We Were Wrong About Who Was Profitable — and We Had the Data to Prove It
What a Global Air Express Network Taught Us About Causal Costing, and Why That Lesson Lives in Everything Asher & Company Does Today
By Pedro San Martín — Asher & Company
The year is 2003. I am part of ALG Software — the Armstrong Laing Group — at that point operating under the banner of ALG (Global Intelligence), the consulting practice built on top of the HyperABC and Metify ABM platform. Altam Global Intelligence is what Asher & Company is today: the same methodology, the same people, a different name for a different era.
ALG had built HyperABC — the activity-based management platform that would eventually be acquired by SAP BusinessObjects and become what is today known as SAP PCM, the direct ancestor of Oracle EPCM. By 2003, the platform had already been deployed at some of the most operationally complex organizations in the world. One of them was the world's largest international air express company.
A firm that had, quite literally, invented the industry it dominated. Operating in 225 countries, 59,000 employees, 217 aircraft, 14,635 ground vehicles. Serving every corner of the planet.
And they had a problem that was invisible to every financial system they operated.
They believed their banking customers were unprofitable. They had a spreadsheet that said so. And they were starting to make commercial decisions — pricing, discounting, relationship investment — based on that belief.
The spreadsheet was wrong. Not wrong because someone made a calculation error. Wrong because it was built on the wrong question. It asked how many shipments does each customer send? when the question it needed to answer was what does it actually cost, activity by activity, to serve each customer?
Those are not the same question. The gap between them was costing the company millions of dollars a year in mispriced contracts. And it was about to get worse — because the competitive environment was intensifying, with rivals aggressively expanding their international operations after dominating their domestic markets.
What we did next — and how we did it — is the foundation of everything I practice at Asher & Company today.
The Battlefield Context
Before I explain what we found, I need to explain how we thought about the problem. Because the methodology matters more than the case. The case is history. The methodology is what you can use on Monday morning.
The framework we used was the CAM-I Cross — the conceptual model developed by the Consortium for Advanced Management International that became the intellectual backbone of activity-based costing. It looks simple. It is not.
The CAM-I Cross has two axes that operate simultaneously.
The vertical axis — the cost assignment view — traces money from resources (people, technology, facilities) to activities (the work those resources perform) to cost objects (the products, customers, and channels that consume the activities). This is the ABC chain that most people understand.
The horizontal axis — the process view — traces performance from cost drivers (what causes the work to happen) through activities (the work itself) to performance measures (how well the work is being done). This is the management dimension that most implementations ignore. It answers not just what does this cost? but why does it cost this, and what does that tell us about our processes?
Most companies that implement ABC only use the vertical axis. They trace costs from the GL to products. They call this "activity-based costing." It is a useful exercise. It is not the same as the full CAM-I model. What we built used both axes — and that is why it produced a result that changed commercial strategy, not just cost reports.
I. What the Old Model Was Doing Wrong
The company's original cost model had 30 cost pools and five allocation drivers. The primary driver was shipment count — the number of packages each customer shipped per period.
This seemed logical. It was not.
In air express logistics, shipment count is correlated with cost but not causally connected to it. What actually drives cost is a combination of weight, distance, customs complexity, pickup frequency, special handling requirements, and post-shipment service contact rate. A single 50-kilogram industrial component requiring customs declaration and special handling generates ten times the activity cost of ten 100-gram document envelopes picked up on a pre-scheduled daily route.
The model treated them identically — because both generated "shipments."
The result was a systematic inversion. Banking customers — who shipped lightweight documents, had predictable daily pickup patterns, rarely needed customs processing, and almost never called the service center — appeared expensive because their shipment count was high. Heavy manufacturers — who shipped infrequently but required complex customs handling, irregular pickups, and frequent service contacts — appeared cheap because their shipment count was low.
Management was looking at a profitability ranking that was the exact mirror image of reality. Every pricing decision, every contract renewal, every commercial strategy built on that ranking was pointing in the wrong direction.
This is what a cost model failure actually looks like. It is not dramatic. There is no single wrong number that alerts you. The model produces plausible-looking outputs. Margins look reasonable. The business continues. And underneath, value is being destroyed systematically, invisibly, at scale.
II. What ABM Revealed — and How We Built It
The activity-based model we built used 106 activities organized into 15 categories, from pickup and delivery through customs processing, international transport, customer service, sales, IT, HR, and finance. Each activity had a precise operational definition, a primary cost driver, and a data collection methodology.
The critical insight was not the number of activities. It was the choice of drivers.
For pickup and delivery, the driver was not shipment count — it was courier minutes per outbound pickup. A scheduled daily pickup at a bank headquarters consumed 8 minutes. An on-demand industrial pickup requiring route adjustment and weight verification consumed 35 minutes. Same "shipment." Different cost.
