In today’s automotive industry, knowing what a component does is no longer enough. The bigger question is: what should it really cost? That question sits at the heart of Bengaluru-based Advanced Structures India (ASI), an engineering firm that is bringing a new level of intelligence and transparency to automotive cost engineering. Through its Cost Labs, powered by proprietary technologies xcPEP and xcPROC, ASI helps manufacturers uncover the economics hidden inside components, assemblies and manufacturing processes—from target cost analysis and competitive benchmarking to cost reduction and digital transformation.
Operating across six industry verticals worldwide, ASI combines engineering expertise, manufacturing intelligence, physical teardowns and AI-powered analytics to challenge assumptions that often become embedded in product costs over time.
For Pratik Shukla, CTO and Co-founder of ASI, the philosophy is straightforward: you cannot manage what you cannot see. And nowhere is that more relevant than inside the automotive industry's black-box components—the bought-out assemblies that arrive with a price tag, but often very little visibility into how that price was built.
The cost you cannot see is often the cost you cannot control, according to Shukla. Walk into the purchasing department of almost any automotive company and you will find a familiar pattern. Steel is costed. Plastic is costed. Machining hours are costed. Labour is negotiated down to the last rupee.
Then there are the black boxes: chargers, radar sensors, motor controllers and electronic control units, among other bought-out assemblies that arrive as finished products with a single part number, a single supplier quote and, often, very little visibility into what sits inside.
That is where a surprising amount of cost can disappear. ASI has made this area of decoding component costs its specialty. Its Cost Labs, powered by proprietary technologies xcPEP and xcPROC, work with leading manufacturers across six industry verticals worldwide to bring engineering depth into target costing, competitive benchmarking, cost reduction and digital transformation.
At the heart of ASI's approach is a deceptively simple question: what does this part really cost to make? Shukla believes the answer cannot come from a supplier quotation alone. It has to come from understanding the part—its materials, components, processes, manufacturing location, volumes and engineering choices—and then testing what happens when those variables change.
To demonstrate the point, ASI went beyond spreadsheets. It independently procured products from the market, opened them up and costed them component by component.
Three examples—a two-wheeler EV charger, a 77 GHz radar sensor and a motor controller—show just how much money can sit behind a seemingly innocuous part number. And sometimes, the biggest savings are hiding in plain sight.
The Black-Box Problem
Bought-out assemblies create a peculiar cost problem. The OEM buys them as complete units. The supplier quotes them as complete units. Over time, the annual negotiation becomes an exercise in adjusting last year's price rather than rebuilding the cost from the ground up.
Nobody necessarily does anything wrong. The problem is simply cost visibility. When the OEM does not know the cost structure inside an assembly, its negotiating position is inherently limited. And when the supplier has proprietary designs or specialized processes, the gap becomes even wider.
This is why should costing matters. It is not about second-guessing the supplier. It is about establishing an independent engineering view of what the product should cost under defined assumptions. The question goes all the way back to economist Ronald Coase's famous 1937 problem: what should a company make, and what should it buy?
Today, answering that question requires considerably more than instinct. So ASI opened the boxes.
Two Chargers. Same Job. Very Different Cost.
Proof is in the pudding as they say. In May 2025, ASI examined two electric two-wheeler chargers in detail from different OEMs with broadly similar specifications. The first surprise was outside the electronics.
Charger A used a four-piece extruded aluminium housing. Charger B used a two-piece moulded plastic housing. The difference was significant. Charger A required 55 assembly-level components. Charger B needed just 16. The simpler packaging architecture made Charger B's casing approximately 30% cheaper.
Then came the smaller decisions. Charger A carried 10 additional fasteners and an extra connector pair on its power cord, adding roughly 6% to assembly cost. Inside the charger, the same story continued. One design used a bridge rectifier. The other achieved the function using four individual diodes, at an estimated 40–70% lower cost.
Even the PCB laminate offered a potential saving. Charger A used FR-5, a laminate that costs 10–15% more, while B used FR-4 for an application where ASI's assessment indicated that FR-4 could survive the same duty cycle.
Then there was component size. Approximately 45% of A's board components were below 2 mm, compared with 39% for B. Smaller components can require more sophisticated and expensive placement capability. Individually, none of these decisions looks dramatic.
Together, they tell a very different story.
ASI identified an 11% reduction strategy for the charger PCB. For Charger A, the analysis identified a 9.3% saving opportunity. The lesson is not that one charger was "wrong."
It is that two products can meet essentially the same requirement while carrying very different cost structures. Granular comparison exposes the difference.
