Reading a plant in one site visit
You get one day, a safety briefing, a walk that someone else has planned, and a lunch. What you leave with determines everything that follows. Most consultants leave with impressions. This lesson is about leaving with data.
On this page · 16 sections
- 01The one idea
- 02The parts — what to ask for, before you arrive
- 03First principles — why the average lies and the deviation does not
- 04What must be true
- 05The quantitative anatomy — the walk, in order, with what to look at
- 06What goes wrong here — how day one fails
- 07Scenarios from the field — ask for the register, not the improvement plan
- 08The numbers that matter — the day-one scorecard
- 09Build the model — the day-one readiness score
- 10Upgrade paths — the three questions that find the backlog
- 11The frontier, and who is running it
- 12What’s changing now
- 13Check yourself
- 14Mini case — the plant that could not produce a heat balance
- 15Go deeper
- 16Carry forward
§ 01 The one idea
A site visit is not an inspection. You are not going to spot a problem by looking at a kiln that a hundred people have looked at every day for twenty years. A site visit is a data-collection exercise with a walk attached, and the walk’s real function is to establish that you are worth talking to.
So the visit has two products. The first is a set of documents that let you size opportunities in ₹/tonne once you are back at a desk. The second is a judgement about which of the plant’s own people already know where the money is — because on almost every site, somebody does.
The plant knows its process infinitely better than you do and will always win a technical argument. What the plant usually does not have is the ₹/tonne translation, the comparison against other plants, and permission to say certain things out loud to head office. Those three are what you bring. A day-one plan built on out-technicalling the plant fails; one built on translating and comparing works.
§ 02 The parts — what to ask for, before you arrive
Send this list a week ahead. What arrives, what arrives late and what never arrives are all findings.
| Ask for | What it lets you do | If it does not exist |
|---|---|---|
| The monthly ₹/tonne cost sheet, 24 months | Rebuild the stack (F3) on the plant’s own numbers rather than the parent’s, and separate imported cost moves from operational ones. | Almost always exists. If it is not shared, the constraint is trust or authority, not data — escalate rather than reconstruct. |
| Monthly heat and power balances | Locate the losses instead of guessing. Everything in C5 and C7 depends on this. | A plant that cannot produce a monthly heat balance does not have one. That is itself the first finding, and it usually means the last one was done by a vendor during commissioning. |
| 12 months of DCS trends kiln feed, fuel rate, ID fan, cooler, free lime | See the variability, not the average. Most opportunity in a mature plant is in the spread, not the mean. | Rare — the historian exists. If access is refused, ask for weekly averages and standard deviations instead. |
| Lab data: LSF and free lime, with standard deviations | Judge control quality. The mean tells you the target; the deviation tells you whether they hit it. | A plant that reports means without deviations is not managing variability, which is where the fuel is. |
| Kiln stoppage log, 24 months | Reliability is a cost line in disguise. Every stoppage is a reheat and a quality excursion. | If the log lives in a supervisor’s notebook, there is no reliability programme. Say so gently and early. |
| Maintenance shutdown history and the deferred list | See §10. The deferred list is the upgrade backlog, already prioritised by people who know the plant. | Ask verbally. It always exists in someone’s head even when it is not written down. |
| Fuel GCV certificates and the fuel mix by month | Fix the ₹/’000 kcal conversion for this plant (F4 §03) rather than carrying ₹1.90 from the last engagement. | Always exists — it is a payment document. |
| The CII award submission or the energy-conservation register | See §07. This is the single highest-value document on the list. | Its absence is a finding, and a large one. |
| CCTS baseline and notified intensity target | Carbon is a compliance line from March 2026. This constrains fuel and clinker-factor decisions. | New enough that some plants are still assembling it. Ask who owns it. |
§ 03 First principles — why the average lies and the deviation does not
The single most useful statistical idea on a site visit is that a cement plant is a system that is punished for variability, not for its mean. Understanding why turns a vague instinct into a quantified argument.
The kiln has a hard constraint: free lime must stay below about 1.5%, or the cement fails on strength. Free lime falls as you burn harder and rises as you burn softer. So the plant sets a fuel rate that keeps free lime under the limit on its worst hours, not on its average hour. The target is set by the tail of the distribution, and everyone pays the fuel bill for the tail all the time.
Suppose free lime averages 0.9% with a standard deviation of 0.35. To keep 95% of hours below 1.5% you need the mean about two deviations below the limit — 1.5 − 0.70 = 0.80%, so the plant is already running slightly harder than its average suggests.
