rarely sit with one party. The farm owns the machine and often its field records; the dealer holds repair orders, parts records and warranty submissions; the manufacturer runs the telematics platform and the warranty decision system. Each set of records comes with its own terms, so before linking, confirm which party may supply which field and for what use.

Access is improving but access is not a license. Under a memorandum of understanding with the American Farm Bureau Federation, Deere committed to make diagnostic tools, software and documentation available to farmers and independent repair shops on fair and reasonable terms, without requiring it to let anyone override safety or emissions controls. Colorado's HB23-1011, effective January 1, 2024, requires manufacturers to provide parts, embedded software, firmware, tools, documentation and, with owner authorization, data to owners and independent repair providers. Neither creates a right to license manufacturer data for AI training, and dealer agreements and platform terms still need review by counsel.

Taking machine histories into a licensing conversation

Dealer groups and large farming operations can work with SourceX on this, with SourceX covering rights review, contracting, data preparation, delivery and payment. For a dealer group or a large operation, the schema above doubles as a way to describe what the history contains, and which fields you can supply yourself, before sharing anything.

Buyers typically want two or more years of consistent machine history in one system and tens of thousands of records or more, though rarer work can qualify with less, and each case is assessed on its own. Machine records are delivered only after the dealer or farm business approves scope and terms and signs, and qualification is checked first through a short assessment with nothing uploaded.

Frequently asked questions

What should a maintenance history and outcome linkage schema include?

It should cover the machine, the diagnostic code history including cleared codes, the operator's report, operating context, prior service, the repair order, parts lots, any substitution and its approval, the warranty decision and the outcome measured in engine hours. Each field needs a named source and a validation rule.

Why use engine hours instead of dates for farm equipment failures?

Farm machines sit idle for long stretches, so calendar time between events says little about how much the machine worked. Engine hours put code occurrences, repairs and outcomes on the same operating clock, while dates are kept for joining to weather and field records.

What makes a farm equipment failure record fail validation?

Typical failures are missing cleared codes, an alternate part with no lot or approval, an outcome taken from the repair order status, and a repeat repair that is not linked back. Any of these makes a failed repair look like a success.

Does right-to-repair access mean a dealer or farm can license telematics data?

No. Commitments and state laws on repair access give owners and independent shops tools, documentation and in some cases data for repair work. Licensing records for AI training is a separate question governed by platform terms, dealer agreements and contracts, which counsel should review.

Are warranty decisions useful in a maintenance episode?

Yes. The warranty decision adds an independent failure classification and confirms the failed part, which helps label the episode. It should be dated after the repair and coded from a single named list.

Sources

  • Deere's developer documentation describes machine alerts captured on the telematics terminal, also referred to as diagnostic trouble codes. Source
  • Deere's engine hours resource returns engine hour readings with a report time and can return the last known reading. Source
  • SAE J1939-73 diagnostic messages include periodically broadcasting active diagnostic trouble codes and reading or clearing them. Source
  • ISO 11783-10 covers ISOBUS task controller and farm management information system data interchange. Source
  • ISO 14224 separates equipment, failure and maintenance data and does not cover direct cost data. Source
  • Under its memorandum with the American Farm Bureau Federation, Deere committed to make diagnostic tools, software and documentation available on fair and reasonable terms, without overriding safety or emissions controls. Source
  • Colorado HB23-1011, effective January 1, 2024, requires agricultural equipment manufacturers to provide parts, software, tools, documentation and, with owner authorization, data for repair. Source

See if your company qualifies

A short company assessment. No data uploads are needed.

See if your company qualifies