candidate-social-research

Candidate Social Profile Research

You own one outcome: a written findings summary that tells HR what a candidate's public LinkedIn, Instagram, Facebook and TikTok presence shows about two things, job-relevant conduct and credential accuracy. Every finding meets four conditions:

• It is attributed to the right person by a counted anchor threshold.
• It is cited with URL, post date and access date.
• It maps to one of six defined categories.
• It contains nothing the hiring decision may not lawfully consider.

The summary goes to a decision-maker who never sees the raw profiles. Anything you write is what they know. Anything you leave out, they cannot be influenced by.

The work is judged mostly on what you leave out. A weak report lists everything found: a hijab in a profile photo, a pregnancy announcement, a march placard, a hospital check-in. Each of these contaminates the decision and creates discrimination exposure under Title VII, the ADA and ADEA, the Equality Act 2010, UK GDPR and EU GDPR. A strong report does three things in a fixed order. It confirms identity against at least 3 independent anchors. It reports only material in categories A–F. It states, in one fixed sentence, that excluded material existed, without a word about what it was.

When to use

• A recruiter sends a candidate name, CV and role and asks for "a social media check" before final interview or offer.
• An offer is pending conditional on background screening, and the employer's written policy includes a public social media review.
• A hiring manager forwards a LinkedIn URL and asks whether the candidate's stated job history matches their public profile. Run Steps 1, 2 and 6 for credentials only. Steps 4–5 still apply to anything you see along the way.
• HR receives a tip about a finalist, such as "check her TikTok", and needs it assessed against policy rather than acted on informally. The tip does not lower the attribution threshold. A named handle from a tipster counts as zero anchors.
• A role with safeguarding, financial authority, public spokesperson or confidential-data duties requires documented pre-hire online conduct review.

Do not use this when:

• The request is a regulated background check: criminal records, credit, right-to-work, DBS. Route it to the licensed screening vendor or consumer reporting agency. In the US, a third party compiling this for an employer is preparing a consumer report under the FCRA. That triggers stand-alone disclosure, written authorisation and pre-adverse-action notice duties that this skill does not satisfy.
• The subject is a current employee under investigation. Use the workplace investigation process, which has its own notice, representation and proportionality rules.
• The goal is sourcing or headhunting. Use a talent-sourcing skill. This skill assumes a named applicant already in process who has been notified.
• HR wants the review before first-round interviews. Decline. The review runs only at final-interview or conditional-offer stage, so protected information cannot screen out candidates at volume.

Inputs

If you hold only some required inputs, you may complete Steps 1–2: build anchors and record which accounts exist. You may not open post feeds, record content, or report findings until role duties, consent confirmation and jurisdiction are all in hand. Deliver a "Scope incomplete" summary instead. It contains Section 2 (attribution table) and Section 7 (limitations), and lists each missing input by name. Do not fill gaps with assumptions about the role or the law.

Method

1. Build the identity anchor set before opening any platform. From the CV and application, list every verifiable identifier on a separate line in your evidence log:

   • legal name and any variant on the CV (initials, maiden name, anglicised spelling)
   • each employer, with month/year start and end
   • each city of residence or work
   • each educational institution, with graduation year
   • professional email, personal website or portfolio handle
   • any supplied photo

   Each line is one anchor. Two employers are two anchors, and a city is one anchor however many times it appears.

   • Minimum 5 anchors. With 4 or fewer, flag "low-anchor candidate" in Section 7 and raise the Confirmed threshold in Step 2 from 3 to 4.
   • Record the common-name count now. Search the full name plus the primary city on LinkedIn and count matching profiles. More than 3 means "common name": raise the threshold to 4 and note it in Section 7. The thresholds do not stack. Low-anchor and common-name together still require 4.

2. Locate and attribute accounts, in a fixed platform order. Work LinkedIn first, because it carries the most anchors. Then do TikTok, Instagram and Facebook. A Confirmed LinkedIn profile linked from another account's bio counts as 1 anchor for that account.

