How to read this. Use this sheet after a mock round. A friend, or you alone with a recording, gives each of six rows a score from 1 to 4. The scores are my own practice tool. They are not a company rubric and they do not predict an interview result.
1. How to use it
In short: Score the evidence you heard, not the answer you meant to give.
Fill in one row at a time, right after the round. Write one example from the round next to each score. A score without an example is a guess.
The sheet follows the five steps of the method, with minutes of 4, 10, 8, 18 and 5. Communication is the sixth row. It covers the whole round.
Use it with the mock interview script. The interviewer fills it in. If you practice alone, listen to your recording and fill it in yourself.
2. What each score means
| Score | Meaning in plain words |
|---|---|
| 1 | Missing. The candidate did not say it, or said it only after a hint |
| 2 | Named, not explained. The candidate used the right word but not the reason |
| 3 | Clear. The candidate said what, why, and what fails otherwise |
| 4 | Clear, and the candidate gave the cost of the choice or a way to undo it, without being asked |
3. The score sheet
| Row | Minutes | 1 | 2 | 3 | 4 | Score | Example from the round |
|---|---|---|---|---|---|---|---|
| Scope and constraints | 0 to 4 | No questions, or accepts every request | Asks questions, does not say what is out | Says what is in and out, and why | Also says what the phone does to the design | ||
| Data and contracts | 4 to 14 | Talks only about the server | Names a database | Says which copy the app trusts and which ID makes a retry safe | Also says who sets the order and how a repeat arrives | ||
| Client structure | 14 to 22 | The ViewModel owns everything | Draws layers with no owner | Says who owns the work and why it must live longer than the screen | Also names the first idea that fails and the rule that follows | ||
| Make the hard part work | 22 to 40 | Only the normal case | Names tools such as WorkManager without the reason | Walks the send path and says what fails | Also says the first idea, why it fails, and the cost of the fix | ||
| Break it, ship it, roll it back | 40 to 45 | No failures named | Names one failure | Names the failures, what the user sees, and a kill switch | Also says what to watch after release and how old app versions are handled | ||
| Communication | All 45 | Talks without structure, runs out of time | Follows the steps but misses the time calls | Keeps to the minutes and says assumptions aloud | Also answers a cut-off, then returns to where it was |
The descriptions come from the five steps and the "Common weak answers" in the chat app case study. The row for step 4 matches the pattern in the case study: first idea, why it fails, what I would say. For a worked example, read the teardown.
4. How to read the total
In short: The total is a rough guide. The lowest row matters more than the sum.
Six rows with 4 points each give a range of 6 to 24. This is my own reading, not a standard:
| Total | What I take from it |
|---|---|
| 6 to 11 | Learn the method first. Read the chat app case study before the next round |
| 12 to 17 | The method is there. Work on the rows with a score of 1 or 2 |
| 18 to 21 | Most rows are clear. Practice the cost of each choice |
| 22 to 24 | Strong round. Repeat with a harder case, or add a changed constraint |
Look at the lowest row first. A total of 18 with one row at 1 is a worse sign than 16 with all rows at 2 to 3. The 1 shows a step you skipped. Skipping a step is the easiest problem to fix.
Compare rounds over time. Write the date and the total at the top of each sheet. A rise of two or three points over a few rounds is a better sign than a single high score.
5. Questions to ask after the round
- Which row had the lowest score? Which case section covers it?
- Did I run past a time call? Which step lost time?
- Did I say the first idea and why it fails, or did I skip to the answer?
- Did I name a kill switch?
Then pick the next case from the study plans.
More in the practice kit
Words on this page (3)
- Kill switch
- A remote flag that turns a feature off at once. It is the one rollback that reaches apps that are already installed.
- ViewModel
- A class that holds the state of one screen. It survives a screen rotation.
- WorkManager
- Android's tool for background work that must run later. The work survives when the app closes or the phone restarts.
Written by Manuel Lorenzo. The site is drafted with AI assistance and edited and reviewed by him.