Now halve the deviation to 0.18 through better raw-mix control and kiln stability. Two deviations is now 0.36, so the same 95% confidence permits a mean of 1.14% instead of 0.80% — a third of a percentage point of free lime, which is worth roughly 10–15 kcal/kg of heat rate at typical burnability.
You did not change a single piece of equipment. You bought heat rate with control quality. This is the mechanism behind a large share of the zero-investment items in the registers of A7 §09, and it is why §02 asks for standard deviations rather than averages. The 10–15 kcal/kg figure is an order-of-magnitude estimate from the free-lime-to-heat-rate sensitivity, not a measured constant — establish the sensitivity on the client’s own data before quoting a number.
The same logic runs everywhere in the plant. Cement mill Blaine, kiln feed rate, coal fineness, cooler grate speed — in each case the setpoint is chosen to survive the bad hour, and every hour pays for it. Ask for the spread of every number you are given the average of.
§ 04 What must be true
A site visit produces impressions by default. Turning it into evidence requires conditions that have to be set before you arrive, because most of them cannot be fixed on the day. These are the ones that decide whether you leave with data.
| What must be true | Why | Target | How you secure it |
|---|---|---|---|
| 1. You have the data before the walk, not after | What you see is only interpretable against what the plant records. A walk without the numbers produces anecdotes; the numbers without a walk produce a spreadsheet nobody believes. The order matters because the walk is where you test the numbers. | request two weeks out |
A written data request before travel, with a named owner. What arrives, how fast, and in what form is itself the first finding — a plant that cannot produce a heat balance in two weeks does not have one. |
| 2. You look at deviation, not at averages | A kiln is set by its worst hour. Monthly averages hide exactly the thing that costs money, and they are what every plant will hand you first. | hourly series, not monthlies |
Ask for the raw hourly logs for one representative month. The refusal, if it comes, tells you more than the data would have. |
| 3. Every number you take away has a basis attached | A figure collected on site without its denominator, boundary and period is unusable by the time you are back at a desk, and you cannot go back and ask. | — | Write the basis next to the number as you take it. This is the single discipline that separates a usable site visit from a wasted one, and it costs nothing but attention. |
| 4. You see the improvement register, not the improvement plan | A plan is what the plant intends and is written for an audience. A register is what the plant has already done, with the saving, the cost and the date — and it contains the plant’s real improvement rate and its real appetite. | — | Ask for the last three years of completed items. If no register exists, the plant has no institutional memory of its own improvements, and that is both the finding and the first recommendation. |
| 5. You have talked to someone who runs a shift | The control room knows which interlocks are bypassed, which instrument is not trusted, and which standing instruction contradicts the procedure. None of that is in any document. | — | Ask for fifteen minutes with a shift in-charge, without the plant head present. Whether that is possible is itself diagnostic. |
| 6. You know what is publicly recorded about the plant before you arrive | A plant’s capacity, its verified emissions, and in India its notified emission-intensity target are now matters of public record. Arriving without them wastes the visit and costs credibility. | — | See §11. Thirty minutes of public-record work before travel changes what questions are worth asking on the day. |
The failure mode of a site visit is not getting the wrong answer. It is coming away with an impression that cannot be defended three weeks later when a plant head disputes it. Rows 1, 3 and 4 exist so that every claim in the final document has a number, a basis and a source behind it. Rows 2, 5 and 6 exist so that the numbers are the right ones.
