By Jon W. Hansen | February 2026
For the Busy Executive
Surface metrics describe. Structural alignment determines outcomes. That was true in 1998 when a single question — “what time of day do orders come in?” — took DND delivery performance from 51% to 97.3%. It was true in 2009 when I participated in Virginia’s JLARC legislative review and disaggregated a 63% favorable survey into five competing realities. Forrester confirmed it independently in 2014 when they recommended Virginia reject an ERP migration in favor of the methodology-first approach I had documented since 2007. And it is the foundation of RAM 2025 in 2026, where multi-model AI scales the same principle across twelve models and thousands of data points. Four eras. Four independent validations. One constant: the industry keeps mistaking statistical surface for structural truth — and the outcomes keep proving them wrong.
Read time: 8 minutes
The Constant
Across twenty-eight years, four technology generations, and four independent moments of validation, one principle has held without exception:
Organizations that measure structural alignment produce outcomes. Organizations that measure statistical surface produce reports.
This is the story of those four moments — not as retrospective, but as a pattern so durable that it has now outlasted every technology cycle the procurement industry has produced. If the pattern holds across DND in 1998, a state legislative review in 2009, an independent Forrester assessment in 2014, and multi-model AI validation in 2026, it is not an observation. It is a law.
1998 — DND (RAM 1998): The Question Nobody Thought to Ask
When SHL SystemHouse, then part of MCI, was managing the Department of National Defence’s MRO procurement platform in the late 1990s, they were delivering 51% next-day against a contractual requirement of 90%. They came to me and said what every struggling procurement organization says: we need to automate.
The first question I asked was: “What time of day do orders come in?”
The answer — most orders arrived at 4:00 PM — unraveled the entire failure chain. Service technicians were incentivized to maximize daily service calls, so they practiced sandbagging — holding all parts orders until end of day. Dynamic flux commodities that cost $100 at 9 AM were $1,000 by 4 PM. Parts shipped late missed customs windows. Delivery performance collapsed. And call close rates were horrific because the parts weren’t arriving to service the calls the technicians had rushed through.
Nobody inside procurement could see this. The failure was cross-boundary, involving four distinct agents: the procurement team, the service technicians, the SME suppliers, and the logistics and customs layer.
The solution wasn’t automation. It was understanding the full agent ecosystem and redesigning how those agents interacted. We built self-learning algorithms, integrated UPS directly into the system, and worked with Canada Customs to pre-clear shipments. Within three months: 51% to 97.3%. Over seven years: 23% consistent cost reduction. The buying group compressed from 23 companies to three.
The technology served the redesign. It did not drive it.
The constant: Surface metrics said this was a technology problem. Structural analysis revealed it was a behavioral problem invisible from inside any single department.
2009 — JLARC: Five Competing Realities Behind One Statistic
A decade later, Virginia’s Joint Legislative Audit and Review Committee launched a review of the Commonwealth’s eVA procurement program — one of the most successful eProcurement implementations in North American public sector history.
What I don’t often mention is this: I was asked to participate in that review process. And I’m quoted in the official legislative record of the Commonwealth of Virginia.
JLARC based part of their assessment on a stakeholder survey. The aggregate result: 63% believed eVA was a benefit to small, minority, and women-owned businesses.
I didn’t challenge the survey based on the number of responses. I challenged it because the aggregate concealed structural disconnects that no single number could reveal.
So I disaggregated the same data across five agent dimensions. What emerged was not one story. It was five.
By job title: C-level executives and vice presidents — 100% believed eVA was not a benefit. Front-line workforce — 100% believed it was. Same program. Same survey. Opposite conclusions.
By company size: Enterprise and large businesses — 100% support. Medium businesses — evenly divided. SMEs — 67% support. The strongest endorsement came from the segment with the least at stake.
By job function: Sales professionals — only 50% favorable. Marketing and creative functions — 100% favorable. The people doing the daily work saw friction. The people seeing the concept saw opportunity.
By gender: Women — evenly split on whether eVA benefited women-owned businesses. Men — 67% believed it did. The program’s own target constituency was less convinced than everyone else.
By age: The 35-to-54 cohort — only 40% favorable. The 55-plus cohort — 100% favorable.
Here is what I wrote in that 2009 post, describing the purpose of the analysis:
“Stimulate the questions that might not have been asked in an effort to shape the investigative process to be more in line with an agent-based versus equation-based methodology.”
Agent-based versus equation-based. In those exact words. April 2009. Seventeen years before RAM 2025.
The constant: Surface metrics said 63% favorable. Structural disaggregation revealed five competing realities that each demanded a different response. The aggregate was technically accurate and structurally meaningless.
