Updated 4 November 2026
If your developer hiring process sends candidates a coding test and then schedules another technical interview, you may be asking them to prove the same skill twice. A separate test is useful when it measures something the interview cannot. If you need to see whether a candidate can write working code and explain the decisions behind it, you may be able to assess both in one live technical interview. Fabric offers a live AI interview with a built-in IDE for that first technical screen. If you also need a separate assessment, confirm the available format and ATS integration before building it into the process. The choice is about the evidence your team needs, not simply reducing the number of rounds.
Before keeping or removing a stage, write down the hiring decision it supports. Are you checking basic coding ability, fluency in a required language, debugging, system design or the way a candidate explains trade-offs? If a separate test measures a distinct skill reliably, keep it. If the same skill is assessed again in the interview, consider combining the two and using the time saved to discuss the candidate's approach in more depth.
Two pipelines, one question
One stage, not two
The same developer shortlist, run two ways.
- Shortlisted
- OA invite goes out
- Handoff to the test platform
- Candidate finds an hour
- Result returns to the ATS
- Handoff to the interview
- Interview is scheduled
- Interview
- Shortlisted
- Interview, with the coding in the IDE
Illustrative. The right-hand route removes the separate test and the handoff between the test and interview.
What you can assess when the coding task is part of the interview
In a live technical AI interview with an IDE, the candidate writes and runs code while explaining their approach. The AI interviewer can ask why they chose a particular solution, introduce an edge case and probe how they respond. Your engineering team gets evidence from the working code and from the reasoning behind it. If the role requires system design, you can include that discussion in the same session, provided there is enough time and the right tools.
Fabric supports live coding tasks inside an AI interview. Candidates write and run code in a browser-based IDE, with execution against test cases. The AI can follow up on time complexity, edge cases and alternative approaches. A separate assessment may still be appropriate when it measures a distinct skill. If you need a standalone aptitude or MCQ test, confirm that exact format and how its results return to your ATS before building it into the process.
Choose the format after you have defined the skill you need to assess and the evidence your engineering team will use to make the decision.
What do you need to test?
When coding is included in the interview, it becomes part of Round 1 rather than a separate test followed by another interview. Tell candidates what language or environment they will use, whether they can run their code and how long the session is expected to last. Confirm the setup and candidate workflow for your role before sending invitations.
What a separate assessment adds to the process
A separate online assessment adds more than the time needed to solve the questions. Someone has to send the invitation, set a deadline, wait for completion, review the result and move the candidate to the next stage. The extra time varies by company. Measure it from your own workflow rather than assuming every assessment adds a fixed number of days.
Pin analysed 4,000+ active roles connected through its ATS integrations in a report published on 28 May 2026 and updated 7 September 2026. The median role had five evaluation stages configured, and 38.5% had seven or more. These are stages configured in an ATS, not proof that every candidate completed that many interviews or assessments.
Pin uses Ashby's 2026 recruiter productivity report estimate that one additional interview round can add roughly a week to a hiring process. That estimate is about an extra interview round, not a measured delay caused by an online assessment, so treat seven days as an illustration rather than an OA benchmark. To calculate your own delay, compare the date the assessment invitation is sent with the date the next interview is booked. Include weekends, reminder windows and time spent reviewing submissions.
The comparison below maps the steps in each route. Use it to identify which handoffs disappear and which assessment activities still need to happen.
The same shortlist, stage by stage
| Stage | Separate OA | Coding in the interview |
|---|---|---|
| Invite | OA link sent through the ATS or email | Shortlist moves to the interview |
| Assessment | Candidate completes the separate test | Candidate codes in the IDE during the interview |
| Result | Score and report return, depending on the integration | Interview report captures the coding and discussion |
| Next step | Interview scheduled after the test | Coding and follow-up discussion happen in one interview |
| Separate stage | One additional assessment stage | Removed only if the interview measures the same skill |
| Potential wait removed | Your measured OA-to-interview delay | Use your own process data; seven days is an illustration, not an OA benchmark |
Enter your own stage timings in the calculator. The starting values are examples; replace them with the median time from your last 20 to 50 candidates if you have enough records.
How much time does your OA stage add?
Replace each example with your own figure. This estimates the wait attached to the separate stage, not the number of candidates you should advance.
A handoff can cost candidates as well as time
HireVue's October 2025 Applicant Dropout: Completion Rate Drivers and Consequences white paper says poor integration between an ATS and other hiring steps can reduce completion by up to 20 percentage points in some instances. This is a company-published claim, not an independently established average, and the paper does not provide enough detail to treat it as a general benchmark.
The examples in the paper describe a familiar problem: a candidate finishes one step but then has to wait for another email, open a link with an unfamiliar company name or work out whether the process is complete. Those are avoidable points of confusion. If you keep a separate assessment, make the handoff immediate, explain what happens next and show the candidate that their submission was received.
The diagram below illustrates the handoff described in the white paper. It is an example of a possible failure, not a claim that every separate assessment creates this experience.