   • Search each platform by name, then name plus each employer, then name plus city. Stop when you have one Confirmed account per platform or have exhausted the three searches.
   • View only content visible to a logged-out visitor, or to a logged-in member with no connection, follow or friendship. LinkedIn requires login, so use an account that has no connection to the candidate and shows its real identity.
   • Count anchors that are independent of each other. A bio saying "Leeds | Brightwell Logistics" gives 2 anchors. A photo resemblance gives at most 1, never 2, and never counts alone.
   • Confirmed: 3 or more anchors (4 or more if common-name or low-anchor). Content may be reviewed.
   • Probable: exactly 2 anchors (or 3 when the threshold is 4). List the account's existence only. Content is not opened.
   • Not attributed: 0–1 anchors. List the handle in Section 2 so HR knows it was checked and excluded, and record nothing else.
   • Not found: no candidate account after all three searches.
   • Where 2 accounts on one platform both reach Confirmed, review both and list both.

3. Set the review window and boundaries. The end date is the access date. The start date is the later of two dates: 7 years before the access date, or the candidate's 18th birthday (or the inferred cutoff from the Inputs table).

   • Full-depth roles get the full window on all Confirmed platforms. A role is full-depth if its duties include any of these:
     • unsupervised contact with children or vulnerable adults
     • authority to approve payments or spend at or above £10,000 / $10,000 / €10,000
     • acting as a named public spokesperson or brand ambassador
     • handling confidential client or employer data as a stated core duty
   • Standard roles get the most recent 3 years. Review posts newest first and stop each platform after 50 consecutive posts with no category hit.
   • Any Severity 2 or 3 finding on a standard role extends that platform to the full 7-year window.
   • Never request connections, follow, friend, message or react. Never use fake, borrowed or colleague accounts, and never ask any person to show you private content. Content behind privacy settings is out of scope even when offered.
   • Material outside the window is counted, not read. Note the count for Section 5 as "Items outside review window: n". If a tip points only to out-of-window content, record "outside window, not reviewed" and stop.

4. Screen content against the six job-relevant categories. Report only material in one of these:

   • (A) Threats of violence, or advocacy of violence against people.
   • (B) Harassment or targeted abuse of a named or identifiable individual.
   • (C) Statements demeaning people because of a protected characteristic.
   • (D) Disclosure of confidential employer or client information: named client plus commercial terms, internal figures, unreleased products, or personal data of customers or colleagues.
   • (E) Admission of, or evident, illegal activity tied to a stated duty. Drink-driving counts for a driving role. Fraud counts for a finance role.
   • (F) Employment-history or credential misrepresentation versus the CV.

   Out of scope regardless of how it looks: profanity, alcohol at lawful events, lawful cannabis where legal, political or religious views, dating and relationships, family, appearance, lawful hobbies, union or collective activity, criticism of a former employer that discloses nothing confidential, and job-seeking posts.

   Ambiguous content. Satire, quote-posts, replies quoting someone else, reposts or stitches without added comment, or content plausibly voiced by another person are handled by category:

   • If it touches A–C, record it as Context unclear. It gets no severity, is listed after graded findings, and generates one clarifying question.
   • If it touches D–F, it is reported only if the candidate authored the disclosing element.
   • Otherwise drop it.

5. Filter protected information out completely, before it reaches your notes. Never record any of the following: race, ethnicity, colour, religion or belief, age, disability or health (including mental health and medication), pregnancy, maternity or family plans, sexual orientation, gender identity or reassignment, sex, national origin, marital or civil partnership status, union membership or activity, and political affiliation.

   Also remove lawful off-duty conduct protected in the role's jurisdiction:

   • California: Lab. Code §§ 96(k), 98.6, 1101–1102.
   • Colorado: C.R.S. § 24-34-402.5.
   • New York: Labor Law § 201-d. This covers political activities, legal recreational activities and legal use of consumable products off premises and off hours.

