Two Buying Models, One Bad Assumption
I've spent about 11 years handling capital equipment procurement for heavy process plants—pumps, separators, dewatering systems, the kind of stuff that weighs more than a car and takes 14 months to arrive. In that time I've run projects both ways: single-vendor integrated delivery (what I'll call Model A) and split-sourcing where we buy each package direct and stitch it together ourselves (Model B). And I've come to a conclusion that surprised even me.
Here's what I'm comparing, and why this specific comparison matters. When you're specifying something like a complete dewatering line or a paper machine rebuild, you face a fork in the road: hand the whole scope to one integrated OEM who owns design, manufacturing, and commissioning, or break it into packages and manage the interfaces yourself. In January 2021, I was certain Model B was smarter. By March 2024, I'd changed my mind on most (not all) projects.
The three dimensions below are the ones that actually decide it. I'll compare each directly rather than describing A first and B later, because that's how the decision really works—you're not choosing features, you're choosing trade-offs.
Dimension 1: Where Responsibility Sits When Something Breaks
This is the dimension that cost me the most, so I'll start here.
Model A (integrated OEM): One throat to choke. If the pump cavitates because the upstream tank level control was designed wrong, that's still the OEM's problem, because they sold you the tank, the pump, and the control philosophy as one performance guarantee.
Model B (split-sourced): You own the interfaces. Each vendor delivers exactly what the spec says and nothing more. When the pump builder says "the NPSH available was never in our scope" and the tank fabricator says "we built to your drawing," you're the one holding the invoice.
In September 2022, I ran a mid-size separation upgrade on Model B. Three packages—feed system, separator skid, discharge conveyor. On paper, the interfaces were clean. In the field, the feed system's discharge flange sat 40mm off from the separator inlet because two vendors used different reference points on the same structural drawing. Neither would move. We paid a third-party fabricator roughly $19,000 to bridge the gap, plus a 9-day commissioning delay that pushed our startup window past a contractual deadline. That single interface issue ate the entire "savings" we thought we'd captured by splitting the packages.
Here's the counterintuitive part, though. Model B is not automatically worse. On a project where each package is genuinely independent—say, replacing two unrelated pump trains in different buildings—split-sourcing is faster and cheaper because there are no shared interfaces to fight over. The interface risk only exists where scopes touch. If your scopes don't touch, Model B wins cleanly. I wish someone had told me that in 2021.
Dimension 2: Sticker Price vs. What You Actually Pay Over 20 Years
The purchase order number is what gets approved. The total cost of ownership is what gets you in trouble.
Model A: Higher upfront number. Integrated OEMs bundle engineering, commissioning support, performance testing, and often a spare-parts commitment into the contract. That bundle shows up on the PO as maybe 12-18% above the sum of split-sourced packages, in my experience.
Model B: Lower PO number, but you absorb interface engineering, multi-vendor contract administration, and—this is the killer—divergent spare-parts logistics. I've seen a plant hold six different seal designs because six vendors each specified their own. That's six sets of training, six inventory lines, six lead times during an outage.
The mistake I made was comparing only the PO. My experience is based on roughly 40 mid-size capital projects, mostly in pulp & paper and minerals processing. If you're working at a much larger scale—say, a greenfield multi-line facility—the economics may shift because you can afford a full-time interface engineering team that makes Model B genuinely competitive. I can't speak to that scenario from direct experience.
One thing I've never fully understood: why identical pumps from two vendors can have such wildly different 10-year maintenance profiles even when the datasheets match. My best guess is it comes down to internal manufacturing tolerances that never make it onto the spec sheet. If someone has hard data on that, I'd genuinely like to see it.
Dimension 3: How Each Model Handles Technology Evolution
This is where the industry has genuinely changed, and where old buying logic breaks down.
Five or six years ago, "integrated vs split" was mostly about price and schedule. Not anymore. What was best practice in 2019 doesn't fully apply in 2025, because equipment now has to talk to plant-wide digital systems—condition monitoring, predictive maintenance platforms, historian integration.
Model A advantage: When one OEM supplies the rotating equipment and the controls, the data model is coherent by default. Vibration sensors, PLC tags, and maintenance triggers all speak the same language. You plug in and it works.
Model B reality: You inherit whatever communication protocols each vendor chose. I've watched a plant spend nearly $55,000 over eight months building middleware to translate three different condition-monitoring formats into one dashboard. The hardware was fine. The data was the problem.
But here's where it gets interesting. On pure mechanical performance, some specialized split-source vendors are actually ahead of the integrated giants, because their entire business depends on one technology. That was true 10 years ago when digital integration barely mattered, and it's still true now—just for different reasons. The fundamentals of mechanical reliability haven't changed, but the execution environment around them has transformed.
I'm not 100% sure this holds for every equipment category. For something like a complete paper machine rebuild—where Andritz Paper Machinery Ltd, Voith, and similar integrated suppliers operate—the coherence argument is strong. For a standalone dewatering screen, it's much weaker.
How I Decide Now (It's Not "A Is Better")
After two expensive lessons, this is the rule I actually use:
Choose integrated delivery (Model A) when:
- Three or more packages share physical or control interfaces
- Unplanned downtime costs more than $8,000/hour (then interface delays are existential)
- You don't have in-house interface engineering capacity
- The plant needs unified condition-monitoring data from day one
Choose split-sourcing (Model B) when:
- Packages are genuinely independent with no shared interfaces
- You have a strong internal engineering team that can own integration
- One specific vendor has a clear technology lead on a critical component
- Schedule pressure favors buying standard, off-the-shelf units fast
To be fair, I get why procurement teams default to Model B—the PO looks better and budgets are real. But the savings are often borrowed from future maintenance, not earned.
One last thing. When your project timeline runs past five years, the vendor's own financial stability becomes a line item in your TCO calculation. Public financial disclosures are worth checking—though I'd caution that market noise around a stock price isn't the same as operational health, and shouldn't be treated as a procurement signal on its own. That's a due-diligence task, not a trading decision.
If you're running a genuinely independent multi-package project right now, I'd honestly like to hear whether the interface math worked out for you. My sample is 40ish projects, and I don't pretend that's the whole picture.