How a handoff loses candidates, as HireVue's paper describes it
The same paper reports that, in one company's data, applicants with higher assessment scores were more likely to drop out later in the process, and about two-thirds of those who left said they had accepted another offer. That is one company's observation, not proof that a coding test caused candidates to accept another job. It does underline why the time between stages should be measured and managed.
Can an unsupervised coding test still tell you what a candidate can do?
Karat's 2026 engineering-interview survey of 400 leaders across the US, India and China found that 71% said AI is making technical skills harder to assess. That's leaders' opinion, not a measurement of hiring outcomes. Karat's report also argues that AI weakens the signal from automated code tests and take-home projects because candidates can generate working code without showing how they reached the solution. That is the report's interpretation of the changing assessment landscape, not a controlled study comparing the predictive validity of different formats.
CoderPad's 2026 State of Tech Hiring survey included more than 650 developers, recruiters and hiring leaders worldwide. Separately, CoderPad's platform data shows global technical assessments up 48% compared with mid-2023. That is more tests, not proof of better hiring decisions.
This evidence is a reason to review how you assess technical skills. It does not prove that a live AI interview is more secure or more predictive of job performance than a separate coding test. If outside assistance is a concern, examine the controls and evidence for the format you choose. Fabric's approach to interview integrity is described on its cheating detection page. Whichever format you use, measure how long candidates wait between stages and whether the extra round improves the hiring decision.
What a live technical interview should not replace
A live discussion can add context to a coding task, but it cannot replace every technical assessment. Keep the distinction clear:
- Grading a code file it cannot run. The code needs to run in an execution environment. A spoken explanation alone cannot establish that the solution compiles, passes tests or handles the required inputs.
- Judging a diagram the candidate drew somewhere else. If the role requires architecture or visual modelling, ask the candidate to work with the relevant diagramming or whiteboarding tools. A verbal description can supplement, but not replace, the artefact when the artefact itself is part of the job.
- Treating a spoken answer or a transcript as the record of code that never ran. A candidate may explain a plausible solution without producing working code. Assess the implementation and the explanation as separate pieces of evidence.
- Setting one coding bar for every seniority level. Define expectations for the level. A junior candidate may need to demonstrate fundamentals; a staff engineer may need to explain architecture, reliability, maintainability and trade-offs.
How to run the combined process fairly
The ATS
Fabric's integrations page lists 31+ connections across the hiring stack, not 31+ ATS integrations. Its current ATS & HRIS section lists Greenhouse, Ashby, Lever, Workday, BambooHR, Keka, Darwinbox, Workable, Zoho Recruit, GreytHR, HROne and Freshteam, and describes two-way sync for jobs, candidates, stages, scorecards and interview transcripts. Confirm the exact connector, fields returned and plan eligibility for your ATS before assuming a coding interview or assessment result will sync back. See the current integrations list.
Do not assume every assessment platform sends the same information to every ATS. HackerRank's documentation for its Zoho Recruit integration specifies which interview and test results are transferred and which details are not. Check the current documentation for the system you use, then test whether recruiters receive the code, scorecard, question-level results and reviewer comments they need.
Integration coverage can differ between interview and assessment products. HackerRank's current Workable integration documentation lists support for tests, including invitations, total scores and report links. It does not list interview support, question-wise scores, reviewer comments or code display in Workable. Check the current connector and plan for your own workflow before relying on a particular result or report field.
Before choosing a combined round, check what the ATS receives: the overall score, question-level scores, code submission, interviewer notes, recording and report link. A single score without the underlying evidence may not be enough for a hiring manager to make or review a decision.
The rubric
Set the rubric before candidates start. Define the problem, allowed tools, time limit, test cases and criteria for the role and seniority level. Score correctness, approach, readability, debugging and communication separately where relevant. Use the same criteria for every candidate applying to the same role.
Keep the recruiter or engineering reviewer accountable for the decision. Review the code and the interview evidence, check unusual scores and allow a candidate to be reconsidered when the evidence is incomplete. If automated rejection is enabled, test the thresholds against real examples before the drive and keep an auditable record of the criteria used.
What you tell candidates
Tell candidates in the invitation that the coding task will take place during the interview, what language or environment they will use, whether they can run their code and how long the session is expected to last. Do not let them discover during the call that a separate test is not coming.
Fabric describes the interview as a live two-way video interview that can be joined in a phone browser without an app. It also says candidates can rejoin after a dropped connection and that AI disclosure and recording consent are built in across 35+ languages. Test the coding experience on the devices and networks your candidates use: an IDE may have different practical requirements from a standard video call.
Use this checklist before opening the drive. Resolve any unanswered item with your hiring manager, engineering reviewer or the Fabric team before inviting candidates.
Before you remove the OA stage
of 6 ticked
Pilot the combined interview on one developer role
Choose one open developer role and run a small pilot. Compare the combined interview with your current assessment route on completion, time between stages, reviewer agreement, code quality and candidate feedback. Keep the role, rubric and decision criteria consistent so you can see whether the new format gives your team better evidence with fewer handoffs.