Number Typing Test
Digits, decimals, currency and dates — on the number row or the keypad
Every other drill on this site keeps your hands near home position. This one does not: it sends them up to the number row and leaves them there. That single property decides which fingers you use, why the keypad exists, and which figures a keypad cannot type at all.
What you will actually be typing
Figures, one after another, with nothing between them. What follows is a real run from the generator at each level — captured output rather than examples chosen to look tidy — and beneath each one, what that level measures.
551 285 931 512 672 295 267 407 350 591 733 756 611 994 343 753 861 208
278.95 2435 693782 1212 197,790 429.54 18066 849,546 995,347 489,132 476929 3719 28877 775,876 89179 648036 4641 06925
$27,861.95 2435776 7826112 03/04/2019 $42,923.54 1806638 12/03/2022 $77,536.47 $94,554.29 371966 10/25/2025 784891 676480367 06/19/2022 09/16/2018 $99,854.25 06370788 08478892
Notice what changes between them. Easy is pure digits. Medium introduces exactly two other characters, the comma and the decimal point. Hard adds the slash, the hyphen and the dollar sign. Five punctuation marks across the whole drill — and as the rest of this page shows, those five characters account for almost everything that makes one level harder than another.
Who this test is for
People who key figures in volume: invoicing and payroll clerks, bookkeepers, bank and back-office staff, claims processors, stock and inventory teams entering quantities, and anyone facing an assessment with a separate numeric section. It is also the right drill for a specific and common problem — typing well on letters while still hunting for digits, which is easy to hide in ordinary work and impossible to hide here.
It is a drill rather than a simulation. The figures arrive bare, with no record around them and no field to tab into, because the point is to isolate the movement. Once the movement is solid, the context belongs somewhere else — the pages listed at the end put the same digits back inside records and ledgers.
| Durations | 30 seconds, 1 minute, 2 minutes or 5 minutes |
|---|---|
| Content | Generated figures — integers, decimals, comma-grouped thousands, dates, references and currency |
| Mode | Word mode only. Each figure is scored as its own unit |
| Input device | Number row or numeric keypad — the test cannot tell which you used |
| Scoring formula | Standard 5-keystroke word. A figure counts only when every character in it matches |
| Backspace | Enabled unless you switch it off above the test |
| Digit share | 100% of characters on Easy, 89.8% on Medium, 84.4% on Hard |
| Token length | 2–3 characters on Easy, 4–7 on Medium, 5–10 on Hard |
| Number convention | US grouping — comma for thousands, point for decimals. See the conventions section below |
| Repeat content | None. Figures are generated fresh on every run |
| Report card | Net and gross WPM, accuracy, consistency, and a heatmap that only has ten keys to blame |
| Cost | Free and account-free. Everything is worked out on your own machine |
The one drill where your hand leaves home and stays there
We measure how far a piece of text keeps the hand from rest by counting the keyboard rows between each keystroke and the home row — the method is set out on the alpha-numeric page, which uses the same measurement to make the opposite point. Applied here it produces the most extreme result on the site. Easy scores 2, and 2 is the ceiling: the digits sit two rows above home, every character is a digit, so there is nowhere further for the average to go. Ordinary prose scores 0.68.
The second figure is the interesting one. On Easy, adjacent keystrokes change keyboard row 0% of the time — never. The hand goes up once and does not come back down until the run ends. No other text on this site does that: prose changes row on 69.3% of adjacent keystrokes, and the alpha-numeric drill on 74.5%.
| Text | Number row | Rows from home | Row change |
|---|---|---|---|
| Ordinary prose | 0.1% | 0.68 | 69.3% |
| Alpha-numeric, Medium | 29.1% | 1.04 | 74.5% |
| This drill, Easy | 100% | 2 | 0% |
| This drill, Medium | 89.8% | 1.9 | 24.5% |
| This drill, Hard | 89.5% | 1.89 | 23.8% |
This is the opposite problem to the one the alpha-numeric drill poses. There the digits are scattered among letters and the hand alternates constantly, which is what makes that text slow. Here the digits are all there is, so the alternation disappears and a different cost takes its place: sustained reach. Your fingers are working at full extension for the length of the run, and the fatigue that produces is why numeric accuracy tends to fall off later in a long session rather than at the start.
It is also, and this is the practical point, the reason the numeric keypad exists. A separate block of keys is only worth having if the hand can settle into it and stay. On text like this it can.
