# Thinking in estimates: units, scaling and educated guessing: Why engineers estimate before they calculate

Source page: https://www.materialinsights.org/learn/engineering/thinking-in-estimates/p1/

[All courses](https://www.materialinsights.org/learn/) / [Understanding Engineering](https://www.materialinsights.org/learn/engineering/) / [Module 2: Thinking in estimates: units, scaling and educated guessing](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/) Part 1 of 5

## Why engineers estimate before they calculate

Most real problems arrive with missing numbers and no time. Engineers cope by working out a rough answer first, using round figures that can be checked. This part explains why a rough answer is worth having, shows how to build one by splitting a large unknown into small known pieces, and sets out how to handle units, which are the quickest way to catch an error.

### 1.1 A rough answer is often all a decision needs

Suppose a contractor quotes a price for a pump and overhead tank for a six-floor building, and a neighbour says it sounds high. You could ring three more contractors, or spend five minutes working out roughly how much water the building uses in a day, how much storage that implies, and what a tank and pump of that size should cost. If your answer lands within a factor of two of the quote, you stop worrying. If it lands ten times lower, you have found a question worth asking.

That is the case for estimating before calculating. A precise calculation needs inputs that are often unknown at the start of a job: the final load, the price of steel next quarter, the efficiency of a machine not yet built. An estimate needs only inputs you can guess. It tells you whether the precise calculation is worth doing, what its answer should look like, and which input matters most. Afterwards the estimate becomes a guard: if a detailed model and a rough number differ by a factor of ten, one of them is wrong, and it is seldom the rough one.

The precision usually wanted is called the order of magnitude, meaning the power of ten that the answer is nearest to. A stream carrying 3,000 litres a minute and one carrying 8,000 are both in the thousands. One carrying 30,000 is a different kind of stream. Many decisions turn only on this. Whether a village needs one pump or a dozen is settled by the power of ten.

Several of the reasoning tools here are treated at length in Sanjoy Mahajan's book on educated guessing, and the rules for units and digits follow the textbook by Kosky and colleagues. The order and examples are this course's own. Module 1 showed that engineers act on partial knowledge; this module is the arithmetic of doing so.

Value chain The estimator's toolbox, and where this module teaches each tool

- **An unknown quantity you need to know roughly**

  - **Split it into known pieces** part 1
  - **Check units and dimensions** part 2
  - **Test easy cases and limits** part 2
  - **Round, lump and take out the big part** part 3
  - **Draw a picture or find an analogy** part 3
  - **Scale from a case you know** part 4
  - **Compare with a real value and decide** part 5

 Source: Mahajan, Street-Fighting Mathematics (MIT Press, 2010), reasoning tools discussed across chapters 1 to 6, arranged here in this course's own order and extended with units, scaling and decision checks  Common misreading

#### 'An estimate is only a lazy guess'

A guess has no reasoning behind it. An estimate sets out its steps and assumptions so that someone else can check and change them. Say 'about 30 million tonnes' and show that it came from 300 million households each burning 0.3 kg a day, and anyone can argue with the 0.3 and see what moves. That openness is what separates the method from a hunch. It is also why a wrong estimate is still useful: it can be corrected one step at a time, whereas a wrong hunch cannot.

- Mahajan, Street-Fighting Mathematics (MIT Press, 2010) · Preface, and ch. 1, 'Dimensions'
- Kosky et al., Exploring Engineering (Academic Press, 2013) · ch. 2, 'Elements of Engineering Analysis'

### 1.2 Break a big unknown into small known pieces

The most widely used trick is to split a number you cannot look up into several you can roughly guess, multiply them, and accept the result. It is often called a Fermi estimate, after the physicist who enjoyed posing such questions. Nothing about it is deep. Its strength is that errors in the pieces partly cancel: one factor is a little high, another a little low, and the product lands closer than you would fear.