§ 05 The quantitative anatomy — the walk, in order, with what to look at
| Stop | Look at | Ask |
|---|---|---|
| The mine face | Bench height, overburden, whether the face is in good rock or chasing grade. Haul road condition and one-way distance. | “What is the current strip ratio, and what does the mine plan say it will be in five years?” The answer sets the raw material cost line for the plant’s remaining life. |
| The stacker-reclaimer | Pile geometry and whether the reclaimer is actually cutting the full face. Number of piles. | “What blending ratio does this achieve, and how do you know?” Pre-blending is where raw-mix variability is cheapest to remove. |
| The raw mill and the kiln feed silo | Mill running hours against kiln hours. Whether the silo is a blending silo or a storage silo — they are different machines with the same shape. | “What is the LSF standard deviation into the kiln, and what was it into the silo?” The ratio is the silo’s blending efficiency. |
| The preheater tower | Walk up it. Look for coating and build-up at the bottom stages, and at the number of cyclone stages against the plant’s age. | “Where do you get build-up, how often, and what do you do about it?” Volatile cycles (C10) are the usual answer and nobody volunteers them. |
| The kiln hood and the cooler | Secondary air temperature and cooler exhaust temperature. Grate loading — CII design is <45 TPD/m² and overloaded plants run 50–70. | “What is the cooler’s recuperation efficiency, and when was it last measured rather than calculated?” |
| The cement mill | Whether there is a roller press or pre-grinder. Separator generation. Product Blaine against the standard’s requirement. | “What Blaine do you grind to, and what does the standard need?” Over-grinding is common, invisible and expensive. |
| The packing plant and the truck yard | Truck turnaround time. Queue length. Whether bulk loading exists and is used. | “What is your direct-dispatch percentage?” ACC disclosed 52% in Q1 FY27 — it is a real, tracked and comparable metric. |
| The control roomthe most important stop | How many alarms are standing. Whether the operator is running on auto or manual. Whether anyone looks at the trend screens. | Nothing at first. Watch for ten minutes. A control room running on manual with fifty standing alarms tells you more about achievable heat rate than any document on the list. |
§ 06 What goes wrong here — how day one fails
| The failure | What to do instead |
|---|---|
| Accepting the planned tour. The route is chosen to show you a working plant, and it works. | Ask for two specific unplanned stops: the control room for ten silent minutes, and the packing plant at shift change. Both are cheap to grant and neither is on the tour. |
| Asking questions you could have answered from the filings. Capacity, utilisation, product mix, EBITDA/t. | Arrive with the F3 stack already rebuilt. Then your questions are about deviations from it, which signals you did the work and gets much better answers. |
| Interviewing only the plant head. | The plant head knows the strategy. The kiln operator on nights knows where the fuel goes, and the shift electrical engineer knows which motor has been oversized since commissioning. Ask for twenty minutes each, without the plant head present, and say why. |
| Taking averages and leaving. | Every average you accept without its deviation is a number you cannot use (§03). |
| Treating a refusal as an obstacle. | It is data. Distinguish cannot from will not: “we don’t have that” is a capability finding, “I’d need to check” is an authority finding, and they lead to different recommendations. |
| Missing the shutdown window. Sizing a retrofit without asking when the plant can next stop. | Ask on day one (§10). A ₹20 crore project with a nine-month wait for a shutdown slot is a different proposition from the same project with a slot in six weeks. |
| Confusing the plant with the decision. | Two-thirds of India’s new capacity is split grinding units. The economics you are diagnosing may be set by a grinding unit three states away, and nobody at this plant will mention it. |
§ 07 Scenarios from the field — ask for the register, not the improvement plan
Indian plants that enter the CII National Award for Excellence in Energy Management submit a document that lists every energy project they did, with investment and annual saving, project by project, for three years. It is written for a jury, so it is complete and it is quantified.
What those registers actually contain:
- JK Cement, Muddapur: 149 projects over FY21–23 for ₹2.21 crore — an average of about ₹1.5 lakh and 0.034 kWh/t per project. The single largest item in three years is 10.6% of the total.
- Chettinad Cement, Kallur: 73 projects for ₹0.88 crore, returning ₹11.1 crore a year. Average project cost ₹1.2 lakh.
- Orient Cement, Devapur FY23: nine items with printed SEC impact — 0.434, 0.141, 0.039, 0.019, 0.015, 0.015, 0.014, 0.005, 0.005 kWh/t. Eight of the nine are under 0.15. Four of the nine cost nothing.
- JK Cement, Nimbahera FY21: six projects, ₹8.5 lakh invested, ₹108.2 lakh a year saved. Sub-one-month payback, no capital project among them.
If your client has one of these, you have their whole improvement history in a spreadsheet on day one. And if they do not have one, that absence is your first finding: a plant with no register does not know what it has already tried, which means it will propose things it has already done.
The CII award archive has been intermittently offline. Cite these as “CII National Award submission, [plant], [year]” and keep local copies of anything you rely on. A citation that cannot be retrieved is not a citation.
§ 08 The numbers that matter — the day-one scorecard
Twelve numbers. If you have all twelve by the end of day one, you can size the engagement from a desk.
| Number | India reference | Basis to confirm | Your plant |
|---|---|---|---|
| Thermal energy | 740 avg / 670 best | kcal/kg of clinker. CII v7.0, May 2025. | — |
| Electrical energy | 70–80 | kWh/t of cement, and at what clinker factor. | — |
| Clinker factor | 0.68 | CRISIL FY25. Best comparable disclosure: Ambuja 0.637. | — |
| Kiln fuel cost | ₹1.66–1.95 | Per ’000 kcal. From the cost sheet, not from a rule of thumb. | — |
| Power cost | ₹5.6–6.0 | Per kWh, blended. Ask the green share separately. | — |
| Total cost | ₹4,004 | Per tonne of cement, grey basis. Confirm the denominator (F3 §05). | — |
| EBITDA | ₹1,005 sector | Per tonne. Q1 FY27 coverage aggregate. | — |
| Utilisation | ~72% | Of what capacity — clinker or grinding? | — |
| Lead distance | 249–445 km | Primary or full cement lead? Not comparable otherwise. | — |
| Free lime, mean and deviation | 0.5–1.5% | The deviation is the number that matters (§03). | — |
| Cooler grate loading | <45 design | TPD/m². CII-benchmarked plants run 41–47; overloaded plants 50–70. | — |
| Days to next shutdown | — | And what is already scheduled into it (§10). | — |
§ 09 Build the model — the day-one readiness score
The ₹60/t reference is deliberately conservative: it is roughly what the documented zero-and-low-investment registers in A7 §09 deliver, not what a capital programme delivers. Use it to establish a floor you can commit to, then argue for the capital case separately.
