Operational Discovery Report
“The warehouse floor and the ERP are two different companies.”
Halden Logistics — Manufacturing & Distribution
112 people mapped · 41 workflows · 18 cross-team patterns · 3 agents scoped · Interviews conducted February 10–24, 2026
The map, in one paragraph.
Halden Logistics is a 500-person manufacturer and distributor operating two parallel systems of record that don't talk to each other: the ERP that finance and leadership trust, and the physical reality of the warehouse floor that the operators trust. We interviewed 112 people and built a map of 41 workflows, 18 cross-team patterns, and 29 pieces of institutional knowledge. The reconciliation gap between the ERP and the floor is staffed by 14 people whose entire job is closing it, and it costs somewhere between $380,000 and $440,000 per year in labor alone. The full map is visible at a private URL. This document is a snapshot as of February 24, 2026.
The three patterns the map revealed first.
The warehouse floor and the ERP are two different companies.
The map revealed that in practice, Halden operates two parallel systems of record. The ERP contains what leadership, finance, and external partners believe to be true about inventory, orders, and fulfillment. The whiteboards, clipboards, printed pick lists, and shift handover notes on the warehouse floor contain what's actually happening. These two versions of reality are never more than roughly consistent, and they're reconciled by a team of 14 people — inventory analysts, shift supervisors, and two full-time ‘data cleanup’ roles in operations — whose entire job is closing the gap between them.
We mapped the reconciliation workflow by interviewing every person on it. The workflow has 23 steps, spans 4 departments, involves 3 different spreadsheets and 2 Slack channels, and includes at least 5 moments where someone prints something, writes on it by hand, and then types the handwritten version back into a different system. Nobody we interviewed could describe the whole workflow end-to-end. Each person knew their 3–4 steps and had a rough sense of what happened upstream and downstream.
The reconciliation is, on average, 9 days behind reality. This means decisions made by leadership on Monday are based on what the warehouse floor looked like the previous Wednesday. When we asked finance whether they knew this, two of the three people we interviewed said yes and had stopped mentioning it because nothing had ever been done about it. The third said no and was visibly unsettled.
Estimated annual cost of the reconciliation gap: $380,000–$440,000 in labor, plus unquantified downstream effects on inventory carrying cost, customer commitments, and invoice disputes. Estimated cost to close: moderate — this is infrastructure work, not a simple automation.
Your best shift supervisors are your worst-documented asset.
The map revealed that three shift supervisors — two on the day shift, one on nights — are independently solving problems that the formal systems can't. They know which orders to prioritize when the ERP says everything is equally urgent. They know which customers will accept a 2-day delay and which won't. They know which of the pickers are reliable on which kinds of orders. They know that the Tuesday morning inbound from one specific vendor always arrives miscounted and needs a manual recount.
None of this knowledge exists anywhere written down. All three of these supervisors described being called at home on weekends and evenings when something goes wrong. Two of them are within three years of retirement. One has been training a replacement for six months and told us, unprompted, that the replacement ‘still doesn't have the feel for it.’
If any one of these three people leaves, Halden loses operational knowledge that would take months to reconstruct and may never be fully recovered. This is not a theoretical risk — one of them mentioned that their partner has been pushing them to retire early. We are not their therapists and didn't press, but the signal was clear.
Estimated urgency: high, and driven by retirement risk rather than recoverable hours. The work is less about automation and more about knowledge capture — interviews, observed workflows, decision rules turned into systems the rest of the team can query. Cost to address: moderate. Cost of not addressing: very high and hard to reverse.
The quality report is being typed from paper every single day.
The map revealed that at the end of every shift, quality inspectors fill out a paper form with roughly 40 fields per line item inspected. The forms are stacked in a tray at the end of the shift, and early the next morning, an operations analyst types the contents of the forms into the ERP's quality module. This takes her approximately 3 hours per day. She has been doing it for 4 years. She was hired to do process improvement work and spends 60% of her time on this data entry instead.
When we asked the quality inspectors why they write on paper first, the answer was that the ERP interface on the shop floor terminals is ‘too slow and clunky to use in real time’ and that they've tried switching to direct entry three times and reverted each time. When we asked the analyst why nobody has automated the paper-to-ERP transcription, the answer was that nobody has ever asked her what she does all day.
This was the single most consistently frustrating thing we heard about in the operations interviews. Six separate people mentioned it without prompting.
Estimated time recovered: 15 analyst hours per week, plus the unquantified strategic value of getting that analyst back on the process improvement work she was hired for. Cost to address: low. This is the easiest high-impact fix in the report.