   Rules for findings that brush against protected information:

   • If a qualifying finding also reveals a protected characteristic (for example, a client disclosure filmed at a place of worship), describe only the qualifying element. Paraphrase rather than quote.
   • If the finding cannot be stated without the protected element, drop it and add a line to Section 7: "One item could not be reported without disclosing excluded information."
   • The only permitted reference to excluded material is the fixed sentence: "Information outside job-relevant categories was observed and excluded." No counts, no platforms, no hints.

6. Capture, verify and grade each finding.

   • Capture. For each finding record:

     • full URL, platform, post date and access date
     • a verbatim quote of 40 words or fewer, or a neutral description of 40 words or fewer for video and images
     • a screenshot with the URL bar and system clock visible, saved to the evidence log as YYYYMMDD_platform_findingN.png

     If a post is deleted before capture, write "Observed, not captured, deleted before archiving" and assign no severity.

   • Verify Category F. Check each CV employment and education claim against at least one source the candidate does not control, in this order:

     1. the employer's website, press releases or annual report
     2. a regulator or registry (Companies House, FCA register, state licensing board)
     3. a news article or conference programme
     4. the institution's published alumni or graduation list

     CV versus the candidate's own LinkedIn alone is an inconsistency, capped at Severity 1.

   • Use a fixed level ladder for title comparisons:

     Level	Titles
     1	Associate / Executive / Assistant
     2	Senior / Specialist
     3	Lead / Manager
     4	Head of / Director
     5	VP / C-level

     Count the difference in levels between the CV title and the independently sourced title.

   • Grade severity:

     • Severity 3 (Escalate): any Category A. Or Category B or C with 2 or more separate posts within any 24-month span.
     • Severity 2 (Discuss): a single B or C. Any D. Any E. Or an independently sourced F where the title is inflated by 1 or more levels, a date differs by more than 6 months, or a qualification is not held.
     • Severity 1 (Note): an independently sourced F with a date difference of 6 months or less and the same level. Or any CV-versus-own-profile inconsistency.

   • Age discount. A single post older than 3 years, with no repeat of the same category within the last 24 months, drops one level. Severity 1 is the floor. The discount never applies to Category A, or to D where the information is still commercially live (for example, a contract that has not expired). State the discount in the finding: "Severity reduced from 2: single post, 4 years old, no recurrence."

   • Relevance. Each finding's relevance line must quote or closely paraphrase a duty from the role description. If no duty fits, the finding is not job-relevant: delete it.

7. Draft the summary in the template, then check it.

   • Use the template in Output format exactly, with all 7 sections, in order.
   • Write observable facts with dates, not character conclusions. Write "Video of 9 Nov 2023 names client Pennine Foods, contract value £340,000 and renewal date." Do not write "candidate is indiscreet."
   • Order findings Severity 3, then 2, then 1, then Context unclear. Within a level, most recent first.
   • End with one clarifying question per finding and per Severity 1 inconsistency, phrased as an open question HR can put to the candidate. Never add a hire/no-hire recommendation, risk score or overall rating.
   • Run every Quality bar line against the draft before sending. Any failed line is fixed, not explained.

Judgement calls

Thoroughness vs proportionality. Depth follows the role's duties, not the seniority of the title. The four full-depth triggers in Step 3 are the whole list. A Head of Marketing with no spend authority, no spokesperson duty and no client data gets the 3-year, 50-post standard review. Any Severity 2 or higher finding extends only that platform to 7 years, not every platform.

Reporting a Probable account vs staying silent. Report its existence in Section 2 and Section 7, for example "A Probable TikTok account (@handle) was identified; content not reviewed." Never report or summarise its content. It converts to Confirmed only when the candidate confirms the handle in writing through HR, or a further independent anchor is found. Re-run Step 4 on it only after conversion.

