Fact fluency

Building fact fluency when your kid already thinks they’re bad at math

Written 10 September 2026 by Daniel, who built this

A kid who has to work out 7 × 8 from scratch every single time is spending almost all of their attention on the arithmetic, and has none left for whatever the arithmetic was sitting inside. From the outside that looks like being slow at math generally. From the inside, to a nine year old keeping score of their own performance, it looks like proof. Fact fluency is often most of the gap, and unlike being good at math, it’s a thing you can actually move.

What is happening when they say it

The sentence lands hard because it sounds like a verdict about the child rather than a description of a skill. The cognitive story underneath it’s much more boring than that.

Working memory is the small amount of mental space you get to hold things in while you think. It’s small for everyone. When a fact isn’t stored, the kid has to calculate it, and calculating it occupies that space. Daniel Willingham's column in American Educator uses long division as the example: a long division problem has small subtraction problems buried inside it, and a student who retrieves those answers instantly keeps their working memory free for the division itself. A student who has to work them out is running two problems at once, and the one that gets dropped is the one they were assigned.

His summary of the underlying research is that before facts are learned to automaticity, calculating them "does indeed require working memory," and that with enough practice the answers can be pulled from memory instead, "thereby incurring virtually no cost to working memory."

So a kid who stalls on a word problem is frequently not bad at word problems. They ran out of room.

Why reassurance bounces off

They aren’t wrong about the evidence, which is the problem. They really are slower than the kid at the next desk, and they can see it every day. A parent saying otherwise contradicts something the child observes directly, so it gets filed as a parent being kind and it changes nothing.

There’s a second loop here, and it’s not a soft one. Sian Beilock and Daniel Willingham, writing in the same series, describe math anxiety as consuming the same resource: "Math anxiety robs people of working memory." An anxious kid is solving the problem and managing the worry at the same time, so there’s less room for the math and the performance drops, which then supplies fresh evidence for the belief that started it. They also note the likely origin: a shaky grasp of the basic building blocks early on is one of the things that predisposes a child to become anxious about math in the first place.

Both loops run through the same bottleneck. That means the intervention is the same in both directions. Take the arithmetic off the working memory bill and you get the room back twice.

"I don’t know it" and "I know it slowly" are different problems

This is the distinction I would most want a parent to have, and almost nothing scores for it.

A kid who doesn’t know 7 × 8 at all will guess, or stall, or leave it. A kid who knows it slowly is doing something quite different: counting up, using fingers under the table, going from 7 × 7 and adding another seven. That’s real math and it works. It is also the stage children are meant to grow out of. The developmental research describes a progression from counting everything, to counting on from the larger number, to retrieving the answer directly, and the entire value of the last step is that it costs nothing.

Both kids write 56. Both get a check mark. On a worksheet, in most apps, and on most tests they’re indistinguishable, so a kid who is quietly finger counting every fact can score well for a year while the gap widens underneath. That’s the case that slips through, and it’s the only reason timing matters at all.

The two need different responses, too. Not knowing a fact is an exposure problem and it resolves fast. Knowing a fact slowly is a repetition problem, and it needs the same fact to come back on many separate days until the counting drops away on its own. Treating the second one as solved because the answer was right is how it survives for months.

Drop the level further than feels reasonable

If your kid has already decided they’re bad at math, the first thing to change isn’t their fluency. It’s the evidence. Put them somewhere they’re getting nearly everything right, and then leave them there longer than feels productive.

It will feel like a waste. A ten year old doing addition inside ten isn’t being stretched in any obvious way. But a child who has spent two years being visibly slower than the person beside them is holding a mountain of evidence and one afternoon of you saying otherwise, and the only thing that touches the mountain is more evidence. Easy work they finish counts. Hard work they fail is just more of the same.

There’s a mechanical reason as well. Automatic recall is built by repeating facts the kid can already get right. Practicing facts they can’t do doesn’t build recall at all. It builds fluency at counting, which is the habit you were trying to retire.

PiggyBankMath has an optional five minute assessment at setup that places the kid where they actually are, and it’s happy to place them below where school has them, which is the useful direction. After that the range widens on its own as facts get locked in, so you’re not the one deciding when they’re ready. You can re-run the assessment any time you think it got them wrong.

Don’t put a clock in front of them

Speed is the signal you care about, and a visible timer is close to the worst thing to hand an anxious kid. Those two facts are in tension, so I should say plainly how the app resolves it.

PiggyBankMath measures how long each answer takes. It never shows a countdown, never marks a question wrong for being slow, and never pays less for a slow round. What the timing changes is only whether a fact is treated as learned: a fact counts as mastered after three fast, correct answers across at least two different days, and a correct-but-slow answer simply doesn’t advance it. The fact stays in rotation and comes back later. From the kid's side, nothing happened.

So the honest answer to whether you should time your kid is that you should know the time and they should not. If you’re doing this on paper, that means you watching, and not a stopwatch sitting on the table between you.

The parts I’m not sure about

I would rather put these here than leave them out.

Willingham is careful about exactly the claim this page rests on. He writes that you would expect interventions improving automatic recall to also improve proficiency in harder mathematics, and that the evidence on that point is "positive but limited." The mechanism is well established. The size of the payoff further up the chain is less settled than I would like, and I’m building on the direction of that evidence rather than on a finished result.

Fluency is also not all of math. A kid who doesn’t understand what division is for will not be rescued by knowing that 56 divided by 7 is 8. If the real trouble is conceptual, fact practice will make them faster at the part that was already fine and leave the actual problem exactly where it was. It’s worth ruling that out first, usually by asking them to explain a problem rather than solve one.

And some kids have a genuine learning difficulty, for which no app of any kind is the answer. I’m a developer with two children, not a specialist in any of this. If the picture doesn’t move over a couple of months of steady practice, that’s worth taking to somebody who assesses children for a living.