Salesforce Developer Technical Assessment: How to Test Candidates Properly (2026)
A Salesforce developer technical assessment that works in 2026: what 126 Apex job specs demand, a live 90 minute format, and a scoring grid to reuse.
Cyril Hauchard · 13 October 2026 · 11 min
A Salesforce developer technical assessment should run for about 90 minutes, be watched live by a human, and cover four things: the data model, SOQL, a small Lightning Web Component build, and how the candidate reads Apex they did not write. The unsupervised take home assignment that most European companies still send out in 2026 measures almost nothing, because 84 percent of developers now use or plan to use AI tools in their workflow and the file you get back is no longer proof of who produced it.
Methodology. Advert counts, salaries and day rates are advertised figures for the six months to 13 October 2026, from IT Jobs Watch: Salesforce Developer permanent (201 adverts, 137 salaries quoted), Salesforce Developer contract (197 adverts, 135 day rates), Apex (126 adverts, 85 salaries), Lightning Web Components (134 adverts) and SOQL (58 adverts). A co-occurrence is one advert citing both Apex and that skill, so the percentages describe what employers write, not a skills gap survey. AI usage comes from the 2025 Stack Overflow Developer Survey (33,662 answers on AI usage, 26,004 from professional developers). Bad hire costs use the SHRM 2026 Recruiting Executives Benchmarking report and United States Department of Labor guidance. Conversions use European Central Bank reference rates of 12 October 2026, one euro equals 0.84478 pound. The exercises and the grid come from Apiire's European Salesforce desk: 800 or more candidate interviews between January 2024 and August 2026, 47 permanent placements followed up at six months, and 62 searches timed from signed brief to signed offer.
Why the old Salesforce developer technical assessment stopped working in 2026
Two things changed at once, and together they broke the format. The first is AI in the editor. In the 2025 Stack Overflow Developer Survey, 84 percent of respondents use or plan to use AI tools in their development process, up from 76 percent a year earlier, and 50.6 percent of professional developers use them daily. Positive sentiment fell over the same period, from above 70 percent in 2023 and 2024 to 60 percent, so this is habit rather than enthusiasm. An unsupervised exercise handed to that population comes back correct almost every time, so it no longer discriminates.
The second is competition. UK permanent adverts for Salesforce Developer reached 201 in the six months to 13 October 2026 against 72 a year earlier, and contract briefs 197 against 90. The median advertised salary moved to 65,000 pounds, about 76,900 euros, up 8.33 percent, with the 75th percentile at 80,000 pounds and a contract median of 525 pounds a day, about 621 euros. When three companies chase the same shortlist, a five hour unpaid exercise is not a filter, it is a withdrawal mechanism. Across 62 permanent Salesforce searches our median is 43 days from signed brief to signed offer, and the ones that slipped past 60 days almost always carried an extra unsupervised test stage.
What Salesforce job specs actually ask for, and therefore what to test
Design the assessment from your own job spec, not from a generic coding test library. Of the 126 UK permanent adverts citing Apex in the six months to 13 October 2026, here is what employers wrote alongside it, and whether 90 minutes can honestly prove it. If your spec is vague about the role, fix that first with our Salesforce developer job description template: a test can only check what the spec commits to.
| Skill cited alongside Apex | Adverts | Share of Apex specs | Can a live 90 minute session prove it? |
|---|---|---|---|
| Lightning Web Components | 106 | 84.13 percent | Yes, with one small component build |
| CI/CD | 70 | 55.56 percent | Partly, ask which tool deploys their metadata |
| REST integration | 56 | 44.44 percent | Yes, as a callout design question |
| SOQL | 54 | 42.86 percent | Yes, the highest signal per minute on this list |
| Salesforce Service Cloud | 49 | 38.89 percent | No, settle it in the reference call |
| Agile | 46 | 36.51 percent | No, test it in the hiring manager interview |
| Data modelling | 41 | 32.54 percent | Yes, and it predicts seniority best |
| SOAP | 41 | 32.54 percent | No, only relevant with legacy middleware |
| DevOps | 36 | 28.57 percent | Partly, ask who approves their pull requests |
| Git | 28 | 22.22 percent | Yes, by reading one of their own pull requests |
| AI | 27 | 21.43 percent | Yes, by letting them use it and scoring the review |
| Clean code | 20 | 15.87 percent | Yes, through a code review exercise |
Read the last column first. Only five of these can be proved properly in a live session, and they are the five that predict delivery: data modelling, SOQL, Lightning Web Components, integration design and code reading. The rest belong elsewhere. Service Cloud depth belongs in the reference call, Agile in the manager interview, and a CI/CD claim is settled by asking which tool deploys their metadata and who approves the pull request.
The cheapest filter we know. Before any code, ask the candidate to describe the data model of the last thing they shipped: objects, relationships, and why. In our interview log that one question separates levels more reliably than any coding exercise. People who built the thing describe trade offs. People who watched it being built describe features.