The Radar Sensor: When the Electronics Become the Cost
The second teardown moved into a very different category. In March 2025, ASI independently procured a 77 GHz automotive radar sensor, the kind of technology increasingly found in advanced driver-assistance systems. The sensor weighed just 118.8 grams.
Yet inside those 118.8 grams sat around 70 components distributed across two boards, including a copper antenna only 27–33 microns thick, with a 1–2 micron tin finish, built on a 100-micron PTFE substrate. This is where conventional costing starts to struggle.
When drawings or supplier data do not reveal enough detail, ASI physically analyses the material itself. In this case, material identification included FTIR and SEM/EDAX testing. The resulting should-cost model, based on Chennai manufacturing rates for the study period (January 2023) and an annual volume of 240,000 units, came to approximately ₹4,486 per sensor. The breakdown: material ₹3,695, conversion ₹259, overheads ₹532.
Here the numbers reveal the real story. Material represented approximately 82% of the modelled cost. But this was not primarily a story about aluminium, plastic or laminate. Those materials made up roughly 94 of the sensor's 118.8 grams, yet only a small share of its cost.
The real cost sat deeper inside the black box: the semiconductor devices, passives and other electronic child parts that make up the electronic BOM. This is an important shift in thinking. As electronic content increases, the cost battlefield moves away from the factory floor.
It moves towards the BOM and the sourcing desk. ASI's sensitivity analysis made that clear. Change one input at a time, see how much the total cost moves, and rank the variables accordingly. A 10% improvement in the negotiated cost of the child-part BOM could reduce the sensor cost by approximately ₹370 per unit—around ₹8.9 crore annually at the assumed volume.
By comparison, reducing conversion rates by 20% through a move to a lower-cost manufacturing location would save only around ₹52 per unit.
The message to some customers is blunt, to some very clear. Do not negotiate the machine hour when the money is sitting in the BOM.
The Motor Controller: Sometimes the Cost is in the Metal
The third teardown produced almost the opposite result. In March 2025, ASI compared motor controllers from two two-wheeler OEMs. Here, the electronics were not the biggest problem.
The packaging was. Controller A used a three-piece aluminium housing with walls 0.8 mm thicker than the competing design. The additional material increased housing weight by approximately 18%. Controller B's simpler two-piece architecture made its casing approximately 20% cheaper.
Then came the accumulated cost of complexity. Controller A used four additional fasteners, a heat sink 1.8 mm thicker and two steel end plates where B used one. The result was approximately 11% higher assembly cost.
The PCB told a similar story. A contained 161 components compared with 138 in B. About 62% of A's components were below 2 mm, compared with 50% for B. Again, none of these decisions necessarily represents poor engineering. But every additional component creates consequences.
One extra fastener seems insignificant. Multiply it by the joints, tolerances and assembly steps it creates—and suddenly that small engineering choice is costing money on every vehicle.
ASI's what-if modelling quantified individual levers. The packaging changes alone could reduce casing cost by approximately 20% and assembly cost by around 11%.
The thinner housing could recover approximately 18% of housing weight—subject, of course, to structural and validation requirements.
Taken together, and after engineering validation, the identified strategy represented approximately 7% potential reduction in controller cost. The electronics were doing their job. The box around them was simply more expensive than it needed to be.
The Forgotten Cost Hiding Inside Machine-Hour Rates
There is another black box that appears regularly in legacy programmes: the machine-hour rate. A machine-hour rate is not simply the cost of electricity, labour and maintenance, according to Shukla. Part of it can represent the fixed cost of the machine itself—depreciation and financing.
That creates an interesting problem. ASI has encountered cases where machines had been fully depreciated, sometimes years earlier, yet the quoted machine-hour rate continued to carry the same fixed-cost component. That does not automatically mean the supplier is overcharging.
If a supplier must invest in a new machine, finance it and maintain capacity, the fixed cost is real. But if the machine has already been fully depreciated and no replacement investment is required, the question deserves to be asked. Is a cost that no longer exists still being carried in the price? A good cost model makes that visible.
When a 'What-If' is a Question, Not a Recommendation
“Why are we paying for this?” It should be “Why is it there in the first place?”There may be a very good engineering reason for every one of them. And that is exactly where a what-if analysis earns its place. ASI's model does not tell the engineer what to change. It simply puts a number on the question:
“If we changed this, what would the cost become?” The engineer then asks the question that matters: “Can we make that change without compromising performance, safety, reliability or regulatory compliance?” If the answer is no, the saving disappears. And that is not a failure of should costing. It is the model doing its job.
