Why fuel consumption rises, and which causes are worth chasing

Fuel consumption is the most useful number a car gives you, and almost nobody writes it down.

That is the whole argument of this article, and it is worth stating before anything else, because it inverts the way the subject is usually covered. Rising consumption is not mainly a cost problem to be minimised with tips. It is a diagnostic signal — and for a large family of faults it is the only signal there is. A stuck-open thermostat, a slow oxygen sensor, a contaminated airflow sensor and a dragging brake caliper have one thing in common: the car drives normally, nothing lights up on the dashboard, and the only evidence that anything is wrong is that you are filling up more often.

Which means the ability to notice a ten per cent rise is a diagnostic instrument, and it costs nothing but a note on your phone.

The three archive pieces this is written from split the subject three ways: one lists causes, one lists fourteen tips, and one explains how to calculate consumption from two fill-ups. The third was the most read of the three, which suggests readers had already worked out which part mattered. It is also the part that has to come first, because every other section depends on it: you cannot investigate a rise you cannot detect.

Measuring it, so that a change means something

The method needs no equipment.

1. Fill the tank until the pump clicks off. Note the odometer.
2. Drive normally until the tank is low. Do not change how you drive.
3. Fill it again, at the same pump if you can, to the same click-off. Note the odometer and the litres added.

Divide: litres added × 100 ÷ kilometres travelled = litres per 100 km.

The original's worked example is arithmetically correct, and it is worth keeping because the numbers are realistic. The odometer goes from 10,000 to 10,501, so 501 km; the second fill takes 43 litres.

43 × 100 ÷ 501 = 8.58 litres per 100 km — about 27 mpg US, or 33 mpg imperial.

The original omits the three rules that decide whether that number means anything, and without them the method produces figures that swing several per cent for no reason at all and teach you to chase phantoms.

Fill to the same point both times. You are not measuring the fuel you burned; you are measuring the fuel it took to get back to a fixed level. If that level moves, the measurement moves with it. Same pump if possible, same nozzle position, stop at the first click — topping up past it is the single largest source of error, because how much more fits depends on the filler neck's angle and the day.

Go far enough between fills. This is the rule that matters most and the one nowhere in the original. Suppose your two fills differ by one litre, which is easy. Over the example's 501 km that is 0.2 L/100 km — about 2.3 per cent, tolerable. Do the same measurement over 150 km and the same one-litre discrepancy becomes 0.67 L/100 km, nearly 8 per cent, which is larger than most of the faults you are looking for. Measure over a full tank, not over a few days.

Compare like with like. A tank of motorway driving and a tank of school runs are different measurements of different things, and neither is evidence about the other. Winter consumption is genuinely higher than summer — a cold engine spends longer on its warm-up map, and short trips may never leave it.

The baseline is your own earlier figure, not the manufacturer's

The original ends by suggesting you compare your result against the maker's published figure. That is the wrong baseline and it generates false alarms, because official figures come from a standardised laboratory cycle rather than from a road. Real-world consumption sits meaningfully above them for almost every car and almost every driver, and that gap is not a fault.

What you want is your own number, from the same car, measured the same way, a few months ago. A car that returned 8.6 L/100 km all last year and returns 9.8 this month has told you something specific. The same car measured once against a brochure figure of 7.1 has told you nothing at all.

Three or four measurements are enough to establish a baseline. Write them down with the date.

What about the trip computer?

Useful, and not an arbiter. A trip computer generally works out consumption from how long the injectors were commanded open, not from fuel that actually left the tank, so it inherits any error in the thing you might be trying to measure — a fuelling fault can move the real figure and the displayed one differently. Treat it as a good trend instrument and settle arguments with the pump.

What actually raises it

Ordered by how often it turns out to be the answer, and by whether you would notice anything else.

An engine that never gets warm

This is the single best example of why consumption is worth measuring, and it is the first thing to check. Until the engine reaches temperature the module runs it on a fixed warm-up map that is deliberately rich, and it will not start trusting the oxygen sensors until it does. A thermostat stuck open means that never happens: the engine runs cool, rich, and permanently.

The corpus is blunt about what that feels like from the driver's seat — the car drives normally and uses more fuel. P0128 is one of the few codes where replacing the named part first is genuinely right, and it is an inexpensive part. The tell you can check without any equipment is the heater: if it never blows properly hot in cold weather and the temperature gauge sits low, that is your answer. The thermostat article covers the rest, including why fitting a cooler thermostat to cure an overheating problem hides the fault instead of fixing it.