Credential discrepancy vs honest variance. People round dates to the year, and LinkedIn titles drift. A mismatch becomes Severity 2 only with an independent source and one of these: a gap of 1 or more levels on the ladder, a date difference over 6 months, or a qualification not held. "Account Manager" vs "Client Account Manager" is the same level and not a discrepancy. "Jan 2017" vs "2017" is not a discrepancy. A CV-versus-LinkedIn difference of any size without an outside source is Severity 1, labelled "inconsistency for candidate to clarify".

Old content vs current conduct. Pattern and recency outweigh shock value. One post from 5 years ago with no repeat in the last 24 months is reported with its date first and drops one level. A second post in the same category inside the last 24 months removes the discount from both. Two or more in B or C inside any 24-month span is Severity 3 regardless of age.

Tips from third parties. A tip tells you where to look, not what is true. Run Steps 1–6 as if the tip did not exist. Do not name the tipster or mention the tip in the report. If the tipped account does not reach Confirmed, it is listed as Probable or Not attributed like any other.

Rules

• Never review or report content from an account below the Step 2 threshold: 3 anchors, or 4 for common-name and low-anchor candidates.
• Never create accounts, pretext, send connection, follow or friend requests, react to posts, or view private content by any route, including content a colleague offers to show.
• Never record, describe, count or hint at protected characteristics or protected off-duty conduct, in the report, the evidence log or working notes.
• Never use trait words about the candidate, such as "unprofessional", "aggressive", "unstable", "immature" or "risky". Report the dated post.
• Never invent or reconstruct URLs, dates, quotes, client names or employer details. A finding without a captured URL and post date is not a finding.
• Label every attribution as Confirmed, Probable, Not attributed or Not found. Label every finding Confirmed or Context unclear. Use no other labels.
• Issue no hiring recommendation, risk score, traffic-light rating or suitability statement.
• If a Category A finding names or identifies a specific person and is dated within the last 30 days, or states a time or place, stop work and notify HR in writing within the hour so they can contact the police or the person at risk. Finish the report afterwards.
• Never review content dated before the candidate's 18th birthday, even if it falls inside the 7-year window.
• Keep only the final report and the evidence log (screenshots plus the anchor list). Delete working notes on delivery. Tell HR the retention period their policy sets. If they have none, state the default: deletion 6 months after the hiring decision.

Output format

CANDIDATE SOCIAL PROFILE RESEARCH — CONFIDENTIAL
Candidate: [full name]          Role: [title]
Prepared for: [HR contact]      Date: [YYYY-MM-DD]
Consent/notice confirmed by: [name, date]
Jurisdiction(s) applied: [list]

1. SCOPE
Platforms reviewed: [LinkedIn / Instagram / Facebook / TikTok]
Review window: [start date] to [end date]
Review depth: [Full-depth — trigger: (duty) / Standard — 3 years, 50-post stop]
Policy basis: [employer policy name or "default six categories"]

2. IDENTITY ATTRIBUTION
| Platform | Handle/URL | Status | Anchors matched |
[row per platform: Confirmed / Probable / Not attributed / Not found]

3. FINDINGS
Finding [n] — Severity [1/2/3] — Category [A-F]
  Platform/URL:
  Posted: [date]   Accessed: [date]
  Content: [verbatim quote or neutral description, at most 40 words]
  Relevance to role: [one sentence quoting a stated duty]
  Status: [Confirmed / Context unclear]
  [Severity adjustment, if any: reason]
[or: "No findings in job-relevant categories."]

4. CREDENTIAL CONSISTENCY
| CV claim | Public source | Match / Discrepancy | Source URL |

5. EXCLUSIONS
Information outside job-relevant categories was observed and excluded.
Items outside review window: [count or "none"]

6. SUGGESTED CLARIFYING QUESTIONS FOR HR
- [Finding n] [open question]

7. LIMITATIONS
[missing inputs, defaults applied, private accounts, common-name or low-anchor flags, items dropped under Step 5]

• Length. The report body is at most 900 words, excluding tables. Order findings by severity, highest first, then by recency.

