Sanjib Das: Stop Stocking on Gut Feeling (Part 1)

(Part 1 of our series with reliability engineer Sanjib Das)

At SPARROW, we regularly sit down with MRO and reliability practitioners from around the world: engineers, managers and leaders who have spent decades in plants, warehouses and control rooms, not in slide decks.

We do this deliberately.
The most persistent problems in spare parts management rarely come from a lack of theory. They come from what happens on the ground: legacy decisions, cultural habits, imperfect data, and the daily trade-offs between risk, cost and uptime.

In previous interview series, experts like Conrad Greer drew on hands-on experience across oil & gas, utilities and rail to unpack the systemic issues behind spare parts performance. Andrew Jordan took a different angle, exposing how poor master data and organisational neglect quietly undermine even well-funded MRO strategies. This new series continues that tradition, but through the lens of reliability engineering.
‍

Meet Sanjib Das and why we wanted this series

When you talk to Sanjib Das, you quickly realise he is not a “spreadsheet reliability engineer”.

He started as a mechanical maintenance engineer in a chemical plant, then moved into reliability at ExxonMobil in Singapore in 2008, just as the company was building six or seven new plants on the same site. Reliability was built in “from the very beginning from the design stage”, long before it became fashionable.

There was one condition to become a reliability engineer:

“You don’t have to be expert on every aspect. However, you need to have knowledge on each aspect. Meaning you have to be in the maintenance field at least 5 to 7 years. And you have to know the equipment, the process, and how to get the answer from the expert… You have to be able to ask the right question to the right expert, otherwise you just become a postman.”

Over the next 20+ years, Sanjib:

• Ran RCM programmes and root cause analyses across complex assets.
• Was handed responsibility for preservation and spare parts strategies for megawatt motors and other critical equipment.
• Got asked to clean up USD 30 million of stock that hadn’t moved in 10 years.
• Built and refined an Excel-based spare parts criticality and stocking tool that several organisations adopted as their standard.
  

The heart of his approach? Strip emotion out of spare parts stocking and replace it with a simple, explainable mathematical model:

“My job as a reliability engineer is to stop the emotion and make the process in a mathematical way… When it is math, everybody understands.”

This is exactly where our worlds meet. At SPARROW, we build AI-based spare parts planning and stocking tools with that same goal: make decisions transparent, repeatable, and grounded in risk and data – not in whoever shouts loudest.

What this 5-part series will cover

This conversation with Sanjib Das turned into a five-part series on how reliability engineers can take control of spare parts, from first principles to AI-driven decisions.

Here is what to expect:

1. Stop Stocking on Gut Feeling
   How to move from “everything is critical” debates to a tunable, 1–5 scoring model that weighs asset criticality, lead time, failure behaviour, custom vs. standard parts, and cost, with a worked example.
2. The Hidden Cost of “Just in Case” Inventory Why spare parts are not free: shelf-life limits, preservation tasks (like rotating megawatt motors 90° every three months), and the real carrying cost of overstock.
3. Planned vs Unplanned: Spare Strategies that Reduce Downtime
   How a turnaround overran by two weeks because of missing spares, and how better stocking rules, reorder points and vendor vs in-house strategies compress downtime.
4. One Global Spare vs Three Local: When to Pool Inventory
   When it makes sense to stock high-value, long-lead items globally rather than at every site, and the governance you need so plants trust the policy.
5. Excel Is Static. Failures Aren’t. Make Spare Decisions Dynamic.
   How to take Sanjib’s Excel model and connect it to your CMMS (SAP, Maximo, etc.) so stocking rules update with real failure data. And where AI fits in.
   
   

Let’s dive into Part 1.

Why “everything is critical” is costing you money

Most plants still decide what to stock by committee: maintenance, reliability, operations, sometimes procurement. Everyone has a list; everyone has a horror story.

As Sanjib puts it:

“Everything is important for someone. If you ask the mechanical engineer: ‘I want to keep this shaft… this is a million-dollar item.’ If you ask the electrical engineer: ‘I need this, I need this.’”

Without a structure, decisions drift to:

• Emotion: the last painful failure or near-miss.
• Anecdote: “We once lost two weeks because of some little spares.”
• Politics: whose opinion carries more weight in the room.
  

The result is familiar:

• Bloated MRO inventory: Analysts estimate that 40–60% of MRO inventory in manufacturing is excess, obsolete, or very slow-moving.(GEP)
• Dead stock: Even well-run manufacturers see 20–30% of inventory become dead or obsolete over time.(amplio.com)
• High carrying cost: Industry benchmarks put annual carrying cost at 20–30% of the inventory value, meaning a EUR 1 million spare parts store can cost up to EUR 300,000 per year just to hold.(CAI)
  

And yet, despite this, many reliability engineers still struggle to say “no” when someone asks to stock another “critical spare”.

