A mid-size flat-rolled steel plant running two shifts, roughly 900 employees, and an aging fleet of utility systems had a familiar problem: nobody could say with confidence where the plant's energy and material waste actually lived. Compressed air leaks, idle machines left running between heats, and untracked scrap all showed up somewhere in the monthly utility and materials bill, but no single view connected the symptom to the asset causing it. This is the account of how that plant structured a maintenance-driven waste reduction program, the platform evaluation behind it, and the 25-item action list that carried it from diagnosis to a documented annual reduction.
Plant Profile
The Problem: Waste Everyone Suspected, Nobody Could Quantify
Plant leadership had a general sense that idle machines, compressed air leaks, and untracked material handling losses were costing money. What they lacked was a system that connected a specific waste event to a specific asset, a specific shift, and a specific corrective action, which meant every improvement effort started from a guess rather than a record. Three waste categories consistently surfaced in informal walkarounds but never made it into a structured tracking system: machines left running during planned downtime, compressed air escaping through unrepaired leaks, and scrap generated by equipment operating outside calibrated tolerance. None of the three had an owner, a baseline, or a repeatable measurement method before the program began.
Selecting the Platform
The plant evaluated several CMMS platforms against three requirements: mobile work order capture from the floor, asset-level energy and downtime tagging, and reporting that could roll idle-time and leak data up to a dollar figure leadership could act on. Oxmaint was selected as the platform that met all three without requiring a separate energy monitoring system bolted on afterward.
Rollout Sequence
The 25-Item Action List, By Category
The corrective backlog that came out of the tagging phase totaled 25 discrete action items. Rather than working them in ticket order, the team grouped them by waste category and estimated dollar impact, then sequenced the highest-impact, lowest-effort items first.
| Category | Action Items | Example Fixes | Est. Annual Impact |
|---|---|---|---|
| Idle Machines | 9 items | Auto-shutdown logic on conveyor lines, standby state enforcement on utility compressors, operator idle-time checklist | $480K |
| Compressed Air Leaks | 8 items | Fitting replacement on pickle line header, valve stem repacking, quick-connect standardization | $390K |
| Scrap From Calibration Drift | 5 items | Gauge recalibration cadence tied to trend, sensor replacement on out-of-spec loops | $340K |
| Material Handling Losses | 3 items | Conveyor transfer point spillage guards, bin level sensor recalibration | $190K |
Working the Idle-Machine Category First
Idle machines were the largest category by dollar impact, and also the fastest to fix once tagged. Conveyor lines running between heats, utility compressors left loaded overnight, and auxiliary pumps cycling with no downstream demand accounted for the bulk of the 340 idle-time events logged in the first year. The fix in most cases was not new equipment but enforced logic: auto-shutdown timers on conveyor lines idle beyond a set threshold, and a standby-state discipline on utility compressors that had previously been left running "just in case" a demand spike arrived. None of the nine idle-machine action items required capital spend; all nine were logic and procedure changes tracked through work orders.
Working the Compressed Air Category Second
The plant's compressed air leak survey found 210 active leak points across a header that had never been systematically surveyed before, spread across pipe joints, quick-connect fittings, and valve stems on lines installed over three separate plant expansions. Leak severity was scored by estimated CFM loss so the repair backlog could be ranked the same way as the idle-machine list, by dollar impact rather than location or discovery order. Fittings on the pickle line header accounted for a disproportionate share of the estimated loss, both because of their age and because that section ran at a higher pressure than most of the rest of the plant. Standardizing on a single quick-connect fitting type across that section, rather than the mixed inventory that had accumulated over the years, also reduced the rate at which new leaks appeared after the initial repair pass.