A 90 minute Salesforce developer technical assessment that works
This is the format we run with clients: one sitting, screen shared, with the hiring manager or a senior developer present. AI assistants are allowed and declared, because that is how the person will work on Monday. You are not scoring whether they produce code, you are scoring whether they can defend it.
| Stage | Time | What it actually proves | Fail signal |
|---|---|---|---|
| Data model and requirement walkthrough | 20 minutes | Judgement, scoping, ability to say no to a bad requirement | Jumps straight to fields without asking what the process is |
| Code review of a broken Apex trigger | 25 minutes | Governor limit awareness, bulk thinking, peer communication | Misses the query inside the loop, or cannot say it kindly |
| Small Lightning Web Component build | 30 minutes | Hands on build, stubbing, test reflex under time pressure | Silent struggling, no stubs, no test, no narration |
| Debrief on their own production work | 15 minutes | Whether the stories in the CV survive two follow up questions | Switches from I to we as soon as detail is requested |
| Total | 90 minutes | Five of the twelve skills above, verified live | Any stage where the panel disagrees on what happened |
Three exercises, and what each one reveals
1. The broken trigger, 25 minutes
Hand over a working Apex trigger that is wrong in three specific ways: it queries inside a for loop, it assumes a single record, and it ships with no bulk test. Ask for a review out loud, as if it were a colleague's pull request. You learn three things at once: whether they see the governor limit problem, whether they know a trigger can receive up to 200 records in one call, and how they talk to a peer about bad code. The last one is what a take home can never show you, and it decides whether your senior developers will want to work with them.
2. The data model and SOQL question, 20 minutes
Give a business requirement of four sentences at most, and ask for the objects, the relationships and the queries. SOQL appears in 42.86 percent of Apex specs and carries a median of 80,000 pounds, about 94,700 euros, as a named skill, so employers now treat query design as a seniority marker rather than syntax. The signal is in the follow up: ask what happens to that query when the object holds 5 million rows. Someone who has shipped at scale reaches for selective filters and indexed fields. Someone who has not says it will be fine.
3. The small component build, 30 minutes
One Lightning Web Component: a list with a filter and a single server call. Lightning Web Components are cited in 84.13 percent of Apex specs, the highest co-occurrence on the board, and that skill median rose 25 percent in a year to 75,000 pounds, about 88,800 euros. Thirty minutes is deliberately not enough to finish. You are watching how they scope, what they stub, what they say when they are stuck, and whether they write a single assertion unprompted.
The scoring grid, and where to set the bar
Score each dimension from 1 to 4 immediately after the session, independently, before the panel talks. A score agreed in conversation is a negotiation. A grid filled in separately is a decision.
| Dimension | Weight | A score of 1 looks like | A score of 4 looks like |
|---|---|---|---|
| Data model judgement | 30 percent | Lists fields, no relationships, no reason | Proposes two models and explains the trade off |
| Apex and SOQL correctness | 25 percent | Queries in loops, no bulk handling | Bulk safe by reflex, names the limit and the fix |
| Code review and communication | 20 percent | Says it is fine, or attacks the author | Finds all three defects, ranks them, stays kind |
| Build under time pressure | 15 percent | Freezes, no narration, nothing runnable | Scopes down, stubs, ships something testable |
| Ownership of past work | 10 percent | Detail dissolves on the second question | Names the constraint, the call made, the result |
Weight the grid toward what onboarding cannot fix. In our interview log the dimensions that predicted a good first 90 days were data model judgement and the quality of the code review conversation, not build speed. Hire above 2.8 out of 4 weighted, publish that bar before you see the first candidate, and reject any candidate scoring 1 on judgement or communication whatever the total says. Panels that set the bar afterwards move it, and they move it toward the person they already liked.
What a weak assessment costs you
We log the cost of a Salesforce hire that misses the first 90 days at 63,000 to 295,000 euros, once the second search, the salary already paid, the lost team output and the rework of shipped code are counted. SHRM puts the general range at 50 to 200 percent of annual salary, and the United States Department of Labor uses 30 percent of first year earnings as a floor. Against a median advertised salary of 65,000 pounds, a well designed 90 minute session is the cheapest insurance in the process. The signals that surface inside it are catalogued in our red flags when hiring Salesforce developers checklist.
Frequently asked questions
How long should a Salesforce developer technical assessment be?
Ninety minutes in one live session, after a 30 minute recruiter screen. That covers the data model, SOQL, one component build and a code review, and stays short enough that candidates in a market carrying 201 permanent and 197 contract adverts in six months do not withdraw. Past two hours of unpaid work you lose more good candidates than you filter out weak ones.
Should we still send a take home assignment?
Only if you pay for it and only if you review it live. With 50.6 percent of professional developers using AI tools daily, an unsupervised artefact tells you what a model can produce, not what the candidate knows. If you keep one, cap it at 90 minutes, pay the market rate for that time, and open the interview by asking them to change it in front of you.
Can candidates use AI during the assessment?
Yes, and say so in the invitation. AI is already cited in 21.43 percent of Apex job specs, so the tools are part of the job and banning them tests compliance rather than skill. Score the prompt, the review of the output, and the decision to reject a bad suggestion. A candidate who accepts generated Apex that queries inside a loop has failed the exercise, model or no model.
Do Salesforce certifications replace a technical assessment?
No. A certification proves exam knowledge on a fixed syllabus, and in our interview log it correlates weakly with delivery quality above the junior band. Treat it as a screening input, never as the decision. If you are still choosing which role to test for, our comparison of Salesforce developer versus admin sets out which responsibilities belong in which assessment.
Apiire runs a European Salesforce desk and sits in on client assessments, so we can tell you how a given shortlist usually performs against this grid before you book the sessions. If you are hiring, our companies page explains how we work.