An oxygen sensor that has gone slow

The failure mode people expect is a dead sensor. The one that actually costs fuel is a sensor that still works and no longer works quickly — response time is what degrades, not output voltage. The module is then correcting the mixture using a description of the exhaust from a moment ago, so fuelling wanders further either side of target before each correction lands.

A slow sensor frequently sets no code, which is the part worth carrying away, because the absence of a warning light is widely read as proof the part is fine. P0133 exists for exactly this and does not always catch it. The oxygen sensor article has the test that does, and the important half of it is that a static voltage check passes a sensor that is finished.

Fuel trim, which is the one measurement worth learning

If you can read live data — or are talking to somebody who can — fuel trim answers the question "why" faster than any part swap. It is the percentage the module is adding to or subtracting from its base fuelling to keep the mixture right, so it is a direct readout of how wrong everything else is.

The comparison the corpus gives at P0171 is the one to know:

Compare trim at idle against trim at higher engine speed. A vacuum leak has a large effect at idle, when manifold vacuum is high and total airflow is low, and its influence fades as the throttle opens. A fuel delivery problem behaves in the opposite direction, getting worse as demand rises.

That single comparison separates the two largest families of cause before anything is unbolted. Reading live data covers what good values look like.

Note also which way each code costs you fuel. P0172 — too rich — obviously does. P0171 — too lean — does too, because the module is adding fuel to compensate for air it did not measure, and it is doing so with a mixture that is wandering. Running lean is also the more damaging of the two.

Misfire, and what it does downstream

A cylinder that does not fire sends its fuel out of the exhaust unburnt. That is the immediate cost, and there is a larger one behind it: unburnt fuel burning in the catalytic converter instead of in the cylinder overheats it, which is how a misfire left alone turns into a much more expensive bill. A flashing engine light means exactly this and means stop driving. P0300 is a misfire the module could not attribute to one cylinder; P0301 and its neighbours name one.

Drag: brakes and tyres

Both are real, both are cheap to check, and the original gives neither a test.

A dragging or binding caliper generates heat continuously and costs measurable fuel while being small enough not to be obvious from the driver's seat. The confirming check costs nothing: after a run, feel whether one wheel is noticeably hotter than the one opposite it. Seized slider pins and corroded piston seals are the usual causes — and so is a collapsed flexible hose acting as a one-way valve, which lets pressure in and not out, so the caliper drags while the caliper itself is perfect.

Underinflated tyres are the cheapest check available and are skipped constantly. Measure all four cold, against the placard in the door aperture or the filler flap — not against the number moulded into the sidewall, which is a maximum and not a recommendation. The tyres article has more on this, including why the nitrogen sold on the strength of fuel savings is not the thing delivering them.

Five things that get blamed, and mostly are not it

This is where the original is at its weakest, and every item here was checked rather than reasoned about.

A dirty air filter

The tips article calls a dirty air filter "one of the commonest causes of increased fuel consumption". On any car with fuel injection, it is not — and this is measured rather than argued. A 2009 study for the US Department of Energy tested three modern fuel-injected cars with filters progressively clogged to the point of near-undrivability and found no significant effect on fuel economy at all. What did change was acceleration, which improved with a clean filter on every vehicle tested.

The reason is the one our own corpus gives for a restricted air filter: on an engine that measures airflow directly, the module fuels correctly for the reduced air, and the driver simply gets less engine. The mixture stays right. You lose power, not fuel.

Two honest qualifications, because "air filters never matter" would be as wrong as the claim it replaces.

• A carburettor behaves differently. The same study's carburetted comparison vehicle did lose economy — restriction increases the depression across the venturi and pulls extra fuel with no closed-loop control to trim it back. That is where this rule of thumb comes from, and it was true when most cars had carburettors. The carburettor article covers why. It is not true of anything with injection.
• Severely blocked is a different state from dirty. A filter left far past its interval, soaked with oil, or blocked with nesting material can push the airflow signal outside what the module can reconcile, and then fuelling does go wrong. Check it by looking rather than by mileage: hold it to a light, and a filter that passes none is finished whatever the service record says.

