TL;DR

A technical interview is an assessment method used to evaluate a candidate's practical skills in a technical domain. In software engineering, this typically means coding problems, system design discussions, or both. The format is widespread, heavily debated, and only loosely correlated with actual job performance.

What Technical Interviews Actually Test

The standard software engineering technical interview tests a candidate's ability to solve algorithmic problems under time pressure, usually in a language they choose, sometimes on a whiteboard, more often now on a shared editor like CoderPad or HackerRank.

The canonical format: a problem is posed (reverse a linked list, find the shortest path, design a rate limiter), the candidate works through it while explaining their reasoning, the interviewer asks follow-up questions, and performance is evaluated across correctness, efficiency, communication, and problem-solving approach.

System design interviews work differently - a candidate is asked to design a large-scale system (design Twitter, design a URL shortener) and is evaluated on their understanding of distributed systems, trade-off reasoning, and architecture judgment. These are typically reserved for mid-level to senior candidates.

Outside software, technical interviews take different forms: data analysis case studies, financial modeling exercises, network troubleshooting scenarios, or clinical skills assessments. The term covers any evaluation of domain-specific ability.

Why It Matters

Hiring decisions at technology companies have historically relied heavily on technical interviews, and the practice has spread well beyond Big Tech. The logic: in a role where technical skill is the job, you should assess technical skill directly.

The counterargument is well-documented. A 2019 study published in IEEE Transactions on Software Engineering found that typical coding interviews activated stress responses similar to public speaking - which measures anxiety tolerance as much as programming ability. Research from Triplebyte (published before the company's decline) found low correlation between performance in algorithmic coding interviews and on-the-job effectiveness.

Despite the critique, technical interviews persist because the alternatives aren't obviously better. Work sample tests require more setup. Take-home assignments raise fairness concerns about who has time to complete them. Trial periods are slow and expensive. Structured behavioral interviews work for many roles but don't directly surface technical skill.

In Practice

Most companies doing technical hiring use a combination of formats: a recruiter screen, then a technical phone screen or asynchronous coding challenge (HackerRank, LeetCode-style), then a multi-stage on-site or virtual loop that includes coding, system design, and behavioral components.

FANG companies (Meta, Amazon, Apple, Netflix, Google) use algorithmic problem-solving as a primary filter. Preparation culture around this has spawned an entire industry: LeetCode has millions of users, Neetcode.io provides structured problem sets, and candidates routinely spend 200-400 hours preparing for a single company's interview loop.

Smaller companies and startups are more likely to use practical assessments - build a feature, debug a broken codebase, review a pull request - which tend to correlate better with actual work.

Calibration and consistency are common failure points. When five interviewers use different criteria and evaluate independently, the aggregated signal is noisy. Structured rubrics and interviewer training directly improve signal quality.

The Calibration Gap

The technical interview is only as good as the people running it. An interviewer who hasn't thought through the evaluation rubric, who leads the candidate toward their preferred solution, or who conflates 'thinks like me' with 'technically strong' introduces bias at the point that matters most. Training interviewers takes time that most engineering organizations resist spending, which is the main reason technical interview quality varies so widely within a single company.