Sample shortlist · candidates written by us, real screener output
WordPress Developer
This is what the screening delivers: every candidate scored against the brief, with screening notes, their answers and a graded work sample. These sample candidates were written by us and run through the real screener — the scores and notes are its unedited output.
At a glance
Highest score first. Scores within 2 points are level: when we had the screener read the same two applications again, the gap between them moved by up to 2. A must-have in the brief (below the list) costs 5 points when the application does not show it and 2 when it only partly does.
- Sipho D.Goes on to the interviewLevel with Miguel S. (89 vs 88 is too close to call): stronger on the work sample (91 vs 88), Miguel S. on the scenario answer (91 vs 88).89/100
- Miguel S.Goes on to the interviewAhead of Paula E. mainly on the work sample (88 vs 62).88/100
- Paula E.Stops here55/100
2 of 3 clear 70, the score an applicant needs to go on to the written interview. A real shortlist only holds people who clear it and then the interview; the one below it is shown so you can see what the screener turns down.
The brief they were screened against
Role: WordPress Developer · Band: $1,000–1,700/mo
Must-haves: maintains and extends existing WordPress sites, reads the code before changing it, child themes and staging-first updates, PHP and plugin-conflict debugging
Nice to have: WooCommerce, performance work (caching, Core Web Vitals), custom Gutenberg blocks
Scored on 2026-09-23 by the AI screener that grades real applicants: the wordpress-developer rubric, run on claude-sonnet-4-6. Nothing below was edited afterwards — including where it went against a candidate. The rubric and weights are published.
Sipho D. · 1 of 3
SCREEN 89/100How the 89 adds up · Top ~10%
A bar is as long as its part's weight; the filled share is the score out of 100.
- Graded work sample 35%91
Diagnosed a real block-invalidation root cause, named 3,412 affected posts, fixed with a deprecation entry, restored 9 posts from revisions, and added a Playwright CI test — complete, client-ready.
- Scenario judgment 25%88
Reversible-readable-reproducible framework, BackstopJS per-commit comparisons, fork-vs-child-theme tradeoff written to client — specific, ordered, and shows genuine judgment.
- Experience specificity 20%87
Five years, one employer, 30,000 posts, 9 plugins, 23 blocks, LCP 3.8s→1.7s, Git/GitHub Actions/PHPUnit/BackstopJS/New Relic — numbers and named tools throughout.
- Written & spoken English 15%84
Precise register throughout; 'I part ways with the usual advice' shows confident, idiomatic control appropriate for a US client relationship.
- Salary-band fit 5%100
Asks $1,700 a month, inside the $1,000–1,700 band.
Must-haves in the brief: 4 of 4 shown
- maintains and extends existing WordPress sites — shown
the in-house WordPress developer for a US nonprofit news publisher
- reads the code before changing it — shown
I put wp-content into a Git repository exactly as it is today. That baseline is how I find out what people changed by hand
- child themes and staging-first updates — shown
a staging copy on the same PHP version
- PHP and plugin-conflict debugging — shown
I bisect on staging with Health Check's troubleshooting mode, which disables plugins for my session only, and read the stack trace in Query Monitor
The parts, weighted as above, come to 89.
Screening notes
Strongest: The work sample is exceptional — root cause, scope (3,412 posts), fix, recovery of 9 posts from revisions, and a CI test added, all in one answer.
Concern: WooCommerce is self-described as "not my strong side" and one store "some years ago" — thin for clients who need it.
Worth asking: What does the one-month notice period mean in practice — is an earlier start negotiable?
Verdict: A highly experienced, evidenced WordPress developer whose tooling, staging discipline, and block expertise are well above typical for this role and band.
Experience
8 years as a PHP developer, the last 5 as the in-house WordPress developer for a US nonprofit news publisher — about 30,000 posts, 14 editors and a membership paywall — working remotely from Johannesburg. I own the custom theme, nine custom plugins and 23 Gutenberg blocks, all in Git with automated deploys to staging and live. I moved the archive from the classic editor to blocks without breaking old posts, and cut mobile LCP on article pages from 3.8s to 1.7s. Before WordPress I built Laravel apps for a Cape Town agency.
Tools
PHP 8, WP-CLI, Git with GitHub Actions for deploys, Composer, PHPUnit, PHP_CodeSniffer (WordPress Coding Standards, PHPCompatibilityWP), @wordpress/scripts and create-block, Query Monitor, New Relic, Redis object cache, BackstopJS for visual regression, Playwright, LocalWP, Cloudflare. WooCommerce: one store, some years ago — not my strong side.