The number row and the keypad are two different skills
They produce identical characters and almost nothing else about them is the same. One splits ten digits across both hands as reaches; the other stacks them under one hand as a resting position. Learning one does not teach you the other.
| Number row | Numeric keypad | |
|---|---|---|
| Hands used | Both, split at 5 and 6 | Right hand alone, left hand free for a mouse, a document or a Tab key |
| Where it lives | Every keyboard ever made, laptops included | Full-size keyboards only. Absent from most laptops and every compact board |
| Home position | ASDF JKL; — the digits are a reach away, never a resting place | 4-5-6, with a raised dot on 5. The digits are the resting place |
| Mixed with letters | Fine. The hand is already there | Poor. Every letter means leaving the keypad and finding home again |
| Punctuation reachable | All of it | Only . / * − +. No comma, no currency symbol |
| Usual unit | WPM, alongside ordinary typing | KPH, quoted separately in data entry postings |
Which to learn is a question about your work rather than about typing. If figures reach you inside sentences, records or forms, the number row wins outright, because the hand is already on the main keyboard and the journey to a keypad costs more than the keypad saves. If figures arrive in columns, on a full-size keyboard, with the left hand free for a source document, the keypad wins clearly. Most people need the number row and some people need both.
What a numeric keypad cannot type
A keypad has fifteen character keys: the ten digits, the four operators and the decimal point. That is the whole set. It has no comma and no currency symbol, and those two absences matter far more than they sound, because both appear constantly in the figures people actually key.
We measured it directly. For each generated figure, the question is whether every one of its characters exists on the keypad — that is, whether the hand can type it start to finish without moving.
| Text | Keypad-only | What decides it |
|---|---|---|
| This drill — Easy | 100% | Bare 2–3 digit integers. Nothing but digits, so nothing forces the hand off |
| This drill — Medium | 65% | Comma-grouped figures like 197,790 break it. The comma is not a keypad key |
| This drill — Hard | 76% | Dates and reference numbers use / and −, both on the keypad. Only the currency amounts break it |
| Alpha-numeric drill — Medium | 0.3% | Letters are not on a keypad at all, so almost no token survives |
Hard is easier on a keypad than Medium
76% of Hard figures are fully keypad-typable against 65% of Medium ones, which is the opposite of what the difficulty labels suggest. It is entirely a property of the punctuation. Hard is built largely from dates and reference numbers, and both the slash and the hyphen are keypad keys — 04/15/2026 never moves your hand. Medium leans on comma-grouped thousands, and the comma is not on a keypad at all. The only Hard figures that break the pattern are the currency amounts, which need a comma and a dollar sign together.
Here is every non-digit character the drill produces, where it lives on the main keyboard, and whether a keypad can produce it:
| Character | Where it appears | Per 1,000 chars | Main keyboard row | On the keypad? |
|---|---|---|---|---|
, | Medium and Hard — thousands grouping | 59.4 | Bottom row | No |
. | Medium and Hard — decimals | 42.4 | Bottom row | Yes |
/ | Hard — dates | 50.4 | Bottom row | Yes |
- | Hard — reference numbers | 22.9 | Number row | Yes |
$ | Hard — currency amounts | 27.6 | Number row, with Shift | No |
The comma and the decimal point are both on the bottom row of the main keyboard, which is the worst possible place for them: your hand is two rows up on the digits, and a comma drags it two rows down and straight back. That is the whole of the movement difference between Easy and Medium — 0% of adjacent keystrokes change row on Easy against 24.5% on Medium, and the digits themselves are identical. The dollar sign is the other awkward one: it sits on the number row where your hand already is, but it needs Shift, so it costs a second finger rather than a journey.
A note on number formats, since ours is not the only one
This drill generates US-convention figures: a comma every three digits and a point for the decimal. That is a convention, not a rule, and it is worth saying so plainly on a site whose readers are mostly not American.
The international position is more permissive than most people assume. The General Conference on Weights and Measures resolved in 2003 that the decimal marker shall be either the point on the line or the comma on the line — both are correct, and which you use is a matter of local convention. On grouping, the standard goes further than either habit: digits should be separated into threes by a space, and neither a point nor a comma should be used for that job at all. By that measure both 197,790 and 197.790 are local conventions rather than the international form, which would be 197 790.
The Indian system groups differently again — three digits, then twos, so a figure of the same size is written 1,97,790. If that is what your work uses, the mismatch is real but it is smaller than it looks: what you are training here is the reach to each digit and the return from a separator, and neither of those cares where the separators fall. The format-specific muscle memory is worth building on records that match your own, which is what the data entry test is for.