One genuine filter-related fuel fault runs the other way, and it is worth knowing because people meet it while trying to save fuel. A re-oilable performance filter that has been oiled too generously sheds a mist onto the mass air flow sensor's hot-wire element. A coated element loses heat more slowly, so it under-reports airflow and the module leans the mixture, progressively, as the coating builds. Fitting a new sensor without addressing the filter contaminates the new one within weeks.

Premium fuel

The tips article says using super unleaded lengthens engine life and reduces consumption. In an engine that does not require it, higher octane does neither. Octane is a measure of resistance to knock, not of energy content: it does not contain more energy per litre, and an engine calibrated for regular cannot convert the extra knock resistance into anything. AAA's laboratory testing found no significant benefit across any category tested, and estimated that American drivers alone spend around $2.1 billion a year on premium fuel their cars do not ask for.

If the handbook says premium is required, use it — that engine's timing and compression assume it. If it says regular, the expensive pump is a donation.

Worn tyres

The causes article says worn tyres have higher rolling resistance and raise consumption. This is backwards. Rolling resistance falls as a tyre wears — by roughly 20 per cent from new to worn — because there is less tread mass to deform and less squirm in the blocks. Tyre industry figures put a 10 per cent change in rolling resistance at 1 to 2 per cent of fuel economy, so fitting a new set typically costs you a couple of per cent, and that is normal rather than a fault.

None of which is an argument for driving on worn tyres. The reason to replace them is wet grip and aquaplaning resistance, both of which collapse as tread depth goes, and those are worth far more than two per cent of anything. But if you replace your tyres and your consumption rises slightly, nothing is wrong — and if somebody sells you tyres on a fuel-saving argument, that argument only holds against a low-rolling-resistance tyre, not against the worn ones coming off.

Under-inflation is the real tyre-related fuel cost, and it is separate: about 0.2 per cent of economy per psi below the placard figure, averaged across all four, with up to 3 per cent available from correcting a badly neglected set.

A catalytic converter that is simply old

The tips article says converters lose their function after two to three years and should be replaced. This is not close, and it is expensive advice.

A converter on a healthy engine should last 160,000 km or more, and very often the life of the car. The archive makes this error repeatedly — our article on the converter already corrects three further figures from a different piece, which gave 50–60,000 km, then 80,000 km, then about four years, in the space of one paragraph.

The reason it matters is not the wasted money alone. A converter that fails early is a symptom of something upstream — a misfire, oil consumption, coolant from a head gasket, the wrong sealant on an intake gasket. Replace it on age and you replace it again. P0420 is also one of the most misdiagnosed codes there is: it reports that a comparison failed, and a lazy downstream sensor or an exhaust leak ahead of it will produce it with a perfectly good converter underneath.

The same section's advice about non-standard exhausts is directionally reasonable and wildly overstated: a different silencer will not meaningfully change consumption on a standard engine. Removing the converter will, along with setting a permanent fault.

Petrol evaporating out of a parked car

The tips article recommends parking in shade so the fuel does not evaporate, and the causes article treats vapour loss as a notable cause of high consumption. On any car with a sealed evaporative system — which is every petrol car built in the last few decades — fuel vapour does not escape to the atmosphere. It is captured in a charcoal canister and drawn into the engine to be burned later. That is the entire purpose of the system.

So the vapour is not lost, and the fuel gauge is not falling in the car park. What an evaporative fault actually produces is a warning light: P0455 and P0456 are leak codes, and the corpus rates them low severity precisely because they cost emissions compliance rather than fuel.

There is one evaporative fault that genuinely affects running, and it is the opposite of a leak. A purge valve stuck open feeds vapour into the intake continuously, including at idle where the engine can least tolerate it — P0441. And a saturated canister dumps stored fuel into the intake and shows up as running rich, most strongly just after refuelling. If the car runs badly immediately after a fill-up, say so — that detail points straight at the evaporative system and is easy to leave out.

Park in the shade anyway. The reason is that the air conditioning has less work to do afterwards, which is a real if modest saving, and the interior lasts longer.

Two instructions in the original that will cost you a part

Both are given as routine maintenance. Both damage the thing they are meant to preserve.

