In short
- Keep the first attempt intact so the review assesses what you did.
- Score evidence you can locate, such as a timestamp, a test result or a code location.
- Separate an untested claim from a demonstrated defect before you judge a session.
- Repair 1 gap, then retry with a changed constraint to see whether the reasoning transfers.
- Treat a clean recording as evidence for that session only, not as proof of readiness.
You read an interview answer and it looks familiar. You close the page and explain it aloud. Halfway through, you discover that you cannot say who owns the state or what the first test should prove.
That is useful information. Another hour of reading can hide it; a recorded attempt preserves it.
Good iOS interview practice alone produces something you can inspect, such as code, a design, an explanation and the moment a requirement changed. You then repair a specific gap and test whether the repair transfers to a different version of the problem.
This is a proposed practice method, not a validated readiness assessment or a promise of interview success. A partner can add independent judgement and unpredictable follow-ups. You can still do valuable work before that partner is available.
Decide which round you are rehearsing
Ask the recruiter about the actual format, editor, permitted tools and expected scope. If the information is incomplete, record the assumptions behind your practice.
| Upcoming round | Useful solo output | What it cannot fully replace |
|---|---|---|
| Algorithm coding | Unseen solution, tests and complexity explanation | Another person's probing and hints |
| App-oriented coding | A small working feature and failure check | Full assessment of collaboration |
| Swift or SwiftUI discussion | A recorded prediction and supporting example | Independent judgement of technical depth |
| Mobile system design | Requirements, contracts and a recovery walkthrough | Unscripted negotiation of scope |
| Past-project discussion | Specific decisions and honest contribution | Verification of the actual history |
Choose 1 round for today's session. A random mixture can make it difficult to identify what improved.
Keep the first attempt intact
Create a folder for the session with the prompt, recording, code or diagram and a short review note. Record which references were available during the attempt.
Do not edit the first answer into a cleaner transcript and then assess the edited version. The pauses, assumptions and abandoned implementation choices are part of the evidence you need.
Use 2 separate environments when practical. First an attempt with the rules of the real round, then a repair session with references available. A closed-book rehearsal can reveal recall problems, but it should not be presented as mandatory if the employer permits other tools.
The method in 6 steps
- Prompt. Pick 1 prompt for 1 round before you open any reference.
- Answer without help. Use the rules and time limit of the real round.
- Record. Capture screen and voice, and keep the first attempt unedited.
- Score. Fill in the scoring sheet below, with evidence for every score.
- Correct. Repair 1 gap with references open.
- Retry. Change the constraint and attempt a different version of the problem.
A worked example of the method
Take this prompt for a 35-minute attempt. Add a sort control to a product list. The user can select ascending or descending price, some products have no price and the rows must keep their identity. Allow 3 minutes to clarify, 14 to implement and 5 to verify. A second requirement, written beforehand and hidden, is revealed at minute 22. The last 13 minutes go to the follow-up and the review.
Say the contract before you write the comparator. For the baseline, assume 1 currency, unknown prices last in both directions, an ID-based tie-break for equal prices and a complete list already on the device. Say where the selected order lives and where the visible list is derived. “I'll sort the array” does not yet define a result.
Then check the awkward values. Use known, equal and missing prices in both directions, an empty list and a list with 1 item, and confirm that each ID stays attached to its product. Explain why the comparator is a valid ordering. Equal elements cannot each be “less than” the other, and missing-value handling must stay coherent, because Swift's sorting API specifies requirements for the ordering relation.[1]
At minute 22 the hidden requirement arrives. The catalogue has many pages and only the first is loaded, but the control must order the whole catalogue. Pause before typing. Sorting the loaded slice says nothing about products the device has not received.
Name the contract you now need from the backend, namely supported sort options, deterministic page ordering and cursors tied to the query. Changing the sort order should invalidate the old paging state, so an old response cannot append into the new session. You do not have to build a backend in the remaining time. Name the changed responsibility and sketch the state transition. If the API lacks global sorting, say so rather than presenting a locally sorted first page as the catalogue.