2014 — Forrester Validates: Independent Confirmation Without Citation
Five years after the JLARC review, the structural truth I had documented since 2007 was independently validated — by an institution that never referenced my work.
Virginia faced pressure to replace eVA with PeopleSoft’s Cardinal ERP procurement module. Classic technology-stack-progression thinking: ERP is more advanced, the organization should climb the stack, deploy the more capable system.
Virginia did something different. Instead of assuming the technology upgrade was the answer, they commissioned Forrester to independently assess whether the organization was ready for the change.
Forrester’s conclusion was unambiguous. eVA provided better functionality for procure-to-pay, better integration with non-Cardinal ERPs and with suppliers, and on a net present value basis, eVA’s total five-year costs were approximately 10% lower than Cardinal’s. Factoring in supplier fees and operational savings to local governments, eVA would have significantly lower costs and significantly lower risks.
Virginia killed the ERP migration. Saved millions.
This was Phase 0 thinking before Phase 0 had a name. The question wasn’t “which technology is more advanced?” It was “are we actually ready for this change, and will it produce better outcomes than what we already have?”
What makes this moment critical is that Forrester arrived at the same conclusion my agent-based analysis had already established — without ever citing it. They didn’t need to. The structural reality was too clear to reach any other honest conclusion. When the methodology is sound, independent analysis converges.
Sid Burgess, commenting on my 2014 post about the Forrester findings, captured the broader pattern: “I’m amazed at how many people say, ‘yeah, we know the ERP module isn’t very good but it’s all the ERP offers, so you know, we’re stuck.'”
eVA is still operational today. Twenty-five years after launch. Nearly 100,000 suppliers. $25 million in annual savings. Three technology generations. The methodology-first approach survived every platform change because the structural foundation never depended on any single vendor’s technology.
The constant: Surface metrics said ERP was the more advanced option. Structural analysis said the organization wasn’t ready — and the existing system was already producing superior outcomes. Forrester confirmed it independently. Virginia acted on it. Millions saved.
2026 — RAM 2025: The Methodology Meets Its Scale
What I did manually in 2009 — taking a single dataset and interpreting it through multiple agent lenses — is what RAM 2025 does at scale with artificial intelligence.
RAM 2025 takes a single question, a single dataset, or a single strategic challenge and runs it through multiple AI models simultaneously. Not to get multiple answers, but to expose what any single model’s lens would miss. Each model brings different training data, different analytical biases, different pattern recognition strengths. The disagreements between models are not noise — they are signal. They reveal the structural fault lines that a single-model consensus conceals.
The JLARC analysis – which came after RAM 1998- was the seed of RAM 2025 before RAM 2025 existed.
One survey. Five agent dimensions. Five different stories. The aggregate told legislators one thing. The disaggregation told them something far more complex and far more useful — if they were willing to look.
RAM 2025 operates on the same principle at exponentially greater scale: multiple models, multiple dimensions, multiple interpretive lenses applied to the same evidence base. The consensus matters. The disagreements matter more. Because disagreements between models — like disagreements between C-suite executives and front-line users in 2009 — are where the real diagnostic information lives.
The constant: Surface metrics from a single model describe. Multi-model structural disaggregation explains. Same principle. Exponential reach.
Why the Constant Matters Now
The documented 50–80% implementation failure rate across procurement technology has persisted unchanged through every technology generation since I started tracking it. Across ERP, eProcurement, S2P suites, RPA, and now AI — the failure rate has not moved.
The analyst ecosystem has never been more robust. Gartner publishes Magic Quadrants. Forrester publishes Waves. Spend Matters publishes SolutionMaps. The Hackett Group publishes benchmarks. And the failure rate holds.
Because every one of those frameworks measures capability at the statistical surface. None of them measure structural alignment at the operational depth where outcomes are actually determined.
That is what DND proved with RAM 1998, what JLARC revealed in 2009, what Forrester confirmed in 2014, and what RAM 2025 scales in 2026.
One constant. Four validations. Twenty-eight years.
Surface metrics are persuasive. Structural alignment determines outcomes.
That was true then. It is even more true now. And until the industry’s research methodology reflects that distinction, we will continue producing surveys, rankings, and quadrants that measure the wrong thing — while the real answers remain hidden beneath the statistical surface, waiting for someone to ask the right question.
The first one is still the same: What time of day do orders come in?
References
Jon W. Hansen is the founder of Hansen ModelsTM and creator of the Hansen MethodTM, a procurement transformation methodology developed over 27 years. He was a direct participant in Virginia’s JLARC review of the eVA program and is quoted in the official legislative records of the Commonwealth. He operates Procurement Insights, an independent blog with archives spanning 2007–2025.
