Scope. This page provides a controlled reader-run protocol and does not claim site-run measurements. For cooking, cooling, storage and reheating, follow the current official guidance linked on the page.
Find your symptom first
This is a diagnostic list, not a set of tips, so start from what you actually saw rather than reading it top to bottom.
| What happened | Most likely mistakes |
|---|---|
| Grainy and separated, with watery liquid | 4, 6, 7, 8 |
| Smooth in the pan, grainy by the time it was in the bowl | 3, 7, 8 |
| Never thickened at all, however long you cooked it | 1, 4 |
| Thickened, then thinned again in the fridge | 1, 9 |
| Lumps of set egg in an otherwise fine custard | 2, 5, 6 |
| Skin, or a rubbery layer on top | 7 |
| Tastes faintly of cooked or scorched milk | 4, 6 |
| Fine at the time, weeping the next day | 4, 9 |
Each mistake below states the mechanism first, because "do not do X" without "because Y" is precisely the writing this site exists to replace. If you know why a step matters you can adapt it to your pan, your hob and your recipe. If you only know the instruction, you are following a ritual.
Ordered by when they happen, which is also roughly the order in which they are recoverable.
Before the heat goes on
1. Choosing the ratio without deciding the target texture
Egg proteins are the structure. How much structure you get depends on how many yolks are dispersed in how much liquid, so the yolk-to-dairy ratio determines the thickness the custard can reach, not the cooking. A thin ratio cooked longer does not become a thick custard. It becomes a thin custard that has been taken to the edge of curdling in pursuit of a thickness it was never able to reach.
This is the mistake behind most "it never thickened" complaints, and it is the one that produces the worst downstream behaviour, because the cook responds by applying more heat and crosses the failure point chasing something structural.
Decide first whether you want a pouring custard, an ice cream base or a set custard, and pick the ratio for that. Then cook to a temperature rather than to a consistency.
2. Letting sugar sit on the yolks
Sugar in contact with yolk draws water out of it. Left sitting, the yolk proteins locally dehydrate and form hard granules that never fully redisperse, and you get small firm specks in the finished custard that look exactly like the beginnings of curdling but happened before any heat was applied.
The instruction "whisk the sugar into the yolks immediately" exists for this reason, and it is the reason it is always phrased urgently. Weigh the sugar out first, add it, and whisk it in at once rather than turning to deal with the milk.
There is no universal “sitting” interval stated here: it depends on the mixture and temperatures. If you are testing it, start the next step immediately and record the actual delay.
3. Not preparing the exit before the heat goes on
A custard finishes faster than you can find a sieve. The pan retains heat after the hob is off, the residual heat keeps cooking it, and the interval you spend hunting for a bowl is exactly the interval in which a correct custard becomes a curdled one.
Before anything is heated: fine sieve, cold receiving bowl, and, for anything you care about, a larger bowl of iced water for the receiving bowl to sit in. This is not fussiness. It is the difference between stopping the cook and merely stopping the heat, and they are not the same thing.
During the cook
4. Cooking above the window because low heat feels slow
The most common mistake and the most consequential. An egg custard has a narrow working range and a failure temperature above it, and the width of that range is the whole margin you have. Applying more heat does not shorten the cook proportionally; it steepens the climb, and a steeper climb means you cross the range faster than you can react to it.
There is a second-order effect that makes this worse than it looks. The base of the pan is hotter than the bulk, and the higher the heat input, the larger that gap. At a high setting the layer against the metal is past the failure temperature long before the probe in the middle reads anything alarming. The specks you eventually see were made minutes earlier, at the bottom.
Use the reader-run worksheet in the custard temperature window; do not infer a universal failure temperature from another formula.
5. Treating tempering as a ritual rather than as rate control
Tempering exists to control one thing: the rate at which the yolk temperature rises. Pouring hot liquid into cold yolks raises the temperature of the yolk it touches almost instantly, and that yolk sets on contact into lumps that will not disperse. Adding the hot liquid gradually with constant agitation keeps the local temperature rise slow enough that nothing sets before the whole mixture is warm.
Understanding it as rate control rather than as a fixed sequence tells you when you can shorten it, which is more useful than the ritual. The full argument is in how to temper eggs properly.