Working the Scrap and Material Handling Categories
Scrap tied to calibration drift proved the hardest category to close, because the causal chain ran from a gauge or sensor reading, through a control loop, to a dimensional or surface defect that only showed up several process steps downstream. The team's approach was to trend scrap rate by production line against each line's gauge calibration history, looking for the point where scrap began climbing shortly after a calibration interval had lapsed. That correlation let the team justify moving five specific calibration points from a fixed annual schedule to a shorter, trend-triggered cadence, which addressed the majority of the calibration-linked scrap without recalibrating every gauge in the plant on a tighter blanket schedule. Material handling losses, the smallest category, closed out with straightforward mechanical fixes: spillage guards at conveyor transfer points and a recalibration pass on bin level sensors that had been reporting inaccurate fill levels.
Results: Before and After the Program
What Made the Difference
The plant's earlier waste-reduction attempts had not failed for lack of effort — technicians already knew where many of the leaks and idle machines were. What changed was the ability to attach a dollar estimate and an owner to each finding the moment it was logged, rather than letting it sit in a notebook until someone remembered to act on it. Ranking the 25-item backlog by estimated impact rather than working tickets in the order they arrived meant the highest-value fixes landed in the first two months of the corrective phase, which built the internal case for continuing the program through the harder, lower-visibility items later in the list. Start a free trial to build a similar backlog against your own plant's asset register, or book a demo to see the dashboard structure in detail.
Sustaining the Reduction Into Year Two
A common failure point in waste-reduction programs is treating the initial tagging pass as a one-time project rather than an ongoing discipline. The dashboard rollout in month nine was deliberately framed internally as the start of continuous tracking, not the finish line, with each category's owner responsible for reviewing new findings against the same dollar-ranking method used in the original backlog. By the start of year two, the leak count had not returned to zero, and it was never expected to — new leaks form continuously as fittings age and vibration works connections loose. What changed was the average time between a leak forming and its repair, which dropped substantially once tagging became a routine part of technician rounds rather than an occasional special project.
Internal Reporting Cadence
Operations leadership received a monthly rollup during the rollout and moved to a quarterly cadence once the dashboard stabilized, with each category owner presenting new findings, closed items, and any revised dollar estimates against the original baseline. That regular reporting rhythm, more than any single technical decision in the rollout, is what kept the program from quietly losing priority once the initial momentum of the first few months had passed.
Lessons for Plants Starting a Similar Program
How the Dollar Estimates Were Built
Every tagged item in the 25-item backlog carried an estimated annual dollar impact from the moment it was logged, not just after the fix was implemented. Idle-time items were estimated from the motor's rated power and the average hours per week the equipment sat idle beyond its operational need, multiplied by the plant's blended electricity rate. Leak items were estimated from a rough CFM loss figure assigned by severity tier during the tagging round rather than a precise flow measurement for every single fitting, which kept the survey moving at a pace the team could sustain across the whole plant in the available window. Scrap-linked items were the hardest to estimate up front, since the dollar impact depended on how far scrap rate had actually drifted from baseline on each affected line, a number the team only had confidence in once several weeks of trend data existed. For those items, the initial ranking used a conservative placeholder estimate, which was revised upward for three of the five calibration-related items once the real trend data came in and confirmed a larger impact than originally assumed.
What Did Not Make the Original List
Not every waste source the team suspected going in turned out to be worth pursuing as a standalone action item. A hypothesis that refractory heat loss on one furnace was contributing meaningfully to the energy bill did not survive the baseline data — the trend showed normal, expected variation rather than a fixable drift, and the team explicitly chose to leave it off the backlog rather than force a fix onto a problem that measurement did not confirm. That discipline, rejecting plausible-sounding waste sources that the data didn't actually support, mattered as much to the program's credibility internally as the fixes that did make the list. A backlog padded with items chosen on instinct rather than evidence would have made the entire dollar-ranking exercise harder to trust once leadership started checking results against the original estimates.
Applying This Structure Beyond a Single Plant
Once the first plant's dashboard and categorized backlog proved out, the same structure was extended to a second facility in the same operating group, with the category definitions and dollar-ranking method carried over directly. The second rollout moved noticeably faster than the first, largely because the tagging and categorization framework no longer needed to be designed from scratch, only adapted to that plant's specific asset register and shift pattern.
.png)