Sanjib’s answer was to take the argument away from personalities and put it into a transparent scoring model.

A simple scoring model that changes the conversation

Sanjib’s tool is not a black-box optimiser. It is deliberately simple and explainable.

At its core, it combines five groups of factors, each scored on a 1–5 scale:

1. Asset criticality
   
   • How bad is it if this asset fails?
   • Does it create a safety or environmental risk? A major production or customer impact?
     
     
2. Lead time and supply risk
   
   • How long does it take to get a replacement?
   • Is it off-the-shelf, made-to-order, or custom-engineered?
   • Are there single-source suppliers?
     
     
3. Failure behaviour
   
   • Is it a time-based degradation (e.g. wear) where you have advance warnings?
   • A random failure where you get little or no signal?
   • A bathtub curve / early-life issue?
     
     
4. Part nature and commonality
   
   • Is this a unique part, or does it serve multiple assets?
   • Does the same part sit on assets with different criticalities?
     
     
5. Cost and consequence
   
   • Direct part cost.
   • Downtime impact (lost production, penalties, overtime).
   • Safety, environmental, or media consequences.
     

Each factor gets a value between 1 and 5 (or 1 and 3 for some organisations), with clear definitions agreed up-front. The scores then roll up into a simple index:

“I put some logarithmic factors from one to five… based on the organisation you can tweak it, depending on the needs and budget.”

You can then define decision rules such as:

• High score → High critical spare: stock locally with defined minimum and reorder point.
• Medium score → Medium critical: stock locally or regionally depending on cross-site logistics.
• Low score → Low: do not stock; rely on supplier lead time.

The key is that people are no longer debating feelings; they are debating inputs:

• “Is this asset really safety-critical?”
• “Do we really see random failures here, or are they time-based?”
• “Is lead time genuinely 16 weeks, or is that a one-off story from a decade ago?”

Once those inputs are agreed, the outcome is automatic.

A worked example: same part, different rules

One of the useful aspects of Sanjib’s model is that it automatically handles context – the same part used on different assets can have different stocking rules.

“The same mechanical seal for one pump is critical, for another pump it is not critical, based on the asset criticality… In my calculation it factors this in as well; then it gives you the stock.”

Consider a simplified example with a mechanical seal used on two pumps:

On Pump A, a critical feed pump, failure would shut down the plant. Despite the same lead time, failure behaviour and part commonality, the high asset criticality and costly downtime push the seal into a high-critical category, justifying local stock with defined minimums and reorder points.

On Pump B, a utility pump, the same seal has a much lower impact if it fails. With identical lead time and failure behaviour but lower consequence, the total score drops to medium critical, making a shared or pooled stock sufficient.

The part has not changed — the context has.
That is exactly what Sanjib’s model is designed to capture.

You get:

• Seal for Pump A → High: stock locally, define minimum 2, reorder point 3.
• Seal for Pump B → Medium: consider 1 unit in common stock or regional pool.

You no longer argue whether the seal “is critical”; you see that its criticality depends on where it is installed.

This is exactly what Sanjib built into his Excel tool:

“My tool… calculates also: I have 10 similar pumps, how many do you keep in stock, what should be your minimum, your reorder point… because the same parts could be high in one place, medium in another, low somewhere else.”

Dealing with imperfect data (and still making progress)

A common objection is: “We don’t have enough data to score all this properly.”

Sanjib has heard that many times and built the tool to cope.

“Sometimes there are questions: ‘I don’t have failure history.’ Okay, use your engineering judgement… If you don’t have data, use this. If you have, use the field data. Then it becomes an excellent tool.”

His practical approach:

• Asset criticality: If you have not done a structured criticality analysis, do a fast-track version. He has done this “a thousand times” for different organisations with slightly different parameters.
• Failure data: Where CMMS data is missing, use standards like ISO 55000 guidance or military handbooks as a baseline for typical failure modes and rates.
• Lead time: If you do not have exact supplier data, use buckets (“same day”, “within a week”, “within a month”, “> 3 months”) and assign scores accordingly.
• Default assumptions: Document guidelines for what score to use when information is missing, so engineers do not “inflate the data” to get their favourite outcome.