Scenario answer
Asked: You inherit a WordPress site that hasn't been updated in 18 months. Core and six plugins are significantly out of date. The client says 'just update everything, it'll be fine.' Walk through exactly what you do before touching the Update button — and how you'd handle the update itself if your first check uncovered a potential conflict.
Before anything is updated, I want the site reversible, readable and reproducible, in that order. Reversible: a full backup, database and wp-content, restored somewhere else and seen working, plus the host's snapshot as a second copy. Readable: I put wp-content into a Git repository exactly as it is today. That baseline is how I find out what people changed by hand, and on an 18-month-old site I check three places: core files (wp core verify-checksums), plugin files (wp plugin verify-checksums --all), and the parent theme, diffed against a clean copy of the same version. Then I read functions.php and any custom code for every call into the six plugins. A plugin update rarely breaks the plugin; it breaks the code that depends on it. Reproducible: a staging copy on the same PHP version, and before changing anything, screenshots of the 20 most-visited URLs. After each update I compare them with BackstopJS, so a broken layout on a page nobody thought to check still shows up. Then the changelogs for every version in the gap: minimum PHP and WordPress versions, database migrations, removed features. If the first check uncovers a conflict, it depends what kind: - A plugin's new version needs newer PHP: I hold that plugin back, update everything else, and treat the PHP upgrade as its own piece of work with its own test pass. - Two plugins that clash once updated: I bisect on staging with Health Check's troubleshooting mode, which disables plugins for my session only, and read the stack trace in Query Monitor. - The parent theme was edited directly. Here I part ways with the usual advice. A few template tweaks go into a child theme. But if someone rewrote a third of the parent, I stop treating it as a vendor theme: I fork it into the repository and rename it so WordPress never offers the vendor's update over the top. From then on we own it, including its security fixes, and I put that cost to the client in writing. The update itself happens on staging, one plugin per commit, with the visual comparison after each. Live gets the same commits through the deploy, then wp core update-db and any plugin database upgrades. Nobody clicks Update on production. The client gets a short note: what I'm updating, what I found, and the one thing I need from them — on sites this old, usually a lapsed premium licence.
Work sample
Asked: Describe a real WordPress problem you've diagnosed and fixed: what was the symptom, what did you actually find when you investigated, what did you change, and how did you confirm it was fixed?
SYMPTOM On a Monday morning the publisher's editors started posting screenshots in Slack: older articles with our 'Key facts' box showed 'Block contains unexpected or invalid content.' in the editor. The live site looked fine. Two editors had clicked 'Attempt recovery' on some posts, and on a few of those the box lost its heading. WHAT I FOUND The cause was my own change. On the Friday I had shipped a heading-level option for the block, and in the same change the wrapper went from a div to an aside. A static block stores its HTML in the post. When the editor opens a post, it runs the block's current save function on the stored attributes and compares the result with the stored HTML; if they differ, the block is marked invalid. The front end was untouched because it serves the stored HTML and never runs save. I had tested the change on new posts only. A search of post_content for '<!-- wp:newsroom/key-facts' found 3,412 posts using the block. WHAT I CHANGED 1. Added a deprecated entry to the block's registration with the previous attributes and the previous save function. The editor now recognises the old markup as an earlier version and upgrades it when a post is opened. Nothing in the database had to be rewritten. 2. Restored the 9 posts where 'Attempt recovery' had dropped the heading, from revisions, one at a time, each checked by an editor. 3. Added a rule to the repo, enforced in code review: any change to a save function ships with its deprecation in the same pull request. HOW I CONFIRMED - On staging, a Playwright test opens 60 random posts from the 3,412 in the editor and fails if the invalid-content warning appears. It passed, and it now runs in CI on every block change. - Front end: rendered HTML of 20 posts before and after, compared with diff. Identical. - The editors confirmed the same afternoon, and there have been no reports since. WHAT I'D DO DIFFERENTLY Test a block change against existing content first. New posts are the one case where a changed save function always looks fine.
Availability
Full-time, 1pm–10pm SAST — 7am–4pm US Eastern in summer, 6am–3pm in winter, so at least six hours of overlap. One month's notice.
Miguel S. · 2 of 3
SCREEN 88/100How the 88 adds up · Top ~10%
A bar is as long as its part's weight; the filled share is the score out of 100.