How this is scored, and what it is not scored in
Scoring follows the site standard, unchanged: a standard word is five keystrokes, and a figure earns nothing unless every character in it matches. Four correct digits out of five is simply a wrong figure. That rule is stricter than it sounds on numeric text, because a digit carries no redundancy — there is no context to make a near-miss legible the way there is in a misspelled word.
The consequence worth flagging is at the easy end rather than the hard end. Each wrong figure costs a flat standard word, but an Easy figure is only 3.91 keystrokes once you count the space that commits it — fewer than the five you lose. So a mistake there removes more than the attempt contributed, and Easy quietly becomes the level that punishes haste hardest. The alpha-numeric page works that arithmetic through in full, since it hits the same wall; what it means in practice here is that a strong Easy result tells you about your accuracy, not your pace.
What this page does not report is KPH. Data entry roles quote keystrokes per hour rather than words per minute, and the two are related by a single short piece of arithmetic that the data entry test sets out in full, along with the thresholds job postings actually advertise. It is written once, there.
What the three levels change
All measured, none asserted — 500,000 figures sampled at each level by scripts/token-stats.mjs at build time.
| Level | Length | Digits | Row change | Keypad-only | Needs Shift |
|---|---|---|---|---|---|
| Easy | 2–3 | 100% | 0% | 100% | 0% |
| Medium | 4–7 | 89.8% | 24.5% | 65% | 0% |
| Hard | 5–10 | 84.4% | 23.8% | 76% | 2.8% |
Easy to Medium is the real step, and it is a step in punctuation rather than in digits. Figures roughly double in length, but the change that costs you speed is the arrival of the comma and the point: row changes go from 0% to 24.5%, and the share of figures a keypad can handle alone falls from 100% to 65%.
Medium to Hard is a step in length and variety rather than in movement — row changes barely move (24.5% to 23.8%) and, as set out above, keypad usability actually improves. What Hard genuinely adds is having to hold a longer and more structured figure in mind: a date has three parts and a fixed shape, a currency amount has four. We would rather describe the ladder that way than claim each level is uniformly harder than the last, because on our own measurements it is not.
Where to go next
This drill builds the movement. Everything that gives the movement a purpose — records, ledgers, field validation, the unit an employer will actually quote at you — lives elsewhere on the site, and deliberately so.
| What you found | Where to take it | Why there |
|---|---|---|
| Digits scattered among letters, not blocked | Alpha-numeric test | The opposite motion — the hand alternates instead of parking, and the keypad stops helping |
| You need the KPH conversion and field-accuracy rules | Data entry typing test | Owns the KPH arithmetic and what employers mean by an alphanumeric versus numeric score |
| Figures inside ledgers and financial prose | Accounting typing test | Ten-key as a career skill, and the divide-by-nine check for a transposed digit |
| Currency symbols and operators are the slow part | Symbols test | Isolates the shifted characters that this drill only sprinkles in |
| You still look down to find a digit | Numbers and symbols stage | Teaches the finger assignment from scratch before you drill speed on top of it |
| You are practising a lot and improving slowly | How to type faster | How to structure practice at all — written once there, rather than restated on each drill |
One thing worth doing before you leave: run Easy once and read the heatmap. With only ten characters in play, a weak digit cannot hide the way a weak letter can in prose, and most people find one or two that account for a surprising share of their errors. Those are worth ten focused minutes each rather than another general run.
Frequently asked questions
Should I take this test on the number row or on the keypad?
Either — the test scores whatever reaches the input, and it cannot tell which keys produced it. Which you should choose depends on what you are training for. If your target role quotes KPH and sits you in front of a full-size keyboard, use the keypad, and be aware that 35% of Medium tokens will force your hand off it anyway. If you work on a laptop, or your numbers arrive mixed into text, train the number row, because that is the skill that will actually be available to you.
Why does my speed collapse the moment a comma appears?
Because the comma is on the bottom row and every digit is on the number row, so a figure like 197,790 sends your hand two rows down and two rows back for a single character. On Easy this drill has no separators at all and the row-change rate between adjacent keystrokes is 0%. On Medium the comma and the decimal point push it to 24.5%. That jump is the entire difference in movement between the two levels — the digits themselves have not changed.
Why is Hard easier on a keypad than Medium?
It surprised us too, and it is a property of the characters rather than the difficulty. Hard is built from dates like 04/15/2026 and reference numbers like 555-0182, and both the slash and the hyphen are keys on the numeric keypad — so those tokens can be typed start to finish without the hand moving. Medium leans on comma-grouped thousands, and the comma is not a keypad key at all. The result is that 76% of Hard tokens are fully keypad-typable against 65% of Medium ones. Only the currency amounts on Hard break the pattern, because they need both a comma and a dollar sign.