The failure mode is distinctive and worth recognising: lumps of set egg in an otherwise perfectly good custard. That is a tempering fault, not an overcooking fault, and no amount of gentler heat later will fix it.
6. Stirring the middle instead of the base and corners
The damage happens at the interface between the liquid and the hot metal, and specifically in the stagnant film that sits against the pan and in the corner where the base meets the wall. Vigorous stirring in the centre of the pan moves the bulk around impressively and leaves that film exactly where it is.
Use a flat-edged implement that reaches the corner, and move it across the whole base in a pattern rather than swirling. The point is not enthusiasm; it is contact coverage. This is the same mechanism that scorches dairy on a base that never looked hot, which is covered separately.
After the heat comes off
7. Deciding when to stop by eye
Every visual cue for custard is a lagging indicator. Coating the back of a spoon means the viscosity has already changed enough for you to see it, which means the temperature reached that point some seconds ago and the base of the pan reached it before that. Judging by eye consistently stops you late, and late in a narrow window is the failure.
A probe reads the same bulk that your eye reads, but it gives you a number continuously, which means it gives you the rate of climb. Rate is the thing you actually need to act on and no visual cue provides it.
If you are working without a probe, pull at the first sign of thickening rather than the confident one. The difference between those two impressions is usually most of the remaining margin.
8. Leaving it in the pan
The hob is off. The pan is not. A heavy base holds enough energy to keep pushing a custard upwards for some time, and a custard left standing in the pan it was cooked in can cross its failure temperature with nobody touching it and no heat applied.
Measure carryover after removal in the same session, with the pan and volume recorded.
Decant immediately. Through the sieve, into the cold bowl, and stir for a moment to bring the temperature down. The sieve catches the small amount of coagulation that always occurs at the edges, and the transfer stops the cook rather than merely stopping the heat.
9. Trying to cook a curdled custard back
Protein aggregation is not reversible. Once the strands have clumped and squeezed out the liquid they were holding, no further heating, no additional egg, and no amount of whisking puts the water back inside the network. Additional heat makes the clumps larger.
What can be done is to make the graininess imperceptible rather than to undo it: blend, then strain. The texture will be heavier than intended and the result will be a usable sauce rather than the custard you set out to make. We give no recovery rate for that. If you want one for your own recipe, curdle a stated number of small batches deliberately, attempt the rescue on each, and record how many it saves.
The related error is reheating a finished custard later without respecting the same window. A cooked custard is not immune. Warming it past the failure temperature curdles it exactly as readily as cooking it there in the first place.
Which of these actually accounts for most failures
The honest answer is that nobody has counted, so we do not know.
Our working hypothesis is that mistakes 4 and 8 dominate, because they are the two that operate through the same underlying cause: the difference between the temperature you are looking at and the temperature that is actually doing the damage, once at the base of the pan and once after the heat is off. If that hypothesis is right, most custard failures are not skill failures at all but measurement failures.
This page does not assert a frequency without counted trials. Run the failures deliberately on your own equipment and record what the session shows, including a result that contradicts the expectation.
Nine is the number of distinct mechanisms we could actually identify and defend.
In Plain English
Custard fails in three phases. Before the heat, through ratio and preparation. During the cook, through excess heat, poor tempering and incomplete stirring of the base. After the heat, through late stopping and residual pan heat. Each mistake has a physical mechanism, and knowing the mechanism means you can adapt to your own equipment rather than following instructions that assume somebody else's.
Who this helps: anyone who has made the same custard several times with different results and no idea which step is the variable.
When to use it: after a failure, starting from the symptom table at the top. Before a failure, as a setup check: ratio decided, sugar whisked in at once, sieve and cold bowl ready, probe in the pan.
What to do next: work out which phase your failure lives in. Grains that appeared instantly on contact are tempering. Grains that appeared gradually are heat. Grains that appeared after the hob was off are the pan. Those three answers point at completely different fixes, and guessing between them is why the same failure repeats.
Next Step
The mechanisms here are all downstream of one narrow temperature range and one hard failure point. Read the narrow window that decides a custard for the window itself, what starch does to it, and why a pastry cream can be boiled when a crème anglaise cannot. Any term used here that you would like defined plainly is in the glossary.
Related technique: Your Scrambled Eggs At Eight Temperatures.