-30-
Procurement Insights
From DND to JLARC to Forrester to the AI Era: One Thing Has Remained Constant
Posted on February 14, 2026
0
By Jon W. Hansen | February 2026
For the Busy Executive
Surface metrics describe. Structural alignment determines outcomes. That was true in 1998 when a single question — “what time of day do orders come in?” — took DND delivery performance from 51% to 97.3%. It was true in 2009 when I participated in Virginia’s JLARC legislative review and disaggregated a 63% favorable survey into five competing realities. Forrester confirmed it independently in 2014 when they recommended Virginia reject an ERP migration in favor of the methodology-first approach I had documented since 2007. And it is the foundation of RAM 2025 in 2026, where multi-model AI scales the same principle across twelve models and thousands of data points. Four eras. Four independent validations. One constant: the industry keeps mistaking statistical surface for structural truth — and the outcomes keep proving them wrong.
Read time: 8 minutes
The Constant
Across twenty-eight years, four technology generations, and four independent moments of validation, one principle has held without exception:
Organizations that measure structural alignment produce outcomes. Organizations that measure statistical surface produce reports.
This is the story of those four moments — not as retrospective, but as a pattern so durable that it has now outlasted every technology cycle the procurement industry has produced. If the pattern holds across DND in 1998, a state legislative review in 2009, an independent Forrester assessment in 2014, and multi-model AI validation in 2026, it is not an observation. It is a law.
1998 — DND (RAM 1998): The Question Nobody Thought to Ask
When SHL SystemHouse, then part of MCI, was managing the Department of National Defence’s MRO procurement platform in the late 1990s, they were delivering 51% next-day against a contractual requirement of 90%. They came to me and said what every struggling procurement organization says: we need to automate.
The first question I asked was: “What time of day do orders come in?”
The answer — most orders arrived at 4:00 PM — unraveled the entire failure chain. Service technicians were incentivized to maximize daily service calls, so they practiced sandbagging — holding all parts orders until end of day. Dynamic flux commodities that cost $100 at 9 AM were $1,000 by 4 PM. Parts shipped late missed customs windows. Delivery performance collapsed. And call close rates were horrific because the parts weren’t arriving to service the calls the technicians had rushed through.
Nobody inside procurement could see this. The failure was cross-boundary, involving four distinct agents: the procurement team, the service technicians, the SME suppliers, and the logistics and customs layer.
The solution wasn’t automation. It was understanding the full agent ecosystem and redesigning how those agents interacted. We built self-learning algorithms, integrated UPS directly into the system, and worked with Canada Customs to pre-clear shipments. Within three months: 51% to 97.3%. Over seven years: 23% consistent cost reduction. The buying group compressed from 23 companies to three.
The technology served the redesign. It did not drive it.
The constant: Surface metrics said this was a technology problem. Structural analysis revealed it was a behavioral problem invisible from inside any single department.
2009 — JLARC: Five Competing Realities Behind One Statistic
A decade later, Virginia’s Joint Legislative Audit and Review Committee launched a review of the Commonwealth’s eVA procurement program — one of the most successful eProcurement implementations in North American public sector history.
What I don’t often mention is this: I was asked to participate in that review process. And I’m quoted in the official legislative record of the Commonwealth of Virginia.
JLARC based part of their assessment on a stakeholder survey. The aggregate result: 63% believed eVA was a benefit to small, minority, and women-owned businesses.
I didn’t challenge the survey based on the number of responses. I challenged it because the aggregate concealed structural disconnects that no single number could reveal.
So I disaggregated the same data across five agent dimensions. What emerged was not one story. It was five.
By job title: C-level executives and vice presidents — 100% believed eVA was not a benefit. Front-line workforce — 100% believed it was. Same program. Same survey. Opposite conclusions.
By company size: Enterprise and large businesses — 100% support. Medium businesses — evenly divided. SMEs — 67% support. The strongest endorsement came from the segment with the least at stake.
By job function: Sales professionals — only 50% favorable. Marketing and creative functions — 100% favorable. The people doing the daily work saw friction. The people seeing the concept saw opportunity.
By gender: Women — evenly split on whether eVA benefited women-owned businesses. Men — 67% believed it did. The program’s own target constituency was less convinced than everyone else.
By age: The 35-to-54 cohort — only 40% favorable. The 55-plus cohort — 100% favorable.
Here is what I wrote in that 2009 post, describing the purpose of the analysis:
Agent-based versus equation-based. In those exact words. April 2009. Seventeen years before RAM 2025.
The constant: Surface metrics said 63% favorable. Structural disaggregation revealed five competing realities that each demanded a different response. The aggregate was technically accurate and structurally meaningless.