§ 10 Upgrade paths — the three questions that find the backlog
Add three questions to the day-one list. They take four minutes and they hand you a prioritised upgrade list assembled by people who know the plant better than you ever will.
| Question | Why it works |
|---|---|
| “When is the annual shutdown?” | Every tier-3 and above intervention is gated by it. A project is not expensive or cheap in the abstract — it is in this window or next year. This single date reorders your entire recommendation. |
| “What is already scheduled into it?” | Tells you the plant’s own priorities, the crane and contractor availability, and where your proposal would have to fit. It also tells you what is not negotiable. |
| “What was deferred from the last one, and why?” | This is the answer that matters. The deferred list is the upgrade backlog, already prioritised by people with full information, and each item comes with a reason for deferral — money, time, or a contractor who did not turn up. Two of those three are things a consultant can actually change. |
The deferred list is usually better than anything you would have generated. What the plant lacks is not ideas; it is the ₹/tonne translation and a route to a decision-maker. That is the whole job.
§ 11 The frontier, and who is running it
This is the one part of the course where something genuinely new has arrived, and it changes how a first engagement starts. A consultant in 2026 can know a named plant’s capacity, its kiln count, its audited CO₂ emissions and, in India, its legally notified emission-intensity target — before making contact with the company. None of that was available five years ago. All of it is free.
| Status | What | Who publishes it, and where | What you actually get |
|---|---|---|---|
| LIVE | Plant-level capacity and configuration | Global Energy Monitor — Global Cement and Concrete Tracker. Free, machine-readable. July 2026 release, updated annually. | 3,884 plants, 171 countries, 6.1 bn t/yr cement and 3.7 bn t/yr clinker capacity. Per plant: owner and parent, location, operating status, clinker and cement capacity, kiln count, production method. Read the caveats: locations are flagged exact or approximate, capacity is theoretical maximum, and where only a daily figure existed annual capacity was taken as daily × 365. |
| LIVE | Audited, legally mandated plant emissions | EU Union Registry. Verified emissions per installation per year. Latest year 2025, released April 2026. | Verified emissions, free allocation, compliance status, installation name and address. This is the only audited, legally required, plant-specific CO₂ number published anywhere in the world, and it is a far better anchor than any satellite estimate. Note: the EEA’s better-known ETS data viewer is aggregated by country and activity — installation level is the Registry. |
| LIVE | India: every plant’s own notified target | Greenhouse Gases Emission Intensity Target Rules, 2025, notified 8 October 2025 under the Carbon Credit Trading Scheme. | 186 named cement facilities, each with a target in tCO₂e per tonne of cement equivalent for 2025-26 and 2026-27, against a 2023-24 baseline. Integrated plants span roughly 0.28 to 1.06. For an Indian engagement this is the most useful public number about a plant that exists — it is specific, legally binding, and it is in a gazette notification rather than in anything the company published. |
| LIVE | Modelled plant emissions, worldwide | Climate TRACE. Free, CC BY 4.0. Data years 2015–2024 plus year-to-date estimates at about two months’ latency. | Per asset: asset type (e.g. “integrated dry”), activity in tonnes of cement, capacity, capacity factor, emissions factor and emissions. Cement is a covered subsector but is not listed on the public sector pages — the website alone will tell you it is not covered. It is. Uncertainty is published as ordinal confidence labels, not as an error bar; anyone quoting a ±x% accuracy for it is adding something the data does not contain. |
| PARTIAL | European pollutant releases | European Industrial Emissions Portal (formerly E-PRTR). | Pollutant releases per facility. Production volume data is not yet published, and several countries have incomplete 2023–24 data — so you cannot currently derive an emissions intensity from this portal alone. Pair its emissions with capacity from the tracker. |
| FAILS CHECKING | Two sources that are not what they are said to be | WattTime and Carbon Mapper. | WattTime publishes electricity grid emissions signals. It does not monitor cement plants or industrial facilities, and its historical satellite work was on power stations. Carbon Mapper detects methane and CO₂ super-emitter plumes; a cement kiln is a tall-stack continuous source, which is a poor match for plume detection, and no cement coverage could be confirmed. Both appear on lists of cement-monitoring tools. Neither belongs there. |
| NOT AVAILABLE | Indian environmental disclosure | Form V environmental statements, State Pollution Control Boards. | Filed annually under the Environment (Protection) Rules, 1986, but with no reliable central searchable repository. Publication is inconsistent and plant by plant, often only on individual state board sites. Treat Indian plant-level environmental disclosure as not reliably available from one source, and use the notified CCTS target instead. |
Look the plant up in the capacity tracker and write down its clinker and cement capacity, its kiln count and its parent. If it is in Europe, pull its verified emissions from the Registry for the last three years. If it is in India, find its row in the CCTS schedule and write down its 2025-26 and 2026-27 targets. Then compute the one ratio you can: emissions or target against capacity.