- Graded work sample 35%88
Real symptom, specific root cause (null freight-class field, PHP 8 TypeError), exact fix in child theme, confirmed on staging with cent-accurate shipping totals.
- Scenario judgment 25%91
Ordered, specific protocol: verified backup restore, wp verify-checksums for silent edits, changelog gaps, one-at-a-time staging updates, PHP split-job reasoning.
- Experience specificity 20%85
40 US client sites, 4-year agency tenure, emergency tickets 9→1, LCP 4.3s→2.0s, named tools throughout — specific and numbered throughout.
- Written & spoken English 15%84
Consistently clear, correct register; 'A backup nobody has restored is a hope' shows confident, professional prose suited to US client communication.
- Salary-band fit 5%100
Asks $1,400 a month, inside the $1,000–1,700 band.
Must-haves in the brief: 4 of 4 shown
- maintains and extends existing WordPress sites — shown
6 years maintaining WordPress sites
- reads the code before changing it — shown
Read the code the updates will touch
- child themes and staging-first updates — shown
staging first, and wrote the agency's update checklist
- PHP and plugin-conflict debugging — shown
PHP Fatal error: Uncaught TypeError: count(): Argument #1 ($value) must be of type Countable|array, null given
The parts, weighted as above, come to 88.
Screening notes
Strongest: Scenario answer shows genuine update-discipline thinking — verified restores, checksums for silent edits, split PHP/plugin jobs — backed by a work sample with exact error text and cent-accurate staging confirmation.
Concern: Custom Gutenberg block work is self-described as "a handful, not my main work"; depth there is limited.
Worth asking: How does the candidate handle premium plugin conflicts where the vendor won't backport a fix and the client won't upgrade PHP?
Verdict: A well-evidenced, methodical WordPress maintainer whose ask sits comfortably inside the band and whose answers consistently show the discipline the role requires.
Experience
6 years maintaining WordPress sites, the last 4 at a Manila web agency that runs monthly care plans for about 40 US small-business sites — law firms, dental clinics, a home-services company and three WooCommerce stores. I ran the update cycle for all of them, staging first, and wrote the agency's update checklist; emergency tickets caused by updates went from 9 in the quarter before it to 1 in the quarter after. I also fix slow sites: took one WooCommerce store's mobile LCP from 4.3s to 2.0s.
Tools
WP-CLI, Query Monitor, WP_DEBUG_LOG, LocalWP, WP Engine and Kinsta staging, UpdraftPlus, Health Check & Troubleshooting, WP Crontrol, PHP_CodeSniffer with PHPCompatibilityWP, Git (Bitbucket), WooCommerce, ACF Pro, PageSpeed Insights, custom Gutenberg blocks with @wordpress/create-block (a handful, not my main work)
Scenario answer
Asked: You inherit a WordPress site that hasn't been updated in 18 months. Core and six plugins are significantly out of date. The client says 'just update everything, it'll be fine.' Walk through exactly what you do before touching the Update button — and how you'd handle the update itself if your first check uncovered a potential conflict.
Before the Update button, in this order. 1. A backup I have actually restored. The host's snapshot, plus an off-site copy with UpdraftPlus and a wp db export. Then I restore it into LocalWP. A backup nobody has restored is a hope, and old sites are where you find out. 2. An inventory, written down. wp plugin list --update=available gives me the installed and new version of each plugin, Site Health gives the PHP version, and I run wp core verify-checksums and wp plugin verify-checksums --all. On an abandoned site someone has usually 'quickly fixed' a plugin file, and the update silently deletes that fix. 3. Read the code the updates will touch. Is it a child theme, or was the parent edited directly? If edited, I diff it against a clean copy of the same version first. Then I search the theme and any custom plugin for calls into the six plugins — hooks, shortcodes, template overrides — because that is what breaks when a plugin changes. 4. Changelogs for every version in the gap, not just the latest: a new minimum PHP or WordPress version, a database migration, anything removed or renamed. 5. A staging clone on the same PHP version as live — WP Engine or Kinsta staging, or LocalWP — with WP_DEBUG_LOG on and display off. Then on staging, one at a time: plugins that still support the current core, then core, then the remaining plugins, then the theme. After each, I test the pages that make the client money — home, contact form (and that the email arrives), checkout, login — and read debug.log. Slower, but when something breaks I know which update did it. If the first check finds a conflict — say the booking plugin's new version needs PHP 8.1 and the server runs 7.4 — I split it into two jobs. Everything that works on 7.4 goes live first. Then PHP on its own: staging to 8.1, PHPCompatibilityWP over the theme and custom code, fixes in the child theme, and only then that plugin. I never change PHP and six plugins in one step; if something breaks, you can't tell which change did it. Live: a fresh backup, a quiet hour, the same order, each plugin pinned to the version I tested with wp plugin update <slug> --version=<version>. Rollback is the host snapshot, and the client knows beforehand how long it takes. To the client: 'Yes, everything gets updated — this week, not all at once, and not on the live site first.' One limit: if a premium plugin's license has lapsed, I can't update it until they renew. On sites this old, that's common.
Work sample
Asked: Describe a real WordPress problem you've diagnosed and fixed: what was the symptom, what did you actually find when you investigated, what did you change, and how did you confirm it was fixed?
SYMPTOM A WooCommerce store selling home-gym equipment, about 1,100 products. The host moved it from PHP 7.4 to 8.1 on a Tuesday. On Friday the owner forwarded a customer email: 'There has been a critical error on this website' on the cart page after entering a ZIP code. Only some carts. Small items checked out fine, which is why nobody noticed for three days. WHAT I FOUND I cloned the site to Kinsta staging on PHP 8.1, turned on WP_DEBUG_LOG and rebuilt a failing cart — a squat rack to a Colorado ZIP. debug.log had it on the first try: 'PHP Fatal error: Uncaught TypeError: count(): Argument #1 ($value) must be of type Countable|array, null given' in the child theme's inc/freight.php. A previous developer had written a freight surcharge in the child theme, hooked to woocommerce_package_rates. It read each product's freight-class field and called count() on it. 214 products had been imported without that field, so it came back null. PHP 7.4 only logged a warning, and WP_DEBUG was off, so nobody saw it for two years. PHP 8 throws a TypeError, which is fatal, so any cart holding one of those 214 products died at shipping. WHAT I CHANGED 1. In the child theme: replaced the count() with a check for a non-empty array, and made products with no freight class fall back to the oversized rate instead of silently getting no surcharge. That fallback changes what customers pay, so I asked the owner before choosing it. 2. Exported the 214 products and sent the list to the owner's team to fill in the field. 3. Ran PHPCompatibilityWP over the child theme and the two custom plugins to find the next one. It flagged four places passing null to string functions — only deprecation notices on 8.1, but I fixed them while I was there. HOW I CONFIRMED - On staging, 10 carts across several states, mixing the 214 products with normal ones: no fatals, and for products that had a freight class, shipping matched the old totals to the cent. - Deployed through Git, not the theme file editor, so the change has a history. - Left WP_DEBUG_LOG on in production, writing to a file outside the web root, and read it daily for a week. Zero fatal errors, and oversized-item orders were back the same day. WHAT I'D DO DIFFERENTLY The host had announced the PHP change two weeks earlier, to an address nobody read. Now every care-plan site's PHP end-of-life date is on my calendar, and each site is tested on the next PHP version a month before the host moves it.
Availability
Full-time, 9pm–6am PHT (9am–6pm ET). Two weeks' notice.
Paula E. · 3 of 3
SCREEN 55/100How the 55 adds up · Competent but generic
A bar is as long as its part's weight; the filled share is the score out of 100.
- Graded work sample 35%62
Clear symptom-to-fix narrative with real tools named, but diagnosis stops at hosting server was not sending emails correctly — no log checked, no bounce code cited.
- Scenario judgment 25%58
Covers backup-first and one-by-one updates, but treats staging as optional and updates WordPress core last without explaining why.
- Experience specificity 20%60
Four years, two freelance, named client types and tools, but no numbers (sites count, traffic, revenue) and no decisions beyond standard maintenance.
- Written & spoken English 15%74
Consistently clear and professional; phrasing like 'I would be careful' is slightly informal but register is appropriate for US client-facing work.
- Salary-band fit 5%100
Asks $1,150 a month, inside the $1,000–1,700 band.