2014 — Forrester Validates: Independent Confirmation Without Citation
Five years after the JLARC review, the structural truth I had documented since 2007 was independently validated — by an institution that never referenced my work.
Virginia faced pressure to replace eVA with PeopleSoft’s Cardinal ERP procurement module. Classic technology-stack-progression thinking: ERP is more advanced, the organization should climb the stack, deploy the more capable system.
Virginia did something different. Instead of assuming the technology upgrade was the answer, they commissioned Forrester to independently assess whether the organization was ready for the change.
Forrester’s conclusion was unambiguous. eVA provided better functionality for procure-to-pay, better integration with non-Cardinal ERPs and with suppliers, and on a net present value basis, eVA’s total five-year costs were approximately 10% lower than Cardinal’s. Factoring in supplier fees and operational savings to local governments, eVA would have significantly lower costs and significantly lower risks.
Virginia killed the ERP migration. Saved millions.
This was Phase 0 thinking before Phase 0 had a name. The question wasn’t “which technology is more advanced?” It was “are we actually ready for this change, and will it produce better outcomes than what we already have?”
What makes this moment critical is that Forrester arrived at the same conclusion my agent-based analysis had already established — without ever citing it. They didn’t need to. The structural reality was too clear to reach any other honest conclusion. When the methodology is sound, independent analysis converges.
Sid Burgess, commenting on my 2014 post about the Forrester findings, captured the broader pattern: “I’m amazed at how many people say, ‘yeah, we know the ERP module isn’t very good but it’s all the ERP offers, so you know, we’re stuck.'”
eVA is still operational today. Twenty-five years after launch. Nearly 100,000 suppliers. $25 million in annual savings. Three technology generations. The methodology-first approach survived every platform change because the structural foundation never depended on any single vendor’s technology.
The constant: Surface metrics said ERP was the more advanced option. Structural analysis said the organization wasn’t ready — and the existing system was already producing superior outcomes. Forrester confirmed it independently. Virginia acted on it. Millions saved.
2026 — RAM 2025: The Methodology Meets Its Scale
What I did manually in 2009 — taking a single dataset and interpreting it through multiple agent lenses — is what RAM 2025 does at scale with artificial intelligence.
RAM 2025 takes a single question, a single dataset, or a single strategic challenge and runs it through multiple AI models simultaneously. Not to get multiple answers, but to expose what any single model’s lens would miss. Each model brings different training data, different analytical biases, different pattern recognition strengths. The disagreements between models are not noise — they are signal. They reveal the structural fault lines that a single-model consensus conceals.
The JLARC analysis – which came after RAM 1998- was the seed of RAM 2025 before RAM 2025 existed.
One survey. Five agent dimensions. Five different stories. The aggregate told legislators one thing. The disaggregation told them something far more complex and far more useful — if they were willing to look.
RAM 2025 operates on the same principle at exponentially greater scale: multiple models, multiple dimensions, multiple interpretive lenses applied to the same evidence base. The consensus matters. The disagreements matter more. Because disagreements between models — like disagreements between C-suite executives and front-line users in 2009 — are where the real diagnostic information lives.
The constant: Surface metrics from a single model describe. Multi-model structural disaggregation explains. Same principle. Exponential reach.
Why the Constant Matters Now
The documented 50–80% implementation failure rate across procurement technology has persisted unchanged through every technology generation since I started tracking it. Across ERP, eProcurement, S2P suites, RPA, and now AI — the failure rate has not moved.
The analyst ecosystem has never been more robust. Gartner publishes Magic Quadrants. Forrester publishes Waves. Spend Matters publishes SolutionMaps. The Hackett Group publishes benchmarks. And the failure rate holds.
Because every one of those frameworks measures capability at the statistical surface. None of them measure structural alignment at the operational depth where outcomes are actually determined.
That is what DND proved with RAM 1998, what JLARC revealed in 2009, what Forrester confirmed in 2014, and what RAM 2025 scales in 2026.
One constant. Four validations. Twenty-eight years.
Surface metrics are persuasive. Structural alignment determines outcomes.
That was true then. It is even more true now. And until the industry’s research methodology reflects that distinction, we will continue producing surveys, rankings, and quadrants that measure the wrong thing — while the real answers remain hidden beneath the statistical surface, waiting for someone to ask the right question.
The first one is still the same: What time of day do orders come in?
References
Jon W. Hansen is the founder of Hansen ModelsTM and creator of the Hansen MethodTM, a procurement transformation methodology developed over 27 years. He was a direct participant in Virginia’s JLARC review of the eVA program and is quoted in the official legislative records of the Commonwealth. He operates Procurement Insights, an independent blog with archives spanning 2007–2025.
-30-
Share this:
Related