You will arrive knowing roughly where the plant sits and what it is legally committed to. More usefully, you will arrive able to ask a question the plant did not expect — and the answer to “how are you tracking against your notified intensity target?” opens more of a plant than any number of questions about equipment.
One honest limit: none of this tells you the heat rate, the power consumption, the clinker factor or the cost. Those are still only available from the plant, and getting them is what the rest of this lesson is about.
§ 12 What’s changing now
Current as of August 2026 · refresh every six monthsIndia’s Carbon Credit Trading Scheme made emission intensity a compliance obligation with a tradable certificate. Every plant now has a number it must know.
What it means: it constrains the clinker-factor and fuel decisions in every later lesson, and it gives you a lever with head office that did not exist two years ago. Ask who owns it — if nobody can say, that is the finding.Thermal substitution moved from voluntary target to statutory obligation. India’s TSR is around 5–8% against European plants above 50%.
What it means: the binding constraint is almost never the kiln — it is RDF supply contracts, calorific value consistency and chlorine. Ask about the supply chain, not the burner.The plant-by-plant energy registers that make up much of this course’s evidence base are not reliably retrievable online.
What it means: if the client has submitted, ask them for their own copy on day one — they will have it. And keep local copies of anything you cite. A citation you cannot retrieve will not survive a challenge.§ 13 Check yourself
§ 14 Mini case — the plant that could not produce a heat balance
A 1.8 Mt/yr plant, running 762 kcal/kg against a sector average of 740. The engagement was scoped as a thermal-efficiency review, and the pre-visit list went out a week ahead.
What arrived: the cost sheet, the fuel certificates, twelve months of production data, and the lab report with means. What did not arrive: the monthly heat balances, the free-lime standard deviations, and the stoppage log.
On the morning of the visit the process head explained that the last full heat balance was done by the kiln supplier during commissioning, eleven years earlier. The lab reported means because the LIMS had been configured that way in 2015 and nobody had changed it. The stoppage log existed as a WhatsApp group.
What that made the finding. The scoped question was “why is this plant 22 kcal/kg above average?” The answer that mattered was: nobody here can tell you, because the plant does not measure the things that would answer it. Twenty-two kcal/kg is about ₹28 a tonne of cement — ₹5.0 crore a year on 1.8 Mt. But it cannot be recovered by a project, because there is no basis on which to choose one.
The recommendation that went in first was not a thermal project at all. It was a measurement programme: a quarterly heat balance, deviation reporting in the LIMS, and a structured stoppage log. Cost, roughly ₹30 lakh in instrumentation and two people’s time. It unlocked nothing directly and it made the rest of the engagement possible.
Eighteen months later the plant was at 741 kcal/kg. No single project accounted for it.
A composite scenario. The 740 benchmark, the ₹1.90 conversion and the 0.68 clinker factor are sourced; the plant is not a real one.
The transferable move. When the data you asked for does not exist, that is the finding — not an obstacle to the finding. A plant that cannot measure cannot improve deliberately, and saying so on day one is more valuable than any project you could have proposed instead.
§ 15 Go deeper
§ 16 Carry forward
- Send the nine-item list a week ahead, and read what does not arrive as carefully as what does.
- Ask for the standard deviation of every average you are given — the setpoint is chosen for the bad hour, and everyone pays for it all the time.
- Spend ten silent minutes in the control room. Standing alarms and manual operation tell you the achievable heat rate faster than any document.
- Ask the three shutdown questions. The deferred list is a prioritised upgrade backlog assembled by people who know the plant better than you will.
- Ask for the energy-conservation register. If it exists you have their whole improvement history; if it does not, that is your first finding.