For customs processing, the driver was a complexity index. A document envelope required zero customs intervention. An industrial component required full declaration, classification lookup, and potential inspection. The cost differential was 40x.
For customer service contacts, the driver was inbound contact rate per account. Banking customers averaged 0.3 contacts per shipment. Manufacturing customers averaged 2.8 — tracking requests, proof of delivery queries, exception handling, damage claims. The cost differential was 9x.
When these activity costs were traced to customer segments, the profitability picture inverted completely.
III. The Scale Problem — and How We Solved It
Demonstrating ABM in one country is a proof of concept. Deploying it across 225 countries is an organizational transformation. The gap between those two things is where most ABM implementations fail.
We approached the scale problem as a methodology design problem, not a technology problem. The software worked. The question was: how do you make the implementation reproducible, fast, and locally owned — without losing analytical integrity?
The answer was standardization at the framework level, localization at the data level.
We developed a standardized chart of 106 activities — the same 15 categories, the same driver definitions, the same calculation logic — adaptable to any country's operational reality. We built a five-day training programme. We developed an API connecting local data systems directly to the calculation engine, automating the monthly cost update.
The result was a 16x compression in implementation time:
- First pilot: 4 months
- Second pilot: 4 months
- Third pilot (the dress rehearsal): 5 weeks
- Global rollout: 3–4 weeks per country, 1–2 weeks for smaller markets
From four months to five weeks. Not driven by better technology. Driven by better methodology design — the kind of design that comes from asking how do we make this reproducible? rather than how do we make this perfect?
By the late 1990s, 80% of all costs in the network were modeled using ABM. Twenty-five countries operated the full 106-activity model. Fifty more operated a simplified 50-activity version calibrated to their market complexity. The decision to use a simpler model in small markets was not a compromise — it was a design principle: deploy the right level of analytical sophistication for the decision complexity at hand. This principle is at the heart of how we design Decision-to-Value architectures at Asher & Company today.
IV. What This Produced — and Why It Still Matters
The commercial output of the ABM system was the Pricing Evaluation Tool — a pricing application that used activity cost data to calculate the appropriate price for any customer based on their specific shipment profile. For the first time, a sales manager could walk into a customer negotiation with a number causally connected to the cost structure of serving that customer.
The Pricing Evaluation Tool evaluated more than USD $500 million of revenue per year. Each one percentage point of margin improvement generated USD $5 million in additional profit. The total ABM system investment was USD $6–7 million.
But here is what I want you to take from this case:
The USD $5 million per percentage point is not the lesson. The lesson is why the old system was producing the wrong number — and what it tells us about how most organizations are still making profitability decisions today.
The old system used a single driver (shipment count) to allocate costs that were actually driven by a complex combination of weight, time, customs complexity, and contact rate. It produced a precise-looking number that was structurally wrong. Management trusted the number because it came from the general ledger. They were wrong to trust it.
This pattern — precise-looking numbers built on structurally flawed cost allocation — is not a 1990s problem. It is a 2024 problem. I see it in every engagement Asher & Company undertakes. The tools are more sophisticated. The data is more abundant. The fundamental error is identical: costs are being allocated using proxies rather than traced using causal drivers.
V. From Altam Global Intelligence to Decision-to-Value
When SAP acquired Armstrong Laing in 2006, the software moved to a new home. The methodology stayed with the people who had built it — and eventually became the foundation of what is today Asher & Company.
What we learned from implementing ABM at global scale across one of the most operationally complex organizations in the world is this: cost intelligence is not a reporting exercise. It is a decision architecture.
The difference between a cost report and a decision architecture is causality. A cost report tells you what was spent. A decision architecture tells you what caused the spending, what would happen if the drivers changed, and what action would produce the most valuable outcome.
In 2003, building that architecture required 106 activities, a multi-year programme, and a team that worked 15-hour days. Today, the same architecture can be built faster, at higher granularity, with AI-assisted driver identification and real-time updating — using Oracle EPCM as the cost foundation and modern EPM tools as the planning and decision layer.
The question the air express company was trying to answer — which customers are actually profitable, and why? — is the exact question our clients ask today. The tools have evolved from HyperABC to Oracle EPCM to AI-powered profitability models. The question has not changed.
At Asher & Company, we call the answer to that question Decision-to-Value: the direct, causal, traceable connection between the data your operations generate and the decisions that move your enterprise value. It is not a software platform. It is a way of thinking about cost and profitability that we have been refining since the days when we looked at a spreadsheet that said banks were unprofitable — and asked, for the first time, the right question.
What is actually driving this cost?
