A backlog estimate with no working kept
VThe Amtrak Office of Inspector General (OIG) published report OIG-A-2026-004 on 10 April 2026, auditing how Amtrak governs the assets behind its state-of-good-repair backlog. The company estimates the backlog at about $47 billion and has a target to clear it by 20401.
VThe estimate multiplies a standard unit cost for each asset class by the units not in good repair, and officials said the real cost could be higher or lower. Amtrak did not save which specific assets it counted, so no later repair can be shown to have reduced the figure, and the company does not use the number to track progress1.
VAmtrak's own estimate of the time to clear the backlog, given sufficient funding, rose from 10 years in its FY2017 to FY2021 plans to 15 years in its FY2022 to FY2024 plans. With no performance metric, the OIG says the causes cannot be separated1.
Two records, and no routine check that they agree
VAmtrak holds its official inventory in a geospatial system (GIS) and tracks maintenance work in Maximo. Comparing 18 of its 35 asset classes, the OIG found that all signal machines in Maximo appeared in GIS, but only 69 percent of defect detectors did1.
VThe OIG traced 1,726 assets in Maximo but not in GIS and 227 in GIS but not in Maximo. Substation alignment rose from 47 to 96 percent after the OIG shared its findings. Some gaps were retired or new assets, and at the time of review no process monitored whether the inventory was complete. The results cannot be applied to the classes the OIG did not review1.
AA backlog figure inherits the gaps of the register under it. When two records disagree, the number of assets behind a figure depends on which is asked, and only a routine check can say which is right.
The next estimate arrives before the fixes
VAmtrak management agreed with both recommendations. It set 30 September 2028 for the governance framework (strategy, policy and plan) and 30 September 2031 for the data actions, including condition assessments. It expects to publish its next plan and backlog estimate in 20261.
VOfficials involved in the work told the OIG they were unaware of the asset management policy, which sits in an appendix to the FY2024 plan, or that it did not inform their work1.
AA new estimate published before the controls exist will face the same problem as the old one. Unless the list of assets counted is kept from the start, the first honest test of the 2040 target waits for the data fixes.
Treat the figure as a conclusion that needs its working
AKIS PLUS LTD's view is that a headline backlog belongs with the record that produced it. Three controls would test it: a retained baseline listing every asset counted, with its unit cost and date; a routine completeness check between the inventory and the maintenance record, with a named owner for each difference; and reconciliation of every restated figure to the one before.
IIf the 2026 estimate is published with its baseline retained, the 2040 target can be tested for the first time. Without it, the target date may keep moving with nothing to measure the move against.
OUR PERSPECTIVEA backlog is a number until someone can show which assets are in it.
Three questions to ask this week
- For each headline figure we give a board or funder, can we rebuild it today from a list of assets we kept, and who holds that list?
- When did we last check that our asset inventory and our maintenance record describe the same assets, and who owns each difference?
- When we restate a figure, can we explain every change from the last version, asset by asset?