I've sat through more turnaround close-out meetings than I can count. They tend to end the same way. Someone produces a lessons-learned report, or a deck, sometimes a bound document if the operator is rigorous about it — and it goes on to a bookshelf. Two years later, when the next event starts, almost nobody opens it. The team that wrote it have moved to other roles, or other companies, or retired. The new planning team starts from work order templates, poor quality SAP notifications, and their own judgment.
I used to think that was a discipline problem, that operators just needed better document control. I no longer think that. The report isn't the thing that's missing. The person who could have told you, without opening it, that the catalyst handling job always overruns by two days because of an access restriction nobody puts in the scope — that person is the thing that's missing. And that person has moved on, taking your plant knowledge with them!
The 2026 turnaround season is a stress test for this problem
This is shaping up to be one of the busiest turnaround years on the US Gulf Coast in some time — according to ExecGraph, eleven major facilities have events scheduled in 2026, including Motiva's Port Arthur refinery, the largest single-train refinery in North America. Each of those events draws on the same regional pool of planners, schedulers, and skilled craft labour.
Article: Gulf Coast operators are front-loading maintenance. Most will leave value on the table.
At the same time, a survey reported by Petrochemical Update found that 62% of refinery operators say they struggle to source skilled workers for turnarounds. That figure isn't just about headcount — it's also about the people who plan and execute your STO events, the roles where experience compounds most and is hardest to replace quickly.
Put those two facts together, and you get a straightforward mechanism: more events, competing for the same shrinking pool of experienced people, at the exact moment a large cohort of that pool is also reaching retirement age. The result isn't that turnarounds go unplanned — they get planned by people with less direct memory of how that specific unit, that specific equipment list, was tackled last time, the time before that, and the time before that.
A turnaround is not a one-off event
Tom Lenahan made this point two decades ago, and it's more relevant now than when he wrote it. A turnaround is not an isolated project. It's one event in a sequence. The strategic objective — fewer, shorter, better-prepared events over time — depends on each event making the next one better. That only happens if what was learnt on event N actually changes how event N+1 is scoped, planned, scheduled, executed, and lessons learnt are captured.
Lenahan's framework assumes that transfer happens through people: the planner who ran the last event sits down with the team planning the next one and says, “here's what we got wrong, here's what took longer than estimated, here's the access problem on Unit 4 that nobody flags until it's too late.” That assumption was reasonable when planners stayed with one operator, on one site, for a career. It's a much weaker assumption today.
Where the knowledge actually lives — and where it doesn’t
It's worth being precise about what “institutional knowledge” means in a turnaround context, because it isn't one thing. It's at least three things, and degrade differently:
- Estimates. How long a job actually took last time, versus what the original estimate said. This is recorded — somewhere — in whatever cost and schedule system the operator used, but it's rarely connected back to the equipment record that can be used for the same job next time.
- Sequencing knowledge. Which jobs have to happen in a particular order on a particular unit, for reasons that aren't obvious from the equipment list — a scaffold access route that only works in a specific configuration, a permit dependency between two work areas that share an isolation point, a discipline that has to finish before another can start because of physical interference. This is almost never written down. This understanding lives in the head of whoever executed the unit last time.
- Planning and Productivity Factors. The adjustment an experienced planner makes to a standard estimate because they know the contractor, the equipment, and the site. The underlying idea — that raw historical durations need a judgement factor applied based on specific context — runs through his case studies.
All three of these are currently stored in the same place: a person. When that person leaves, the lessons-learned report captures a small percentage of the first category, some of the second, and none of the third.
The comparison that matters: a report validates against past plans and recorded actuals from previous events.
The alternative to “the planner remembers” isn't “a better lessons-learned report.” A report is a static document, written once, read rarely, and disconnected from the planning tool where the next event's work orders are actually built. Even a well-written report can't be queried at the point a planner is sitting in front of a specific work order, asking “has this job overrun before, and by how much?”
The alternative that actually closes the gap is a system that captures executed actuals — durations, resource usage, sequencing dependencies — against the equipment library itself, so that the next time that equipment (or one similar) is planned, the actuals are sitting there rather than buried in a report from two events ago.
This is the mechanism behind iPlanSTO's closed-loop execution feedback. As work packs are executed, planned vs actuals, and schedule variance, are tracked against the original plan and written back to the equipment library. The next time a planner builds a scope for a similar job — same equipment type, same discipline — the AI Agent validates the new plan against those recorded actuals from previously executed work packs, not against the original estimate that may never have been accurate in the first place. The planning library is corrected by what actually happened, rather than reinforced by what someone guessed five years ago.
This matters most for exactly the second category above — sequencing knowledge — because it's the category that's hardest to write down and easiest to lose. iPlanSTO's activity logic linking capability lets an operator encode dependencies like shared permits, discipline ordering, and same-equipment sequencing once, as a rule attached to the work order type or the unit, rather than as something a veteran planner mentions in a handover meeting that may or may not happen. Once it's a rule, it applies whether the planner building next year's scope has done five turnarounds on that equipment tag, or none.
This is the next stage of a problem we’ve already written about
We've covered before how two planners working from the same activity can produce two different estimates for the same job (read more here) — the planning inconsistency that comes from estimation being a matter of individual judgment rather than shared, validated data. Institutional knowledge loss is what happens to that problem when the more experienced of those two planners isn't in the building anymore.
If your current planning process relies on the better planner's judgement to catch the gaps the worse planner's plan would miss, you have a single point of failure that retirement, resignation, or a competitor's job offer can remove without warning. The fix for planning inconsistency and the fix for institutional knowledge loss are, in practice, the same fix: move the validation reference from “what an experienced person remembers” to “what was actually recorded the last time this job was executed”, recording it against the equipment record.
What this means for the 2026 season specifically
Operators running events this year are competing for planners against their neighbouring facilities, doing the same thing at the same time. Some of that competition is unavoidable — it's a labour market problem, and no software changes the number of qualified turnaround planners in the Gulf Coast region this year.
What can change is how much of the planning quality for any given event depends on which specific individuals you manage to secure at the end of this cycle. An operator whose work order and equipment library carries actuals, sequencing rules, and lessons learnt from its last three events can put a less experienced planning team on this year's event and still start from where the last event finished, rather than starting from the templates and a lessons-learnt deck nobody can find on the bookshelf.
The operators who come out of the 2026 season ahead won't necessarily be the ones who won the fight for the most experienced planners — there are only so many of those to go around. They'll be the ones who adopt a planning system to avoid a reset to zero every time someone walks out the door.
About the Author: Ross Coulman is the Managing Director of IAMTech, a global leader in industrial asset management and technology solutions. With over 20 years of experience in the sector, Ross has driven IAMTech’s growth from a start-up to a trusted partner for the oil, gas, chemical, and power industries worldwide. Passionate about innovation and sustainability, he champions the use of digital transformation to enhance efficiency, safety, and compliance across complex industrial operations
Tom Lenahan's Turnaround, Shutdown and Outage Management (Elsevier, 2005) is the source for the framing of turnarounds as a sequence of events rather than isolated projects.
Sources for 2026 season data:
