For US, UK, EU and Australian employers hiring engineers in Nigeria, Kenya and South Africa

The code test is the half you already run.

Verifying a remote developer in Africa means proving the things a coding test does not: that the person who wrote the code sat the session, that they can read a spec or an API changelog without inventing the missing half, that they can trace logic they did not write, and that they raise a problem before a deadline slips. Measured under proctoring, in one sitting, paid per candidate in US dollars.

Verify a shortlist free +91 99000 11601

At a glance

Verifying Remote Developers Hired in Africa: the facts, with their units

What AssessAll does NOT do — read this first	It does not execute your candidate's code. The platform's own validation layer refuses to store a coding question at all: "Coding questions are not currently supported — the grading pipeline does not execute code." There is no IDE, no test-case runner in the candidate player, and no language-specific problem library. If what you need is code executed against hidden test cases, buy a code-test platform. This page is about the half that platform does not measure.
What it does measure, for a technical hire	Seven live catalogue modules, all language-agnostic: Pseudo-code & Programming Logic (20 questions, 25 min), Technical Reading Comprehension (16 questions, 24 min), Flowcharts & Process Logic (18 questions, 22 min), Deductive Logic & Syllogisms (22 questions, 20 min), Spreadsheet & Data Literacy (20 questions, 22 min), Computer & Digital Fundamentals (25 questions, 18 min) and Workplace Judgment SJT (16 scenarios, 25 min). Plus written and AI-scored spoken English from a 48-product English catalogue.
What a verification costs, publicly	1 credit = US$0.50. A three-module technical verification runs roughly 20–40 credits — US$10–20 per candidate — and the candidate pays nothing and needs no card or account. New organisations get 250 free credits, which is enough to verify a first shortlist before spending anything. There is no seat licence, no platform fee and no annual minimum.
The arithmetic that can rule us out	Divide a code-test subscription by our per-candidate price. Codility publishes a Starter plan at US$1,200 a year (checked 4 September 2026); at US$15 per verified candidate that is about 80 candidates a year. If you screen engineers at a steady volume well above that and code is the only thing you want measured, a subscription is the cheaper instrument and you should buy one.
Who publishes a price in this category	Of the two code-test platforms checked on 4 September 2026, Codility publishes plan prices on its own pricing page — Starter US$1,200/year, Scale US$6,000/year, Custom on request. HackerRank's pricing page routes a buyer to a demo request; the figures circulating for it come from third-party procurement aggregators, not from HackerRank. Ask every vendor for a per-candidate number in writing at the volume you actually expect.
The integrity band, with its arithmetic shown	Every attempt is scored, not judged. Points per occurrence: window blur 1, ambient speech 1, gaze off screen 1.5, tab switch 2, fullscreen exit 2, paste blocked 2, no face 2, secondary-camera disconnect 2, print-screen blocked 3, screen-share stopped 3, additional face in frame 4, suspiciously fast completion 4, critical flag 5, AI-styled answer 3 or 8. Under 5 points is Low, 5–14 Medium, 15 or more High. The platform's own code calls these "investigative weights, not verdicts" — nothing auto-rejects anyone.
For a developer screen, the weight that matters is paste	Paste-blocked and print-screen-blocked are the two counters that separate "wrote this" from "moved this here", and they are weighted 2 and 3. None of it proves an AI assistant was used. It tells you which attempts to look at, and the evidence is attached to the attempt so you can decide for yourself before an offer.
What AssessAll has NOT published	No validity study, no local norm group and no benchmark for any of these instruments. The platform has not passed its own internal evidence gate on any assessment, so there is no AssessAll norm to quote and this page quotes none. If a vendor offers you an African developer benchmark, ask how many sittings, from how many organisations, and what share came from the largest one.

A code test tells you the code ran. It does not tell you who wrote it

Every unproctored technical exercise now measures the candidate plus whatever assistant they had open, and the artefact looks the same either way. That is not an argument for abandoning code tests — it is an argument for knowing which parts of your evidence were produced under observation and which were not. The parts that were not are the parts you should stop weighting.

Cross-border engineering hires rarely fail on syntax

They fail on the ticket that was read too fast, the ambiguity nobody flagged, the changelog whose breaking-change note was skipped, and the deadline that slipped for three days before anyone said so. None of those appear in a passing test suite, and all of them are measurable — reading under time pressure, tracing logic you did not write, and judgement about when to escalate.

You cannot audit a vetted-marketplace bar

Buying pre-vetted talent moves the assessment behind someone else's wall. The bar is real and often good; it is also unpublished, unversioned, and not comparable between two providers or between this year and last. If you never see the instrument or the score, you cannot re-cut your own bar when a cohort underperforms — you can only change supplier.

Candidate-paid screening does not work in this corridor

Card penetration and currency friction make "pay to take our test" a dead end in exactly the markets with the deepest junior and mid-level engineering supply. Here the employer pays per candidate in US dollars; the candidate opens a link on a phone, needs no account and no card, and the player resumes where it stopped when a connection drops.

What you get

Built for verifying remote developers hired in africa

Program logic, without a language

Pseudo-code & Programming Logic — 20 questions in 25 minutes on reading and tracing program logic: loops, conditions, accumulators, nested iteration, integer division, string handling, value-versus-reference on a simple copy, function composition, a bug-spotting trace, and concept checks on FIFO and linear time. Language-agnostic — no syntax knowledge required, only the ability to execute logic in your head. Every traced answer in the bank was verified by running equivalent code.

Reading the documents engineers actually misread

Technical Reading Comprehension — 16 questions in 24 minutes on a product manual, an API changelog and an SOP, probed for conditions, exceptions, deadlines and precedence, plus compliance checks against specs, config rules, release notes and warranty clauses. Every answer is decidable strictly from the text, which is the point: the failure mode being measured is the assumption, not the vocabulary.

Executing a process exactly as specified

Flowcharts & Process Logic — 18 questions in 22 minutes tracing two realistic multi-branch flows under different inputs, plus loops with exit conditions, retry counting, decision tables, parallel-path durations, dependency ordering and escalation timing. This is the competency behind on-call runbooks, deployment checklists and anything automated.

Deduction, including the refusal to conclude

Deductive Logic & Syllogisms — 22 questions in 20 minutes: categorical syllogisms, modus ponens and modus tollens with the invalid forms most people fall for, disjunctive elimination, chained implications, contrapositive versus converse versus inverse. Several items are keyed to "may or may not", so the discipline of refusing an unwarranted conclusion is itself scored — which is exactly the habit you want in someone reading an incident timeline.

Data and digital fundamentals

Spreadsheet & Data Literacy (20 questions, 22 minutes — absolute references, IF/COUNTIF/VLOOKUP semantics, format-versus-value, the single-column sort disaster, median-versus-mean under outliers, truncated axes, CSV and mixed units) and Computer & Digital Fundamentals (25 questions, 18 minutes — files and formats, cloud concepts, networks and IP, plus a security block on passwords, phishing, 2FA, HTTPS and ransomware backups).

Judgement, escalation and ownership

Workplace Judgment — 16 realistic scenarios in 25 minutes: errors already sent to a client, conflicting senior instructions, confidential data exposed, impossible deadlines, scope creep, Friday-evening discoveries. Scored with graded partial credit, so a plausible-but-suboptimal response earns partial marks and the test separates sound judgement from merely avoiding the obviously wrong answer.

Working English, written and spoken

Written accuracy, register, concision and comprehension, plus AI-scored speaking and listening from the candidate's own recorded voice — a task no chatbot can sit for them. For a distributed engineering team English is not a nice-to-have; it is the medium the entire job happens in, and "fluent" on a profile becomes a score you can set a bar against.

Proctoring whose evidence you can open

Identity-baseline photo at the start, face-presence and additional-face monitoring, tab-switch, paste and screen-capture counters, and a High/Medium/Low integrity band computed from the published weights above — with the underlying evidence attached to the attempt rather than summarised away. A band you cannot inspect is just another claim.

One link, any phone, no accounts

Delivery is a share link or a QR code. Candidates create no account and need no card, the player is mobile-ready and resumable when a connection drops, and results rank the moment each person submits. That is what makes screening practical across Lagos, Nairobi and Cape Town without scheduling a single call.

How to build the screen

Seven steps, in order, that you could start on Monday

1
Decide what only code can answer — and buy that separately
Write the shortest honest list of things you cannot infer: does this person write idiomatic Go, can they debug a race condition, is their SQL sane under load. Those need code executed. Nothing on this page replaces a code-test platform or a live pairing session, and any vendor telling you otherwise is selling you a proxy as a substitute.
2
Write down the failure you are actually trying to avoid
Not "a bad developer" — the specific bad outcome you have already lived through. Went dark for four days. Shipped against a spec they had misread. Could not be understood on a call with the client. Each of those points at a different module, and a screen assembled without this sentence measures whatever the vendor happened to bundle.
3
Verify the reading layer first — it is the cheapest and it moves the most
For a hire you will manage asynchronously across five time zones, the ability to read a spec, a changelog and a ticket precisely is doing more work than any other single measure. It is a 24-minute module, it is language-agnostic, and it is the one most likely to reorder your shortlist on the first run.
4
Trace logic without picking a language
A pseudo-code trace asks the candidate to execute logic in their head with no syntax to hide behind and no library to reach for. It is not a substitute for a code test; it is the thing you can run on 300 people in a week when a code test on 300 people is neither affordable nor reviewable.
5
Put judgement in the same sitting, not in a later interview
Escalation, ownership and how someone handles a conflicting instruction are the behaviours remote hires fail on, and they are the behaviours an interview is worst at reading, because an interview rewards the candidate who performs well in an interview. Sixteen scenarios with graded partial credit, taken by everyone, is a fairer and more comparable reading than four people's impressions.
6
Proctor — then read the evidence, not the band
The band is a weighted arithmetic summary, and the platform's own code calls its weights investigative rather than conclusive. Open a Medium before you act on it. A flatmate crossing the frame and a second person answering the questions produce different evidence and the same colour. The rule that keeps this honest: no attempt is auto-rejected, by us or by you, on a band alone.
7
Re-cut your bar against who was still there at 90 days
The step no vendor enjoys writing. Six months in, line up every score you collected against who is still delivering, and move the bar where the evidence says — including down, if you have been screening out people who would have been fine. A screen you never audit is a superstition with a dashboard.

Regulation & compliance

South Africa: the one legal fact that changes your configuration

This is not a full compliance guide, and the corridor's other markets are not covered here. It is the single instrument that changes how a South African screen may be built, plus the sentence most pages in this market get wrong. Statuses were checked at the sources named below on the date shown.

Statuses verified at primary sources on 4 September 2026.

SiddharthanFounder, AssessAll — Bodhih Training Solutions

Founder of AssessAll and of Bodhih Training Solutions, a corporate training company in Bangalore. Works on assessment design, scoring and reporting across hiring, L&D and certification programmes.

Last reviewed 4 September 2026

Competitor pricing facts were read at each provider's own published pricing page on 4 September 2026 and are re-verified monthly. The South African statutory position was checked at the sources named in that section on the same date. Nothing here is legal advice, an endorsement, or a ranking.

South Africa

Section 8 of the Employment Equity Act is the reason South Africa is not simply "another market" for an assessment. It bans a category of testing outright unless three conditions are met — and the fourth condition everyone still cites was struck down in 2017 and deleted from the Act on 1 January 2025.

Employment Equity Act 55 of 1998, section 8

In force · amended by the Employment Equity Amendment Act 4 of 2022, commenced 1 January 2025

Section 8 prohibits psychological testing and other similar assessments of an employee unless the test (a) has been scientifically shown to be valid and reliable, (b) can be applied fairly to all employees, and (c) is not biased against any employee or group. A fourth condition inserted in 2014 — certification by the Health Professions Council of South Africa — was declared null and void and of no force or effect by the court on 2 May 2017 in the Association of Test Publishers of South Africa matter, and has now been removed from the Act itself by the 2022 Amendment Act, which commenced on 1 January 2025.

What it means for a hiring assessment

The sentence to stop repeating is "the test must be HPCSA-certified". It must not, and it cannot be — the requirement has been legally dead since 2017 and is no longer in the statute at all. What survives is harder and it does not sit with your vendor: conditions (a) to (c) are conditions on your use of the instrument for that role. Two things follow in practice. First, prefer instruments whose content is visibly job-related — a technical-reading or pseudo-code module read alongside an engineering job description is far easier to defend than a general personality inventory used as a filter. Second, take advice on whether what you are running is "psychological testing or a similar assessment" at all before you run it; practitioners draw a line between a proficiency or job-knowledge test and an instrument measuring psychological constructs, and the answer changes what you must be able to show.

Source: ENS — commencement of the Employment Equity Act amendments (1 January 2025)

HPCSA test classification (not certification)

Administrative · classification only, the HPCSA's stated position since June 2019

After the certification requirement fell away, the HPCSA moved to classifying tests — determining whether an instrument is a psychological measure — rather than certifying them against the Employment Equity Act. Assessment Standards South Africa states the position plainly: registration of a test with the HPCSA is not a certification that the test complies with the Employment Equity Act.

What it means for a hiring assessment

So a vendor telling you an instrument is "HPCSA registered" has told you it is treated as a psychological measure, which if anything places it more firmly inside section 8 rather than outside it. It is not a compliance badge and should not be read as one. AssessAll holds no HPCSA classification or registration for any instrument, does not present any of them as an approved psychological measure in South Africa, and publishes no South African validity study or norm group. If that matters for your use case, it should rule us out for that use case — and any vendor whose answer to the same question is vaguer than this one is worth a second question.

Source: Assessments Standards South Africa — SA legislation

AssessAll holds no South African accreditation, certification or regulatory approval of any kind, and no AssessAll instrument is classified or registered with the HPCSA. Nothing on this page is legal advice and none of it is a claim that AssessAll is compliant with, or certified against, any instrument named here. Statuses were checked at the sources above on 4 September 2026 and are re-verified quarterly. Take South African employment-law advice before assessing employees or candidates in South Africa.

Compare the approaches

Five ways to get evidence about a remote developer, and when each is right

Approach-level rows, not a vendor ranking. Prices are stated only where the provider publishes them on its own pricing page, and the "when to choose this instead" line is the point of every row — including the rows that are not ours.

External facts verified at source on 4 September 2026.

Approach	What it measures	What it costs	When to choose it instead
Code-test platform (e.g. Codility, HackerRank)	Code executed against visible and hidden test cases, in a named language, usually with plagiarism and similarity signals.	Codility publishes Starter US$1,200/year and Scale US$6,000/year, with a Custom tier on request. HackerRank's pricing page routes to a demo request and publishes no plan price (both checked 4 September 2026).	Choose this — in addition to, not instead of, anything here — whenever the role is production coding in a named language and you need to see the code itself. Nothing on this page executes code, and a proxy for coding ability is not coding ability.
Take-home project	An artefact of unverified authorship, produced over unbounded time with unbounded assistance.	No licence cost; a reviewer's hours per candidate, which is usually the largest real cost in this table.	Choose it when you can afford to review it live with the candidate, line by line, and treat the conversation as the measurement. As an unattended filter it now mostly measures who had the better assistant and the most free time — which quietly penalises the candidates with day jobs.
Live technical interview or pairing session	Reasoning under observation, collaboration, how someone recovers when stuck. The highest-signal option in this table.	Senior engineer hours per candidate, plus scheduling across time zones.	Choose it for your last three candidates. It is the wrong instrument for your first three hundred, and using it there is how a pipeline stalls and a hiring manager starts rubber-stamping.
Vetted-talent marketplace or outsourcing partner	Someone else's bar, applied before you see the candidate. Often genuinely rigorous; rarely published, versioned or comparable between providers.	A markup per hour or a placement fee, rather than a per-candidate assessment cost.	Choose it when you want sourcing solved and are content not to hold the evidence yourself. The trade is real: if a cohort underperforms you can change supplier, but you cannot re-cut a bar you have never seen.
Proctored non-code verification (AssessAll)	Technical reading, program-logic tracing, process and deductive logic, data and digital fundamentals, workplace judgement, written and spoken English — under proctoring, with an integrity band whose weights are published and evidence attached.	US$10–20 per candidate at 1 credit = US$0.50, employer-paid, no seat licence or annual minimum; 250 free credits to start.	It is the wrong choice when code execution is the requirement — this platform does not run code, and that is a stated limitation rather than a roadmap item. It is also wrong when you need published local validity evidence to defend a specific selection decision, because AssessAll publishes none; when you screen engineers at steady high volume and syntax is genuinely all you need measured, where a code-test subscription is cheaper above roughly 80 candidates a year; and when the bottleneck is sourcing rather than selection, because measuring a pool you do not have changes nothing.

This table is not a ranking and none of these options is wrong. Most teams hiring engineers across a border end up running two of them: something that executes code for the finalists, and something that measures reading, logic, judgement and English across everyone. Where they are mutually exclusive is budget, so the arithmetic in the key facts above is the one to run before shortlisting a vendor.

Sources: Codility pricing page · HackerRank pricing page

The corridor, market by market

What actually differs between Lagos, Nairobi and Cape Town

The instrument does not change across these markets. Delivery, payment and — in one case — the legal frame do.

Nigeria

Assume a phone, not a laptop, and assume the connection will drop at least once. Delivery is a share link or QR code with no candidate account and no card; the player is mobile-ready and resumes where it stopped rather than voiding the attempt. Schedule nothing — a window that stays open and ranks on submission fits a distributed pool better than a fixed sitting.

Kenya

The same delivery constraints, with an added practical note for high-volume drives: because scoring and ranking happen on submission, a Nairobi-hours cohort and a European-hours reviewer never have to be awake at the same time. The evidence report is the handover, so a hiring manager reads a ranked list rather than attending a sitting.

South Africa

The one market in this corridor with a statute that speaks directly to testing. Read the section above before you configure anything: section 8 of the Employment Equity Act puts validity, fairness and bias conditions on your use of a psychological or similar assessment, the HPCSA certification requirement everyone still cites has been legally dead since 2017 and out of the Act since 1 January 2025, and AssessAll publishes no South African validity study or norm group.

The employer's side of the corridor

Billing is in US dollars at US$0.50 a credit wherever the paying organisation sits outside India, so the per-candidate cost is a number you can put in a hiring budget rather than a quote you have to chase. The candidate never pays and never needs a card, which is the design decision that makes screening in these markets work at all.

Common use cases

US, UK, EU and Australian employers hiring engineers directly in Nigeria, Kenya or South Africa
Screening a large inbound applicant pool before any engineer's time is spent
Adding the non-code layers to an existing code-test process
Agencies and staffing firms evidencing an offshore engineering shortlist to a client
Talent marketplaces verifying their own supply side under their own branding
Re-verifying an existing contractor before a renewal or a rate increase

Pricing for remote developers (africa)

Employer-paid, per candidate, in US dollars: 1 credit = US$0.50, and a technical verification set runs 20–40 credits (US$10–20) per candidate depending on how many modules you include. Volume packs of 250 / 1,000 / 5,000 credits cost US$125 / 465 / 2,175 — up to 13% off. No seat licences, no platform fee, no annual minimum, and 250 free credits on signup, which covers a first shortlist before you spend anything.

See full pricing solutions@bodhih.com

Frequently asked questions

Does AssessAll run coding tests?+

No. AssessAll does not execute candidate code — the platform's validation layer refuses to store a coding question at all, and returns the message "Coding questions are not currently supported — the grading pipeline does not execute code." There is no IDE and no test-case runner in the candidate player. What it measures for a technical hire is the language-agnostic layer: tracing pseudo-code, reading specs and API changelogs, process and deductive logic, data and digital fundamentals, workplace judgement, and written and spoken English — all under proctoring. Most teams hiring engineers across a border run both: a code-test platform for finalists, and this across everyone.

How do you verify a remote developer you will never meet?+

By measuring them under the same monitored conditions as everyone else, rather than trusting a CV, a portfolio or a reference you cannot check across a border. Every candidate takes the same modules through one link, with an identity-baseline photo captured at the start and face-presence, additional-face, tab-switch, paste and screen-capture counters running throughout. You receive scores per module, an integrity band computed from published weights, the proctoring evidence itself, and a ranked position against everyone else in the drive.

Can a candidate just use an AI assistant to pass?+

That is exactly why an unproctored take-home has stopped being useful — the artefact looks identical whether the candidate wrote it or an assistant did. Here the sitting is proctored and timed, spoken English tasks are scored from the candidate's own recorded voice, and paste attempts, print-screen attempts, tab switches, additional faces and suspiciously fast completion are all counted and weighted into the integrity band. To be precise about what that gives you: it is evidence, not proof. The platform's own code calls these "investigative weights, not verdicts", nothing auto-rejects anyone, and the right use of a Medium band is to open the attempt and look.

How is the integrity band calculated?+

Arithmetically, from weighted counters, and the weights are published rather than described. Points per occurrence: window blur 1, ambient speech 1, gaze off screen 1.5, tab switch 2, fullscreen exit 2, paste blocked 2, no face 2, secondary-camera disconnect 2, print-screen blocked 3, screen-share stopped 3, additional face in frame 4, suspiciously fast completion 4, critical flag 5, and an AI-styled answer review 3 or 8. Below 5 points is Low, 5 to 14 is Medium, 15 or more is High. An AI review that concludes the person on camera is not the same person, or that other people were present, forces High regardless of the score.

Is this legal to use in South Africa?+

That depends on what you run and for what role, and it is a question for a South African employment-law adviser rather than for a vendor. The frame is section 8 of the Employment Equity Act: psychological testing and similar assessments of an employee are prohibited unless the test has been scientifically shown to be valid and reliable, can be applied fairly to all employees, and is not biased against any employee or group. The HPCSA certification condition that most pages still cite was declared null and void on 2 May 2017 and removed from the Act by the Employment Equity Amendment Act 4 of 2022, which commenced on 1 January 2025. AssessAll holds no HPCSA classification and publishes no South African validity study — so if your use case requires published local validity evidence, it should rule us out for that use case.

What does it cost, and who pays?+

The employer or the platform pays, per candidate, in US dollars: a technical verification set is 20–40 credits, which at US$0.50 a credit is US$10–20 per candidate, with volume packs up to 13% off. The candidate pays nothing and needs no card — which is the point, because candidate-paid screening does not work in these corridors. New organisations get 250 free credits, enough to verify a first shortlist before spending anything.

When should I buy a code-test platform instead?+

When code is the thing you need to see, and when your volume is steady enough to make a subscription cheaper than paying per candidate. The arithmetic is simple and it is on this page deliberately: Codility publishes a Starter plan at US$1,200 a year, which at US$15 per verified candidate is about 80 candidates a year. Above that volume, if execution against test cases is your only requirement, a subscription is the better instrument. Below it — or where reading, logic, judgement and English matter as much as syntax — paying per candidate holds up better.

Is this a background check?+

No, and the two should not be confused. AssessAll verifies demonstrated capability under proctored conditions and the integrity of the session itself. It does not verify legal identity documents, employment history, education records or criminal history. Use a background-screening provider for those; use this for whether the person can actually do the work, and whether they did it themselves.