Must-haves in the brief: 1 of 4 shown, 2 partly
- maintains and extends existing WordPress sites — shown
4 years building and maintaining WordPress websites for small businesses
- reads the code before changing it — not shown (−5)
- child themes and staging-first updates — partly shown (−2)
if the hosting has a staging option I would do it there first, but many small business hosting plans do not have one, so I would do the updates at night when there is less traffic
- PHP and plugin-conflict debugging — partly shown (−2)
I would deactivate the plugin that is causing the problem, restore the backup if necessary, and contact the plugin's support
The parts, weighted as above, come to 64; must-haves not fully shown cost 9, so 55.
Screening notes
Strongest: Practical, client-facing communication style and a coherent end-to-end fix story in the work sample.
Concern: Staging treated as optional; no mention of child themes anywhere; no evidence of reading existing code before making changes.
Worth asking: How do they handle a site built on a parent theme — do they use a child theme, and can they walk through a real example?
Verdict: A reliable maintenance-level WordPress operator whose application shows gaps in two of the four must-haves, placing them below the brief's core requirements.
Experience
4 years building and maintaining WordPress websites for small businesses. I started at a digital marketing agency in Medellín, and for the last 2 years I have worked as a freelancer for clients in the US and Colombia — restaurants, a law office, an online clothing store and several service businesses. I build the sites with Elementor Pro and Astra, do the monthly maintenance and updates, and fix problems when the clients report them. I am very responsible and detail-oriented, and my clients trust me with their websites.
Tools
WordPress, Elementor Pro, Astra, WooCommerce (basic), Yoast SEO, UpdraftPlus, Wordfence, WP Mail SMTP, cPanel, FileZilla, HTML, CSS, basic PHP
Scenario answer
Asked: You inherit a WordPress site that hasn't been updated in 18 months. Core and six plugins are significantly out of date. The client says 'just update everything, it'll be fine.' Walk through exactly what you do before touching the Update button — and how you'd handle the update itself if your first check uncovered a potential conflict.
I understand why the client wants to update everything, because an outdated site is a security risk. But updating everything at once after 18 months can break the website, so I would be careful. Before touching the Update button, the first thing I would do is a full backup with UpdraftPlus, including the database and the files, and I would download a copy to my computer. Then I would check each plugin's page on WordPress.org to see if the new version is compatible with the latest WordPress, using the 'Tested up to' information, and read the recent reviews and support forum to see if other users are reporting problems. I would update the plugins one by one and check the website after each update to make sure everything is working, and then update WordPress. If the hosting has a staging option I would do it there first, but many small business hosting plans do not have one, so I would do the updates at night when there is less traffic. If I found a conflict, I would deactivate the plugin that is causing the problem, restore the backup if necessary, and contact the plugin's support. If they cannot fix it, I would look for an alternative plugin that does the same thing. I always keep the client informed during the whole process.
Work sample
Asked: Describe a real WordPress problem you've diagnosed and fixed: what was the symptom, what did you actually find when you investigated, what did you change, and how did you confirm it was fixed?
PROBLEM: The contact form stopped sending emails A US client with a cleaning company told me they had not received any messages from their website contact form for about two weeks, which was strange because they usually receive several every week. WHAT I CHECKED: First I tested the form myself. I got the success message but the email never arrived. I checked the form settings in Elementor and the recipient email was correct. It was not in the spam folder either. WHAT I FOUND: The hosting server was not sending the emails correctly, which is a common problem with shared hosting. WHAT I DID: I installed WP Mail SMTP and connected it to the client's Google Workspace account, so the emails are now sent through Gmail and not through the server. I also made sure 'Collect Submissions' is active on the form, so in the future the messages are saved in WordPress too. HOW I CONFIRMED: I sent a test email from WP Mail SMTP and it arrived. Then I submitted the contact form again and the client received it. I asked the client to tell me when the next messages from customers arrived, and a few days later they confirmed everything was working.
Availability
Full-time on US Eastern hours from Medellín. Available in two weeks.
This is the artifact every search delivers
Post the role and the same screener runs on your real applicants — the questions arrive already written, and you read every finalist in full before any fee is due.
The same screener, other roles
- Customer Support VA
- Executive Assistant
- Bookkeeper
- Operations Manager
- Social Media Manager
- Data Entry Specialist
- Content Marketer
- Junior Developer
- Graphic Designer
- Sales Development Rep
- Real Estate VA
- E-commerce VA
- Bookkeeping Assistant
- Virtual Assistant
- Video Editor
- Copywriter
- SEO Specialist
- PPC Specialist
- Shopify Developer
- Accountant
- Project Manager
- Lead Generation Specialist
- Web Designer