Why this update matters right now
The numeric keypad finally has a fresh, published international standard: ISO/IEC 9995‑4:2025. The 4th edition (published November 19, 2025) clarifies the numeric section of keyboards and, crucially, divides it into zones—ZN0 (numeric zone) and ZN1 (function zone)—with guidance on arrangement, number and location of keys, and intended applications from data entry to POS and telephony. For anyone running 10‑key tests, that’s a timely reason to recalibrate content, scoring, and even the hardware assumptions you build into your benchmarks. (iso.org)
What’s in ISO/IEC 9995‑4:2025 (in plain English)
At a high level, Part 4 now explicitly defines the numeric section’s zoning:
- ZN0: the numeric zone, intended for rapid entry of digits and related symbols.
- ZN1: the function zone, intended for operations that complement numeric entry.
The abstract also states the standard specifies the arrangement, number, and location of keys and the allocation of functions—giving test builders a neutral baseline to validate against, regardless of manufacturer quirks. In short: your test setups and “compliance checks” can now reference concrete zones and arrangements, not just tradition. (iso.org)
Layout affects performance: telephone vs calculator evidence
If your candidates switch between a telephone-style keypad (1–2–3 on the top row) and a calculator/computer numpad (7–8–9 on the top), layout interference is real. In a controlled study (N=57), participants were more accurate on calculator-style keypads overall; they were specifically slowed when entering generic numeric strings on a telephone-style layout, and responses on the middle row (4–5–6)—identical in both layouts—were fastest and most accurate. Memory for the exact calculator layout was also worse than for the telephone layout, underscoring how familiarity doesn’t always equal explicit recall. Use this as a design hint: don’t assume layout neutrality in your tests. (sciencedirect.com)
Historically, Bell Labs’ human‑factors work in the 1950s/60s found the telephone layout slightly faster than then‑common adding‑machine layouts—part of why the phone pattern won for dialing tasks. But for today’s desktop data entry, modern experiments show task‑specific performance differences, with calculator layouts advantaged for generic numeric entry. Testing both (or at least being explicit about which your test uses) is smart. (en.wikipedia.org)
Redesign your 10‑key tests for 2026 hiring
Here’s a pragmatic plan you can ship this quarter.
1) A/B test three input modes
- Classic numpad (calculator layout). Baseline for numeric throughput.
- Top number row. Common on laptops and compact boards; expect slower KPH but less hand travel if the mouse is central.
- One‑hand “layer” numpad. With QMK/VIA‑capable boards, you can map a full numpad to a momentary or locked layer under one hand (e.g., OSL/TG/Layer Lock), then compare speed/accuracy vs a physical numpad. Publish your deltas. (docs.qmk.fm)
Tip: Randomize which mode appears first to control for warm‑up effects. Capture device type and whether a physical numpad was available.
2) Score in KPH (with transparent math)
KPH is just characters‑per‑minute scaled to an hour: KPH = CPM × 60. Show this formula in your UI and export. Many job postings still prefer KPH, and candidates recognize it. (typespeedtest.com)
What’s “good”? For data‑entry‑style roles, 8,000–12,000 KPH (≈27–40 WPM using the 5‑chars‑per‑word convention) is a typical expectation band—state your target so candidates know what to shoot for. (typingtest.now)
3) Use stricter, layout‑aware error weights
Not all mistakes cost the business equally. Consider these example weights in your net KPH:
- Substitution (wrong digit): −1 per wrong keystroke
- Omission/Insertion (missing/extra digit): −2 each (harder to detect downstream)
- Transposition of adjacent digits: −3 per transposed pair (often invisible to spell‑checkers and costly in numeric IDs)
- Wrong delimiter in formatted strings (e.g., phone or credit‑card pattern): −2
Why tougher on transpositions and formatting? The research shows row‑specific slowdowns and error patterns when layouts mismatch; emphasizing penalties that reflect real‑world consequences encourages accuracy with structure, not just raw speed. Publish your weighting so results are auditable. (sciencedirect.com)
4) Include a telephone‑layout variant when relevant
If your workflows involve dialing or IVR codes, add a telephone‑layout task to quantify any switch‑cost between patterns. That gives hiring managers data for roles that genuinely span both worlds. (en.wikipedia.org)
Build a free validator for ISO/IEC 9995‑4:2025 (and link it in your post)
Ship a small, open web tool that:
- Checks that a keyboard (physical or virtual) declares a numeric section divided into ZN0 and ZN1.
- Verifies presence and placement of required numeric keys in ZN0 and function keys in ZN1, and that their reported scan codes map as expected.
- Flags non‑standard placements and labels (informationally—do not reproduce paywalled details; use the public abstract as your scope line and ask vendors to confirm full compliance). (iso.org)
Implementation sketch:
- Client: WebUSB/WebHID to read key events; fallback to a virtual on‑screen keyboard grid.
- Test: Prompt the user to press each expected key; record the physical row/column (if exposed), OS code, and timing.
- Output: Pass/flag list with ZN0/ZN1 grouping, a printable report, and a JSON export you can attach to candidate records.
What to measure (and publish) in your benchmark
- Gross vs net KPH (show math). (typespeedtest.com)
- Accuracy by error type (substitution, insertion, omission, transposition).
- Row‑level timing for numpad tasks (top/middle/bottom); report if top/bottom rows are disproportionately slow, which often signals layout interference. (sciencedirect.com)
- Mode deltas: numpad vs number row vs layer numpad.
- Learning effect: best of three short runs per mode; compare run 1 vs run 3.
- Device notes: physical numpad present? external keypad? layer lock enabled? (develop-docs.qmk.fm)
Practical tips for test builders and admins
- Tell candidates which layout a section uses and whether Num Lock or a layer toggle is required; give a 20–30‑second warm‑up in that mode.
- For one‑hand layers, default to a lock/toggle so candidates aren’t forced to hold a modifier while keying strings. (develop-docs.qmk.fm)
- Mix unformatted numeric strings and formatted ones (e.g., (123) 456‑7890, 16‑digit IDs) to surface delimiter accuracy and chunking behavior. (sciencedirect.com)
- Add basic anti‑cheat: flag impossible inter‑keystroke timing patterns or copy/paste artifacts; extreme uniformity is a tell. (typingtest.now)
- Document your scoring policy and publish a sample calculator so candidates can self‑check before a proctored attempt. (typespeedtest.com)
Bottom line
The 2025 refresh of ISO/IEC 9995‑4 gives you a concrete zoning model for the numeric section. Use it to validate hardware and test UIs, then modernize your assessments with KPH‑first scoring, transparent error weights, and A/B tests across numpad, number row, and one‑hand layer numpads. Your resulting benchmark post—and the free validator—will help candidates prepare, help hiring teams interpret scores, and align your platform with the latest standard. (iso.org)