VI. The Data
Global network profile
| Metric | Value |
|---|---|
| Countries served | 225 |
| Delivery stations | 2,321 |
| Ground vehicles | 14,635 |
| Aircraft | 217 |
| Employees | 59,000 |
| Global TDI market share (current) | 43% |
| ABM system total investment | USD $6–7M |
| Revenue evaluated by pricing tool/yr | USD $500M+ |
| Profit per 1% margin improvement | USD $5M |
| Implementation speed improvement | 16x (4 months → 5 weeks) |
The profitability inversion — before vs. after ABM
| Segment | GL rank | ABM rank | Key driver of difference |
|---|---|---|---|
| Banks / financial services | 6th — least profitable | 2nd — highly profitable | Low weight, predictable pickup, minimal customs |
| Technology companies | 3rd | 1st — most profitable | High value, time-sensitive, document-heavy |
| Healthcare | 5th | 4th | Regulatory complexity offset by premium pricing |
| Retail | 4th | 3rd | Moderate complexity, volume-driven |
| Heavy manufacturing | 1st — most profitable | 5th — below average | Complex customs, special handling, high contact rate |
| Logistics intermediaries | 2nd | 6th — least profitable | High volume, extreme price sensitivity, high coordination cost |
ABM activity framework (106 activities, 15 categories)
| Category | Activities | Primary driver |
|---|---|---|
| A — Pickup & delivery | 12 | Courier minutes per stop |
| B — Local operations | 9 | Station throughput index |
| C — Trunking | 8 | Weight × distance |
| D — Customs | 11 | Shipment complexity index |
| E — International sorting | 7 | Units sorted per hour |
| F — International transport | 6 | Kgs × flight hours |
| G — Customer services | 14 | Inbound contact rate per account |
| H — Sales | 10 | Account management hours |
| I — Marketing | 6 | Campaign allocation |
| J — IT | 8 | System utilization index |
| K — HR | 5 | Headcount by activity |
| L — Customer accounting | 6 | Invoice count |
| M — Finance | 5 | Transaction volume |
| N — Management & admin | 9 | Overhead allocation |
VII. Lessons Learned
1. The most dangerous cost model is one that looks precise but is causally wrong. The spreadsheet produced a number for every customer. The number was wrong in a systematic, structural way that inverted the profitability ranking of the entire customer base. Precision in the wrong model is more dangerous than acknowledged uncertainty in a simple one.
2. The CAM-I Cross is not a diagram. It is a discipline. Most organizations implement the vertical axis of the CAM-I Cross and call it activity-based costing. The horizontal axis — cost drivers into activities, activities into performance measures — is what connects cost intelligence to operational management. Without it, you have a better allocation. With it, you have a decision tool. The difference is everything.
3. The driver choice is the model. Switching the primary driver from shipment count to weight × complexity inverted the entire customer profitability ranking. The model itself — the software, the activities, the cost pools — was almost irrelevant. The driver was the model. This is the most underappreciated insight in cost management: the analytical choices that determine what drives cost allocation are more consequential than the system that performs the calculation.
4. Methodology design is more valuable than software selection. The ability to reduce implementation time from four months to five weeks was not a software improvement. It was a methodology design improvement — standardized activity frameworks, structured training programmes, automated data interfaces. The tools that enable Decision-to-Value today are more powerful than HyperABC. But the methodology is the same one we designed at Altam Global Intelligence: start with the right question, trace causality from drivers to activities to cost objects, and make the model reproducible without losing analytical integrity.
5. Scale requires tiered sophistication — not uniform complexity. The decision to use a simplified 50-activity model in mid-size markets and a spreadsheet in small ones was not a compromise. It was a design principle. The goal of a profitability architecture is not maximum model complexity. It is optimal decision quality per unit of analytical investment.
6. The same question — twenty years later. Every engagement I do at Asher & Company begins with the same question we were trying to answer in 2003: which customers, products, channels, and operations are actually creating value — and which ones are consuming it invisibly? Oracle EPCM connected to AI-driven analytics can produce in weeks what took us years. But the question that makes the answer valuable is the same. Organizations that are not asking it — still relying on GL-allocated costs to make pricing, investment, and portfolio decisions — are making the same mistake that produced the inverted profitability ranking. With much more expensive consequences.
Pedro San Martín is Managing Partner at Asher & Company, a specialized advisory firm in Enterprise Performance Management, profitability architecture, and strategic finance across the Americas and Europe. The methodology behind Asher & Company's Decision-to-Value practice traces directly to the Armstrong Laing Group and Altam Global Intelligence, before ALG's acquisition by SAP BusinessObjects in 2006.
asher.company | Battlefield Lessons — real decisions, real consequences, no textbook answers.