Try it on cooking gas. How much LPG does India burn in a year? Start with who burns it. Roughly 300 million households hold a connection, and round numbers are fine. Then ask how much each uses. A 14.2 kg cylinder typically lasts a household of four or five about six weeks, which is a little over a third of a kilogram a day, so call it 0.3 kg. Over a year that is 0.3 x 365, about 110 kg. Multiply by 3 x 10^8 households and the national figure is 3.3 x 10^10 kg, or 33 million tonnes. The table sets it out.

None of the three inputs is exact. Some homes finish a cylinder in a month. Many families who joined under the government's Ujjwala scheme refill far less often. The published consumption for 2024-25 was 31.3 million tonnes, and that includes restaurants and industry, so the household share is a little smaller. The closeness is partly luck: the true customer count of 326.8 million would give 36 million tonnes, 15 per cent high. Either way the estimate sits in the right power of ten, which is all the method promised.

Notice what it teaches beyond the number. The weakest input is the refill rate, not the customer count, so a better survey would spend its effort there. The fuel appears on the site under 'CNG, LPG and biogas distribution', and the stoves that burn it under 'Boilers, furnaces and stoves'.

Data Worked estimate: LPG burnt by Indian households in a year

| Step | Working | Result |
|---|---|---|
| Households with a connection | Active domestic customers 326.8 million, rounded to one figure | 3 x 10^8 |
| Use per household per day | 14.2 kg cylinder lasts about 6 weeks (42 days): 14.2 / 42 = 0.34, rounded | 0.3 kg |
| Use per household per year | 0.3 kg x 365 days | about 110 kg |
| National use | 3 x 10^8 households x 110 kg | 3.3 x 10^10 kg, or 33 million tonnes |
| Check against a real value | PPAC consumption in 2024-25, all uses: 31.3 million tonnes | Estimate about 5 per cent high |

The 6-week cylinder life is an assumed typical figure, not a measured national average. The published total includes commercial and industrial use.
 Source: Inshorts, 'How much LPG did India produce and consume in last 5 financial years' (reporting PPAC data, 21 March 2026), LPG consumption 31.3 million tonnes in 2024-25, https://inshorts.com/en/news/how-much-lpg-did-india-produce-and-consume-in-last-5-financial-years--1774093334297  In the data

#### Turning tonnes into energy to judge the size

At about 46 MJ per kg, 31.3 million tonnes of LPG is 31.3 x 10^9 kg x 46 x 10^6 J/kg, about 1.4 x 10^18 J (calculated). Spread over 1.4 billion people that is roughly 1 GJ, or 280 kWh of heat, each a year: a little under 1 kWh a day (calculated). A conversion is not an independent check, since it restates the same number, but it puts the total into a form you can judge: a 1 kW hot plate running about 45 minutes per person a day sounds like a day's cooking. A true second check needs a different route, such as cylinders sold.

- Construction World, 'India's LPG Q1 sales rise 4.7% as customer base grows to 326.8 million' (reporting PPAC data, 23 August 2024) · 326.8 million active domestic LPG customers of the public-sector oil marketing companies (first quarter of 2024-25), https://www.constructionworld.in/amp/energy-infrastructure/oil-and-gas/indias-lpg-q1-sales-rise-4.7-as-customer-base-grows-to-326.8-million/61155
- Inshorts, 'How much LPG did India produce and consume in last 5 financial years' (reporting PPAC data, 21 March 2026) · LPG consumption 31.3 million tonnes in 2024-25, https://inshorts.com/en/news/how-much-lpg-did-india-produce-and-consume-in-last-5-financial-years--1774093334297
- Mahajan, Street-Fighting Mathematics (MIT Press, 2010) · ch. 3, 'Lumping' (the idea of replacing a messy quantity by a simple one of the right size)

### 1.3 Units are a free error check, and Indian units need careful handling

Every quantity in an estimate is a number times a unit, and units obey the same algebra as numbers. Divide metres by seconds and you get metres per second. Try to add metres to seconds and you have made a mistake, however tidy the arithmetic looks. This is the cheapest check in engineering because it costs no extra work: write the unit beside every number and cancel units like fractions. Kosky and colleagues set each conversion factor in square brackets beside its number, such as [3,600 s/h], so that the unwanted unit visibly cancels and the wanted one remains.

