Nisf A Matrimonial App, Redesigned
Nisf is a matrimonial app for practising Muslims: no photographs, a profile read across five tabs, and a filter layer no general matrimonial site has any reason to build. This is the redesign, starting with the app as it stands.
It is live, with a real user base across Kerala. That is the constraint behind most of what follows: the redesign lands on people who already know their way around, so it has to be recognisable on the first open.
A profile is a code, F‑2249, with an age, a district and a line about how the person covers. No photograph. So the work of the app is helping you decide who is worth a request from that alone.
That decision is where the app is unusual. Under the filters any matrimonial site has (age, location, education, income, height, language) sits a second layer that a general one has no reason to build: path and aqeedah and madhab, how much of the Qur’an someone has memorised in juz, the level of their tajweed, their Arabic, their view on photography of people, the courses they have finished inside the app itself. A profile is read across five tabs, and one of them is Deen.
Around that sits the machinery: a wali’s verification as a status you can see before you write, requests that live in four states, and a purchase wall somewhere in between.
The app as it is today
Thirty-one screens of the live app, running now on real phones. This is the thing being redesigned, not the redesign. Almost every decision further down this page is an answer to something in here.
Where it starts
Two decisions come before any screen gets drawn: what it is set in, and what colour it is. Every screen after them inherits both, so they are cheap to change now and expensive to change later.
Onest, for three reasons. Its letters are open and its x-height is tall, which is what keeps small type legible, and this app is mostly small type, since a filter is a wall of chips. It ships as one variable file covering 100 to 900, so the whole hierarchy comes out of a single download instead of five. And it is friendly without being childish, which matters in an app that is asking people personal questions.
#65558F is not a new colour. It is the purple the app already wears (the icon is that colour), and keeping it means the redesign arrives as the same app rather than a different one that took the name.
It also sits at a useful lightness. White on it clears AA at 6.5:1, and it clears AA again as text on the near-white surface, so one colour does both jobs: the fill of a button, and the ink of a link. A system that needs a second, darker purple for text has two colours to keep in step forever.
Everything else is that colour at another tone. On a near-white surface the tones do the separating, so a card can sit apart from what is behind it without a border drawn around it.
The navigation
The old bar was Material’s own: five tabs with a pill sliding in under the active one. The reason to leave it is that the app ships on both platforms and that bar is Android’s. What replaces it is the shape iOS and Android already agree on, so neither side has to learn anything.
Five tabs became four. Courses and Coaching were one intent wearing two names (somebody who wants to learn does not care which shelf it is filed on), so they are a single tab called Learn. Four tabs also give each one a wider target than five did.
Every icon keeps its label. An icon on its own has to be learned first; a word does not.
And the icons are drawn for this app rather than lifted from a set, so their weight and their corners match everything around them. A borrowed icon set is the usual reason an app looks slightly assembled.
The top of the list
Search, filter and the status chips are all doing one job: narrowing a list of strangers down to a few worth reading. The old screen spread that job over three rows, put sort in the middle of it, and left the last chip cut off behind a scroll arrow.
So filter moves up beside the field it belongs to, and the chips get the width to all fit. Everything that narrows the list is now one tap away, and none of it is hidden.
The placeholder does a quiet job as well. It names what the field takes, id or location, because a box that only says Search has to be guessed at, and a guess that returns nothing reads as an empty app rather than a wrong query.
The study goes under here.