My keyboard has no numeric keypad. Am I at a disadvantage?
Not for this drill, and not for most jobs. The number row is on every keyboard, which makes it the portable skill; a keypad is an advantage only in roles that key long columns of figures on a fixed desktop. It is worth knowing which one an assessment will put in front of you, because a keypad specialist on a laptop keyboard loses more speed than a number-row typist does on a full-size board. If a posting says "10-key by touch" it is telling you a keypad will be there.
Why does this test use commas for thousands when I write numbers differently?
Because it generates US-convention figures, and that is one of several conventions in use rather than a universal one. The international standard is explicit that the decimal marker may be either a point or a comma, and it goes further than most people expect: it says groups of three digits should be separated by a space, and that neither a point nor a comma should be used for that job. The Indian system groups differently again, after the first three digits and then in twos. If your work uses a different convention, treat this drill as training the reach rather than the format, and drill your own format where it appears in real records.
What is KPH and how does it relate to my score here?
KPH is keystrokes per hour, the unit data entry postings quote instead of words per minute. This page does not report it, because the test is scored in WPM like everything else on the site. The conversion is short and the data entry test works it through in full along with the common advertised thresholds, so rather than restate it here, use that page — it also explains why employers often quote two separate scores, one alphanumeric and one numeric.
Should NumLock be on?
Yes, or the keypad sends arrow keys and Home/End instead of digits, and on a typing test that means your keystrokes either do nothing or move the cursor somewhere unhelpful. It is worth checking before you start rather than thirty seconds in. If your keypad appears to be producing nothing at all, NumLock is the first thing to look at; on some compact keyboards the keypad is overlaid on the letter keys and needs a function-key combination instead.
Why is the mistyped-key heatmap unusually useful on this drill?
Because the character set is tiny. Ordinary typing spreads your errors across dozens of letters and punctuation marks, so no single key accumulates enough of them to stand out in a short run. Here the text is 89.8% digits at Medium, which means ten keys absorb almost every error you make and a weak digit shows up quickly and unambiguously. Word mode also judges each token on its own, so one slip cannot shift the comparison and manufacture a page of errors that are not real.
Why does my accuracy fall away late in a long numeric run?
Most likely because your fingers have been extended the whole time. On prose the hand returns to rest constantly between reaches; on this drill it does not, and Easy is the extreme case — the row never changes at all, so the fingers hold their stretch from the first keystroke to the last. That is a different kind of tiredness from the one a long prose test produces, and it shows up as drift rather than as sudden mistakes. If you are training for a numeric assessment, build up to the 5-minute duration rather than repeating short runs, because the endurance is a separate thing from the speed and only the longer runs reveal it.
How should I practise if my job uses different date or currency formats?
Drill the movement here and rehearse the format where it lives. What transfers between formats is the reach to each digit, the return from a separator, and the habit of not looking down — none of which care whether the year comes first or last. What does not transfer is the muscle memory for a specific punctuation pattern, and that is better built on the data entry test, where the figures sit inside records that look like the ones you handle rather than as bare tokens.
Sources and method
Nothing on this page describing the drill’s own text was written from memory. Figure length, digit share, punctuation frequency, row placement, row-change rate and keypad coverage are all output from scripts/token-stats.mjs, a build-time script that reproduces the engine’s generator and draws 500,000 figures at each level from a fixed seed. Because it re-runs whenever the site is rebuilt, the published numbers stay tied to the code that produces them, and the fixed seed keeps them steady between builds.
The keypad-coverage figures test each generated figure against the character set of a standard PC numeric keypad — the ten digits, the four operators and the decimal point. Two honest limits on that: keypad layouts vary at the edges, and in comma-decimal locales the keypad’s decimal key produces a comma rather than a point, which would reverse part of the result. The measurement describes the US layout this drill generates for. The finger assignments in the diagram are the ones most commonly taught rather than a published standard, because no body publishes one.
The prose and alpha-numeric comparisons are measured the same way from our own general passage corpus and our own generator, so the comparison is like for like. On number formatting we have cited the resolution and the brochure themselves rather than a secondary summary. We have deliberately published no 10-key speed or accuracy threshold here: no professional body sets one, advertised figures vary widely by employer, and our other pages take the same position rather than quoting a number we cannot stand behind.
- Resolution 10 of the 22nd CGPM (2003) — the decimal marker shall be either the point on the line or the comma on the lineBureau International des Poids et Mesures (BIPM) · link checked 4 Sept 2026
- The International System of Units (SI Brochure), 9th edition — formatting of numbers and digit groupingBIPM · link checked 4 Sept 2026