The metric SI system helps because its units fit together. A newton is the force that gives a kilogram an acceleration of one metre per second squared. A joule is a newton acting through a metre. A watt is a joule per second. So a power in watts multiplied by a time in seconds is automatically an energy in joules, with no stray constants.

Indian practice adds a few traps. Big counts are in lakh (10^5) and crore (10^7), not millions. Water utilities quote supply in MLD, million litres a day, and cities quote load in megawatts. The 'unit' on an electricity bill is a kilowatt-hour, which is energy and not power, a confusion behind many wrong estimates. Land comes in acres, hectares or square feet depending on the state, and grain in quintals of 100 kg. The remedy is simple: convert everything into one system at the start, then convert back only for the final answer. A classic slip is to mix a lakh crore (10^12) with a crore (10^7), an error of a factor of 100,000 that no calculator will warn you about.

Data Units met in Indian estimates, and what they equal

| Unit | Equals | What to watch for |
|---|---|---|
| Lakh | 100,000 (10^5) | Prices and counts; not the same as a million |
| Crore | 10,000,000 (10^7) | 100 lakh; 10 crore is 100 million |
| Lakh crore | 10^12 | Budgets and market sizes; a trillion |
| MLD | 10^6 litres a day = 1,000 m3 a day = about 11.6 litres a second | A flow rate, not a volume |
| Electricity 'unit' | 1 kWh = 3.6 MJ | Energy, not power |
| Quintal | 100 kg | Grain and crop yields; a tonne is 10 quintals |
| Horsepower (pump motors) | about 750 W | Shaft output, not electrical input (see module 1) |

 Source: Kosky et al., Exploring Engineering (Academic Press, 2013), ch. 2, 'Elements of Engineering Analysis', sections on the SI system and unit conversion; Indian units added here  Worked example

#### Units in motion: a 500 MLD treatment plant

A city plant treats 500 MLD. In SI form that is 500 x 10^6 litres a day, or 5 x 10^8 L / 86,400 s, about 5,800 litres a second, or 5.8 m3/s (calculated). For a population of 5 million, 5 x 10^8 L / 5 x 10^6 people is 100 litres per person per day. Each line carried its units, so each result arrived with the right label. Had any step produced 'litres per second per person' by mistake, the units would have said so.

- Kosky et al., Exploring Engineering (Academic Press, 2013) · ch. 2, 'Elements of Engineering Analysis' (SI units, unit prefixes, conversion factors)

### See it in the table
[H₂O Water](https://www.materialinsights.org/commodities/water/)[St Steel](https://www.materialinsights.org/commodities/steel/)[LPG CNG, LPG and biogas distribution](https://www.materialinsights.org/technologies/e_cng_lpg/)[Blr Boilers, furnaces and stoves](https://www.materialinsights.org/technologies/e_combustion_heat/)[El Electricity](https://www.materialinsights.org/energy/electricity/)

### Check your understanding

A four-step estimate and a detailed computer model disagree by a factor of ten. What is the sensible first reaction?
 A factor of ten is far outside the normal uncertainty of a rough estimate, so something specific has probably gone wrong, and the rough number is often the one that is right.

In the LPG estimate, which input is the least certain and so most worth improving?
 The customer count is published, but usage varies a great deal between households, so the daily use drives the uncertainty.

A report gives a sum as 2 lakh crore rupees. In powers of ten this is:
 A lakh is 10^5 and a crore is 10^7, so a lakh crore is 10^12 and two of them are 2 x 10^12.

[← Module contents](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/)[Mark read · Part 2: Dimensions and easy cases: checking formulas and finding them →](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/p2/)

## Other sections

- [Module overview](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/index.md)
- [Dimensions and easy cases: checking formulas and finding them](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/p2/index.md)
- [The big part first, then rounding, lumping and pictures](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/p3/index.md)
- [Scaling: why size changes everything](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/p4/index.md)
- [When an estimate is good enough to decide](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/p5/index.md)
- [Key terms, quiz and sources](https://www.materialinsights.org/learn/engineering/thinking-in-estimates/review/index.md)