A review note for this attempt might read like this.
02:10, I started the comparator before deciding where unknown prices belong. 08:40, known prices sorted correctly. 19:05, descending order moved unknown prices to the front, contradicting the contract. 24:20, I proposed sorting every page independently. Global catalogue order was not established.
The repair has 2 parts, and they need not become 1 oversized session. First fix and check the missing-price rule. Then rehearse the difference between loaded-slice sorting and server-defined pagination.
If you want this kind of rehearsal with a rubric and a changed-constraint retry, iOS Interview Training is the structured route, with Guided Practice.
Score the attempt on a sheet
Copy the sheet below into your notes and fill it in after every attempt. A row may only carry a score when its evidence cell is filled. If you cannot point to a timestamp, a test result or a code location, leave the score blank.
| Dimension | Score 0 to 3 | Evidence | Retry date |
|---|---|---|---|
| Contract stated before coding | |||
| Code matches the contract | |||
| Identity and state ownership | |||
| Awkward cases checked | |||
| Adapted to the changed requirement | |||
| Explained aloud without notes | |||
| Open question named |
Use 0 for not reached, 1 for needs repair, 2 for prompted (you needed a hint or a reference) and 3 for demonstrated. These are descriptions for your own notes, not a hiring scale. For how interviewers record their own notes after a round, read what interviewers write in the debrief.
Separate an untested claim from a demonstrated defect. If you never attempted the follow-up, mark it 0 and write “not reached” as the evidence. Do not manufacture a passing score because the design feels familiar. A score of 7 out of 10 for communication cannot tell you which line to change, so give every score a location.
Set the retry date in the last column on the day you fill in the row. A row with no date is a gap you have agreed to ignore.
Repair the gap before repeating the whole round
Use references during repair. If the issue is the comparator, reduce it to a pure function with a small fixture. If the issue is identity, build a minimal list where you can see state follow the wrong row. If the issue is architecture, draw the ownership boundary and practise explaining it in 2 minutes.
For SwiftUI identity, Apple's explanation connects persistent view state to identity.[2] Check the relevant framework behaviour, then return to the original symptom. Reading an entire book chapter is optional if a small focused example answers the question accurately.
Write down what the repair established and what it did not. A unit check for ordering does not establish navigation restoration or production performance.
Retry with a new problem that uses the same reasoning
For the next attempt, sort articles by publication date, with missing dates last. Then introduce a server-ranked feed whose order cannot be reproduced locally.
Keep the skill you are testing and change the surface details. You want to see whether you can define the ordering contract, preserve identity and recognise when the data boundary moves.
Repeating the exact fixture immediately can show whether you remember a correction. A changed problem can provide additional evidence about whether you understand it.
Build a week around observed gaps
Use 1 diagnostic session, 1 focused repair and 1 changed-constraint retry as a small cycle. Add a different round only when you have time to inspect its result.
If your loop contains algorithms, schedule algorithm practice explicitly. If it includes SwiftUI diagnosis, schedule a focused SwiftUI rehearsal. For a full mixed rehearsal, run 1 timed session that covers identity, concurrency and recovery together. For scheduling the days before a loop, see the 14-day senior mobile interview prep plan.
Bring in a human reviewer when you need independent judgement, especially if your own review repeatedly cannot distinguish a clear answer from a vague one. Solo practice can prepare that session by giving the reviewer a specific gap to examine. Keep algorithm drills and recorded sessions in separate blocks of the week, so the scoring sheet only covers rehearsals.
Choose structure if assembling the sessions is the bottleneck
iOS Interview Training is $99, paid once per platform, and combines the book and companion code with Guided Practice. You can read the free sample before buying, to decide whether its route and format help you practise consistently.
Keep today's first attempt regardless of which resource you use. The next session should produce a correction you can point to, not merely a more confident feeling about the topic.
Sources
Questions readers ask
© 2026 Mike Salari, Interview Runtime. All rights reserved. Quoting a short excerpt with a link to this page is welcome; republishing the article, in full or in part, is not.



