Esper Othman.
A designer's own mark is the first work a client judges. So I built mine the way I build for clients — decisions measured, alternatives rejected on evidence, and the whole system documented before it shipped.
- Role
- Identity design · Type craft · Print engineering
- Timeline
- Summer 2026
Set type isn't a signature
Most personal marks start and end the same way: pick a nice typeface, type the name, done. For someone selling design-and-build craft, that reads as a shortcut — type alone carries no evidence. The brief to myself: a wordmark with visible, defensible craft cut into it; a compact mark for the places a full name can't go; and stationery real enough to survive an actual press — with every decision honest enough to show a client how their own brand would be made.
Measured, cut, and archived
I treated my own identity like a client engagement with an unusually demanding client: every decision made against a measurement, every alternative kept in the archive with the number that rejected it.
- Set the wordmark in Fraunces (opsz 72, wght 500) — then stopped typing. Five decisions were made at point level in the actual outlines, four of them cut in: a drawn “th” ligature, the E's middle arm scaled to 96%, the O's counter rotated 2°, and a teardrop on the r's terminal that echoes the period. The fifth — an n exit-flick — was explored, measured, and rejected. Advance widths and the ink bounding box stayed identical to the typed original.
- Built the E mark by fusing the wordmark's ink-drop period into the letter itself: the drop replaces the middle arm, so it is load-bearing — remove it and the letter breaks. That is the structural test the rejected marks kept failing. 38 labeled explorations across the wordmark and five mark tracks sit in the archive, the strongest of them killed by a measurement — a drop that collapses to a blob at 32 px, a void too small for the spec's own floor.
- Engineered the business card as production files, not mockups: 85×55 mm with 3 mm bleed on 350 gsm stock, CMYK under SWOP with the teal held to a ΔE2000 of 0.31, total ink peaking at 215% — and every hairline modelled against ink spread, so the spec names what to watch on press: the finest marks halve at 0.02 mm of gain and are gone by 0.04.
- Shipped two editions that share one byte-identical front — the build aborts if the fronts ever differ — with a fully mirrored Arabic back: HarfBuzz shaping asserted letter by letter, set in Amiri and IBM Plex Sans Arabic. Real RTL, not flipped English.
- Put the mark inside the QR instead of beside it — 32 of 442 modules dropped and 52 touched, 11.8% against a failure ceiling measured at 15%, on ECC level H — then decode-tested the shipped PDF itself at 1200 dpi with 3× resolution headroom, as two of 58 automated checks that gate every release of the files.
It ships — and it shows its work
The identity is in use, not in a drawer: the favicon and app icons on this site are the mark's exports byte for byte, and the cards are press-ready files with their print spec. Just as important is what the archive holds — the rejected options labeled and kept, the decisive ones with the measurement that rejected them, versioned deliverables with checksums, and a one-sheet system covering construction, clear space, size tiers, colour states and misuse. A client reading this case study is reading my process on myself.
- Fraunces opsz 72 — exact outlines
- Python + fontTools
- HarfBuzz — Arabic shaping
- SVG / PDF masters
- CMYK — SWOP · 3mm bleed
- QR — ECC H · decode-tested