The eleven mid-tier patterns.
These are presented more briefly. Each was mentioned by multiple interviewees or surfaced as a pattern across teams.
Customer service and operations are using different definitions of 'on time.'
Customer service marks an order on-time based on when it left the warehouse. Operations marks it on-time based on when the customer received it. The gap between these two definitions is currently around 7%, which explains why the CS team and the operations team disagree about performance in every monthly review.
Worth resolving before any reporting automation is built on top of it.
The Monday inbound huddle is 40 minutes long and 70% of it is information that's already in the ERP.
Six people described the huddle as 'the meeting where we find out what the computer already knows.' The remaining 30% — the operator intuition about the week ahead — is the part that matters and doesn't exist anywhere else.
Reclaim 4 hours per week across 6 people, keep the intuition.
Inventory adjustments are happening in a Google Sheet that three people have access to.
When the ERP count doesn't match the physical count, one of three people makes a judgment call about which to trust and updates the ERP. This judgment is not logged anywhere. Over the 6 months we sampled, approximately $1.2M in inventory value was adjusted this way.
Not an automation finding. A governance finding.
Driver dispatch is being routed through a whiteboard in the transportation office.
The whiteboard is photographed every morning and the photo is shared in a WhatsApp group. This is the actual dispatch system. The formal TMS is used mostly for billing.
Worth a separate conversation — may not need 'fixing' so much as formalizing.
New hire training for warehouse staff is done by pairing them with whoever is free, not whoever is good at training.
Consequently, new hire productivity ramps over 6–10 weeks depending on who trained them. The gap between the best and worst trainer is roughly 4 weeks of productivity.
~$8,000 per new hire in recovered ramp time. With 60 hires a year, this is significant.
The daily production report is assembled by three different people working from three different data sources.
By the time it reaches the leadership meeting at 9am, it has been manually typed, re-typed, and compared by hand. The three versions disagree 30% of the time.
~12 hours/week recoverable and the report becomes trustworthy.
Returns processing is a black hole.
When a customer returns a product, it enters a manual workflow that takes an average of 11 days to resolve. During those 11 days, the product is physically present in the warehouse but not in any system.
Operational and financial impact.
Shift handovers rely on a shared notebook that's been in use since 2019.
The notebook lives on the supervisor's desk. When a supervisor is sick, the handover for that shift is lost.
Recoverable into a queryable system easily.
Maintenance requests are submitted via email to one person.
That person is on vacation twice a year for two weeks at a time. The maintenance queue becomes a mystery during those periods.
Routing problem, easy fix.
The safety incident log is a Word document on a shared drive.
It is updated inconsistently, not searchable, and never analyzed. Three safety incidents in the last year had similar root causes that would have been visible in any halfway structured log.
Compliance and safety risk.
The CFO mentioned, unprompted, that she 'doesn't really trust the inventory numbers.'
This was not a finding we went looking for. It emerged in a single interview and we think the CFO would prefer we not dwell on it. We think you should dwell on it.
Not an automation finding. A signal.
The agent blueprint.
For each high-value pattern, a specific AI agent to build — in the order we'd build them, with estimated hours recovered and the reason it sits in that slot.
Quality form transcription agent
Shift supervisor decision capture agent
ERP-to-floor reconciliation agent
A note on what the map didn't reveal.
A few things worth saying, because documents like this tend to oversell.
We didn't find evidence that your people are the problem. We interviewed 112 of them and came away impressed by how much effort is being spent compensating for systems that are making the work harder than it needs to be. The findings in this report are about the systems, not the people.
We didn't find a reason to believe that the reconciliation gap is anyone's fault. It has grown up over years of small decisions, each of which made sense in isolation, none of which anyone is responsible for. Trying to assign blame for it would be counterproductive and untrue.
We didn't find anything that would justify cutting headcount. The 14 people on the reconciliation workflow are not the target of any of these recommendations. If the recommendations are implemented, those 14 people become available for the higher-leverage work that Halden has been putting off for years because these 14 people were stuck doing reconciliation. The cost savings in this report come from redirected effort, not eliminated effort.
This is a snapshot. The real map is live.
This document captures the Plainsight map for Halden Logistics as of February 24, 2026. The live map is hosted at a private URL for the leadership team and continues to update as new interviews happen and patterns shift. If you are the leader who commissioned this map, your URL is in your email. If you are evaluating Plainsight, the closest thing to a live map view is in the three sample companies on our homepage.
This is what your company's map could look like.
Add your team and we'll start the interviews this week. You'll watch your map build in real time.