Say it clearly: trade-off sentences and spoken English

A design round is a spoken exam. The interviewer cannot see your thoughts. They can only score what they hear. Two candidates can reach the same design, and one still scores much higher, because that candidate said why.

A weak answer is a decision with no reason: "I will use four servers because it is the best." Or it is a long silence, then a finished drawing. A strong answer says the choice, what it solves, what it costs, and when the choice would change. It uses short sentences and numbers.

If English is not your first language, this lesson is for you. Most interviewers care more about clear thinking than perfect grammar. Short and simple sentences are easier to follow, and they are easier to say under stress.

At SDE-1 level you should give a choice and a reason. At SDE-2 level you should add the cost with a number. At SDE-3 level you should also say what would make you change the choice. This is our own guide, not an official rule.

Think of a doctor who explains a medicine. A weak doctor says: "Take this one." A good doctor says: "This medicine works fast. It can make you sleepy. If you do not feel better in three days, we will try another one." You trust the second doctor more. You know the doctor has thought about the cost, and has a plan if it fails.

In lesson 1 you learned to say, for every choice, what it solves, what it costs, and when you would change it. Now we make that a habit. A trade-off sentence has three short parts:

1. It solves ___. Name the need the choice meets, with a number if you can.
2. It costs ___. Name what you pay: cost, speed, simplicity, staying up or correct data (the list from lesson 1). Be honest.
3. I would change it when ___. Name a condition that you can check, with a number if you can.

The third part is the one most candidates forget. It shows that you know the choice is right only for these needs and these numbers.

Phrases for each moment

Use simple phrases. You do not need difficult words. Here are phrases for the moments of a round.

If you forget a word, describe it: "the part that stores the data". Then go on. The interviewer will understand.

Thinking aloud

Thinking aloud means that you say your thoughts as you have them. It is not a performance. It is the only way the interviewer can see how you think. Use this order:

1. Say the question you are answering: "I need to decide how many servers."
2. Say the options: "Three servers handle the traffic. Four servers also keep working when one breaks."
3. Choose, and give the reason with a number.
4. Say the cost.

Also say your arithmetic in steps: "500,000 orders divided by 7,200 seconds is about 69 orders per second." If you are stuck, say so, and say what you do know. Silence for a long time gives the interviewer nothing to score.

Some more tips for speaking:

• One idea in one sentence. Pause after each idea.
• Say numbers with units, and use both ways for big numbers: "5 lakh, that is 500,000." Some interviewers think in millions.
• Say "p ninety-nine" for p99. Say "three nines" or "ninety-nine point nine percent" for 99.9% (lesson 4).
• Say the time: "I will take about 5 minutes for requirements" (lesson 2).

We go back to the food-ordering app from lessons 3 and 4. The interviewer asks: "Why four servers?"

Assumptions (as in lesson 3):

• The busiest 2 hours send 1,389 requests per second. This is 10,000,000 requests / 7,200 seconds.
• One server answers 500 requests per second.
• The app must still work if one server breaks.

Answer 1: the weak answer.

"Four servers, because it is scalable and strong, and many companies do it."

This has no number. It does not say what the choice solves. It says nothing about the cost or the change. "Many companies do it" is not a reason.

Answer 2: a better answer, but not finished.

"Traffic is 1,389 requests per second. One server answers 500. So I need 3 servers, and one more, so 4."

This has numbers. But it does not say why one more, or what it costs. It also has no change.

Answer 3: the strong answer. First the arithmetic, then the three parts.

• 1,389 / 500 = 2.8, so 3 servers handle the traffic.
• If one breaks, 2 servers answer 2 x 500 = 1,000. That is less than 1,389.
• With 4 servers, 3 are left after one breaks. 3 x 500 = 1,500, which is more than 1,389.
• 4 servers instead of 3 is 4 / 3 = 1.33 times the cost, which is 33% more.
• Three servers would be enough only if traffic is 1,000 or less, because then 2 x 500 = 1,000 are left after one breaks.

Now say it:

"The busiest two hours send 1,389 requests per second. One server answers 500, so I need 3. If one breaks, 2 servers answer only 1,000, which is not enough. So I choose 4 servers. This solves the busiest two hours, even when one server breaks. It costs one server more than the minimum, which is 33% more than 3 servers. I would go back to 3 servers if the busiest traffic fell to 1,000 requests per second or less."

Each sentence is short. Each part has a number. The last sentence gives a check that anyone can measure.

Thinking aloud in a hard moment. The interviewer asks: "Is the average speed enough as a target?" You do not know what to say first, so you think aloud:

"Let me think. The average can hide slow requests. For example, 98 requests take 100 ms and 2 take 5 seconds. The average is 198 ms, which looks fine. But p99 is 5 seconds. So I would put the target on p99."

Here is the answer to the predict box. All three parts are missing. "It is the best choice" does not say what the choice solves, what it costs, or when it would change. Answer 3 shows a better version.

• Saying the trade-off aloud solves: the problem that the interviewer cannot see how you think. Your choice becomes something they can score.
• It costs time. A three-part sentence takes longer than saying only the choice. If you say it for every small choice, you lose the time needed for deep dives (lesson 2). Use it for the choices that matter: the ones that change cost, speed, or staying up.
• Thinking aloud costs focus. You must speak and think together. Practise on your own first.
• I would change it when: the interviewer asks a yes-or-no question. Then answer yes or no first, in one sentence, and add the three parts only if they want a reason. When time is short, say two parts: what it solves and what it costs.

"Good spoken English means long, formal sentences and difficult words." It does not. Look at the strong answer above. No sentence is longer than 20 words. A listener follows short sentences more easily. A small grammar mistake is fine if the meaning is clear. If you do not understand a question, ask again.

"I should stay quiet until I know the right answer." Silence gives the interviewer nothing to score. If you are unsure, say what you know, and what you are trying to find out. Thinking that you say aloud, even with a small mistake, is better than a perfect answer that nobody heard.

"I should say the trade-off sentence for every choice, in the same words." The three parts are a habit, not a script. If you repeat the same words for small choices, it sounds like a memorised script, and you lose time. Use the parts for the big choices, and say them in your own words.

1. Write the three parts of a trade-off sentence. Why should the third part contain a number or a condition?
2. A site gets 700 requests per second at its busiest time. One server answers 500. The site must still work if one server breaks. (a) How many servers do you need? (b) Write the trade-off sentence with numbers.
3. A candidate says: "Um, maybe we can use, like, many servers? It is robust, scalable and highly available. I think it is maybe good." What is wrong? Write a better version, using the numbers from question 2.