This is also where external benchmarks help in the conversation. If your plants suspect they are “not that bad”, you can show that:

• Studies suggest 50–60% of MRO inventory is excess, obsolete, or slow-moving in many manufacturers.(GEP)
• A world-class level for slow-moving and obsolete parts is around 10% of spare parts stock.(Fiix)

In other words: the default state of the industry is not efficient, so it is worth acting even with imperfect data.

How SPARROW.Plan automates this thinking

What Sanjib built manually in Excel is the same reasoning engine we embedded into SPARROW.Plan. His method starts by weighing a small set of stable reliability factors, like asset criticality, lead time, failure behaviour, part commonality and consequence, and translating them into clear stocking decisions. SPARROW.Plan automates exactly this logic. It continuously pulls updated information from the customer’s ERP (SAP, Maximo, Infor), reconciles it with harmonised master data from SPARROW.Clean, and applies a transparent scoring model to recommend stocking levels, reorder points and pooling opportunities.

Where engineers once had to revisit spreadsheets, SPARROW.Plan recomputes the impact of every new work order, updated failure history, changed supplier lead time or modified bill of materials. Instead of debating spreadsheets, maintenance and reliability teams get real-time, risk-based guidance, the same decision logic Sanjib advocates, but applied consistently, automatically and at scale.

Selling the model internally: from resistance to relief

Introducing a scoring tool will trigger push-back, especially when the result says: “Your favourite spare isn’t critical.” Sanjib is upfront about this:

“After implementing the tool, some people will realise that some spare parts they thought were critical are not in the list. So definitely you will get the push.”

But he also notes that the resistance is often short-lived:

“Because of the nature of the tool (it is math, clear and crystal clear) even a technician can understand… You will face challenge, but it’s not a big hurdle.”

What helped in his experience:

1. Start with the why
   
   • Explain that the goal is not to cut inventory for its own sake, but to put money where risk really is, and remove emotion from the process.
   • Show the scale of the problem: for many plants, 20–25% annual carrying cost means the value of a spare part roughly doubles every 4–5 years if you just let it sit on the shelf.(Reliable Plant)
     
     
2. Walk through the logic with real examples
   
   • Take a few contentious spares and run them through the tool live.
   • Ask engineers to challenge inputs, not outputs: “Should this really be a 5 on safety? Do we have evidence?”
     
     
3. Give clear guardrails
   
   • Document what each score means.
   • Define who is allowed to change scoring rules (typically a cross-functional team).
     
     
4. Make management own the risk
   
   • The tool exposes the risk; management decides whether to stock or to accept it. This moves conversations from “I think” to “We are explicitly accepting this risk profile.”

In one organisation, Sanjib’s tool helped them tackle USD 30–40 million of stock that had never been used in 10 years. The value was not only in cash released, but in freeing up warehouse space, reducing preservation workload, and clarifying which spares truly mattered.

Why this matters for reliability, not just finance

It would be easy to frame this purely as a cost story. Sanjib does not.

For him, spare parts are deeply intertwined with reliability and risk:

“Spare part strategy is not only mitigation from the RCM, it is one of the mitigations… You have other mitigations like procedures, training, instrumentation thresholds, periodic replacement, monitoring… but spare parts are included in reliability as a whole.”

Good stocking decisions:

• Reduce consequences of unplanned failures: You cannot stop every failure, but you can compress downtime “from two weeks to three days” by having the right spares ready.
• Protect safety and environment: by prioritising spares for high-consequence failure modes.
• Stabilise operations: so you are not firefighting unexpected stockouts in the middle of a turnaround.

Conversely, poor stocking strategies hit you from both sides:

“It will increase your operational costs because you do not have the right spares or you do not have the spares at all… On the other side your warehouse is stocked with a lot of unnecessary things… This is all our money and after 5 years warranty is gone.”

The scoring model gives reliability engineers a defensible way to balance these tensions.

Where this leads next (and how AI fits in)

In later posts in this series, we will look at how Sanjib’s framework plays out in practice:

• Part 2 – The Hidden Cost of “Just in Case” Inventory
  
• Part 3 – Planned vs Unplanned
  
• Part 4 – One Global Spare vs Three Local
  
• Part 5 – Excel Is Static. Failures Aren’t
  

For now, the takeaway from Part 1 is simple:

“When it is math, management understands… It is crystal clear, no ambiguity.”

Do your spare parts conversations still revolve around whose story is most convincing? It may be time to put some numbers on the table.
To understand what this could mean for your own organisation, run your numbers in our Cost Savings Calculator. The calculator lets you explore different scenarios based on your inventory size and see where savings could realistically come from. As an example, a company with an inventory of 50,000 spare parts can save up to 6 Million Euros over 3 years.