In the charger study, only 41% of the identified ideas made it through validation and into implementation. The rest were rejected, ruled out or simply did not stand up to engineering scrutiny. That is how a credible cost model should work, says Shukla.
Sometimes, the right answer is not “we can save money.” Sometimes it is: “This part is already well engineered—and well bought.” Good cost engineering challenges assumptions. It puts a price on alternatives and gives engineers and procurement teams better questions to ask. But it never replaces engineering judgement.
From Chemistry to City-Level Economics
True should costing is not simply putting a price against "steel", "plastic" or "electronics". The material itself has to be understood. The questions include: What grade? What composition? Which manufacturer? Which region? What price in that market? What price in that quarter? What certified alternatives exist?
A polymer is not simply "plastic". It may be a base resin plus fillers, additives and reinforcements. A magnet is not simply "magnet material". Its cost depends on its precise rare-earth composition and supply chain.
For every part in these studies, ASI records an average of approximately 25 parameters, with around 5–8 typically driving the cost model.
And manufacturing economics are geographical. The same operation can carry very different economics in Chennai, Pune, Bengaluru, Shanghai or another manufacturing hub. That is why ASI's models use city-specific, time-stamped rates.
In the radar study, for example, the model used Chennai rates for the relevant period. Most importantly, ASI describes its model as a glass box. The formulas are visible. The assumptions are editable and the inputs can be challenged. Because a cost model is only as credible as its least-verified assumption.
Turning Weeks of Costing into Hours
There is, however, a practical problem. This level of costing is enormously data-intensive. Trying to do it manually for thousands of parts is simply not scalable. That is where technology becomes more than a convenience. That's why ASI developed xcPEP, its AI-powered costing platform, to bring speed to the depth of engineering analysis.
Its vision capabilities can interpret a circuit board, identify components and build a BOM in seconds. What traditionally could take 6–12 weeks to develop can be reduced to hours for a complete should-cost assessment, subject to the availability and quality of source information.
The system can then run sensitivity analyses, rank cost drivers and test individual what-if scenarios like change in a material, change a supplier price, change a manufacturing location, change in the volume, change the process.
The model recalculates the impact and exposes where the saving actually comes from. But there is an important caveat. AI does not magically know the cost of a component. Ask a generic language model to predict a price and it can produce an impressively confident answer.
That does not make the answer defensible. For ASI, AI has a role in costing only when it sits on top of primary manufacturing data, engineering logic and traceable assumptions. The objective is not to produce a clever number. It is to produce a number that can withstand scrutiny.
The Question Every Procurement Leader Should Ask
Here is a simple exercise ASI would challenge every automotive CEO, CFO, procurement head and engineering leader to try:
Count your cost models. Now take the 20 bought-out assemblies that account for the biggest chunk of your spending. For each one, ask a simple question: Do we know what this part should actually cost? Not what the supplier says it costs. Not what you paid last year. Not what another programme paid for something similar. Your own independent, engineering-based view of the cost.
For many manufacturers, the answer may be surprisingly uncomfortable. The good news is that building that capability does not have to take years. Twenty robust models in 12 months is a realistic starting point.
Use this sequence: Tear down. Model. Benchmark. Challenge. Validate. Implement. Then do it again.
The real advantage comes when this stops being a once-a-year cost-cutting exercise and becomes part of the organisation's DNA. Some manufacturers are already taking that step, building Cost Labs within their own operations so that cost intelligence sits alongside engineering, procurement and manufacturing—not somewhere outside the annual negotiation cycle.
That is where cost engineering is heading. The winners of the next decade will not necessarily be the companies that negotiate hardest. They will be the companies that walk into the negotiation already knowing what the part should cost—and why.
Opening the Black Box
The three studies discussed here were independently procured and analysed by Advanced Structures India. The objective was not to prove that one design is universally better than another.
It was to demonstrate what happens when a bought-out assembly is no longer treated as a black box. Once you can see the components, materials, processes, manufacturing economics and assumptions inside the box, the conversation changes. Procurement has a stronger negotiating position. Engineering can see which changes are worth investigating. Finance can see where the real cost sits.
Management can make better make-or-buy decisions. And suppliers can be challenged with facts rather than assumptions.
That is ultimately what a Cost Lab is designed to do. Turn cost from a negotiated number into an engineered number. And sometimes, the biggest saving is not hiding in the factory at all. It is hiding inside the box nobody opened.