"Remove the oxygen sensor every so often and clean its tip." Do not. The sensing element is a porous ceramic with a platinum coating, and it is not a surface that can be cleaned. A wire brush, a pick, abrasive paper or a solvent will scratch, dissolve or crack it, and heat from a torch will crack it internally. Contamination of an oxygen sensor is chemical — silicone, oil, coolant, lead — bonded into the pores rather than sitting on the surface, which is why cleaning does not recover one even when it appears to. A sensor slow enough to matter is replaced, and the thing that poisoned it is found first, or the new one goes the same way.

Note the contrast with the airflow sensor above, which can sometimes be cleaned — with the manufacturer's approved cleaner and no physical contact with the wire. The two parts get opposite advice for good reasons and the advice is not transferable.

"Remove and re-gap the spark plugs every 5,000 to 10,000 km." This is workshop practice from an era of conventional plugs and it is actively harmful on modern ones. NGK explicitly advises against re-gapping fine-wire platinum and iridium plugs: the precious-metal tip is laser-welded to the centre electrode, is around 0.6 mm across, and is not designed to withstand any force. NGK will not warrant a broken tip, and their own guidance is that a fine-wire plug slightly out of gap still outperforms a correctly gapped conventional one — so there is nothing to gain and a new part to lose. Repeatedly removing plugs from an aluminium head has its own risks.

The stated service life is wrong in the same direction. The original puts ordinary plugs at no more than 20,000 km, which would have you replacing a serviceable part perhaps twice as often as needed; the interval is the handbook's, and for fine-wire plugs it is several times longer again. The spark plug article covers heat range — which matters far more than the electrode material everybody argues about — and the reason a worn plug is a genuine cause of coil failure rather than an innocent bystander.

While we are here: the original says an injector must produce a 30-degree spray cone and that "any other spray form is rejected". Spray geometry is designed per application — single-stream, two-hole and multi-hole injectors all differ, and published cone angles for port injectors run far wider than 30 degrees. The underlying concern is real, and it is worth keeping: a poor spray pattern puddles fuel on the port wall behind the valve, which then enters the cylinder as liquid rather than mist and burns badly. But there is no universal angle, and clogged injectors are diagnosed by how the engine runs and what the fuel trims say, not by comparing a spray to a number.

Driving: the levers that are real, and how big they are

The behavioural tips are mostly sound, and mostly unquantified, which makes it impossible to tell the ones worth changing from the ones worth ignoring. Numbers, then.

• Smooth throttle and anticipation. By far the largest lever, larger than every part in this article. No number, because it depends entirely on how you drive now.
• Weight. About 1 per cent per 45 kg. Real, and small — clearing the boot is worth doing and will not transform anything.
• Idling. Restarting a modern engine costs roughly ten seconds' worth of idling fuel, so the threshold for switching off is about ten seconds, not the two to three minutes the original gives. Idling burns something like one to two litres an hour.
• Tyre pressures. About 0.2 per cent per psi, as above. Check monthly, cold.
• Gear selection. The original advises against 4th gear at 40–50 km/h. This is backwards. For a given road speed a higher gear means lower engine speed and less fuel; cruising in the highest gear the engine will pull smoothly is correct. What is genuinely harmful is labouring — a wide-open throttle at very low revs, where the engine shudders and can knock — and that is a throttle-position problem, not a gear-number one. If you want to accelerate, change down. If you want to cruise, stay up.
• Air conditioning versus open windows. Genuinely contested, and the honest answer is that it is close. A 2004 SAE study on large-engined American cars found open windows the worse of the two at highway speed; a 2013 test on an ordinary saloon put the crossover above 130 km/h, meaning windows win below that. The difference either way is a few per cent. Use whichever is comfortable and spend the attention on the throttle instead.

What to do when the number moves

If your measured consumption has risen and you want a sequence rather than a list:

1. Confirm it over a second full tank, measured the same way. A single high tank is noise.
2. Read the codes, including stored and pending ones. Many of the faults above set none, but the ones that do will save you the rest of the list.
3. Check the temperature gauge and the heater. A thermostat is cheap and is the commonest cause of a permanent, symptomless rise.
4. Feel the wheels after a run for a hot corner, and check all four pressures cold.
5. Read fuel trim at idle and at speed if you can get to it, and use the comparison above.
6. Then start considering sensors — and even then, look at what the trims say before replacing anything, because the sensor reporting a problem is usually the only part working correctly.

The increased fuel consumption page lists every trouble code in the corpus that names this symptom, which is a longer list than most people expect.

Video guides

Video, for the parts of this that are easier watched than read.