• When over length, cut in this order:

  1. Reduce Severity 1 findings to one line each (date, category, one-clause description).
  2. Remove Not attributed and Not found rows from Section 2, and replace them with one line in Section 7.
  3. Shorten Section 4 to discrepancy rows plus "All other CV claims matched."

  Never cut Section 5 or Section 7.

• Standalone findings. Every finding must be understandable without opening the URL: who, what, when and which duty it touches.

• Fixed sentence. If there are no findings, Section 3 reads exactly "No findings in job-relevant categories." Sections 4–7 are still completed.

Worked example

Input: "Social check please. Daniel Okafor, applying for Senior Account Manager, Leeds, UK. Duties: (1) manage a portfolio of 12 B2B client accounts worth c. £2.1m annual revenue; (2) negotiate renewals and pricing; (3) maintain client confidentiality under our MSA terms; (4) represent the company at client QBRs. CV attached: Account Manager, Brightwell Logistics, Mar 2020 – present; Head of Sales, Corvane Ltd, Jan 2017 – Feb 2020; BA Business Management, University of Sheffield, 2013–2016. Candidate signed screening notice 2 May 2024; LIA on file. Rachel Hughes, Talent Partner."

Analyst working notes (not sent to HR; deleted on delivery):

• Anchors (5):

  1. name "Daniel Okafor"
  2. Brightwell Logistics
  3. Corvane Ltd
  4. University of Sheffield
  5. Leeds

  Not low-anchor.

• Common-name check: LinkedIn search "Daniel Okafor" plus Leeds returns 2 profiles. That is 3 or fewer, so not a common name. Threshold: 3.

• Under-18 cutoff: university entry 2013 → cutoff 1 Sep 2013. This is earlier than the window start, so it has no effect.

• Depth: duty (3), client confidentiality, is a full-depth trigger. Window 2017-05-06 to 2024-05-06, all Confirmed platforms.

• Attribution:

  • LinkedIn /in/danielokafor-leeds lists all 5 anchors → Confirmed.
  • TikTok @dokafor_sells: bio "Account Mgr @ Brightwell | Leeds" plus a link to the Confirmed LinkedIn → 4 → Confirmed.
  • Instagram @dan.okafor: name, Leeds in bio, tagged at a Brightwell team event → 3 → Confirmed.
  • Facebook: one "Daniel Okafor" with Bristol listed, no employer or school → 1 (name) → Not attributed.

• Screening:

  • TikTok: 38 videos reviewed. One Category D hit.
  • Instagram: 112 posts reviewed. No category hits. Some posts excluded under Step 5. 14 posts dated before 2017-05-06 counted, not opened.

• Severity, Finding 1: single D, dated 2023-11-09, under 3 years old, no discount → Severity 2.

• Credentials:

  • Corvane press release of 18 Jul 2019 calls him "Sales Executive" → level 1. CV says "Head of Sales" → level 4. Gap of 3 levels, independent source → Severity 2.
  • End date: CV Feb 2020 vs his own LinkedIn Aug 2019 = 6 months, no independent source → Severity 1 inconsistency.
  • Brightwell and Sheffield: LinkedIn matches. Brightwell's team page lists him as "Account Manager". Sheffield 2016 graduation list on the university news page names him.

Output:

CANDIDATE SOCIAL PROFILE RESEARCH — CONFIDENTIAL
Candidate: Daniel Okafor        Role: Senior Account Manager
Prepared for: Rachel Hughes     Date: 2024-05-06
Consent/notice confirmed by: Rachel Hughes, 2024-05-02 (LIA on file)
Jurisdiction(s) applied: UK GDPR, Data Protection Act 2018, Equality Act 2010

1. SCOPE
Platforms reviewed: LinkedIn, Instagram, Facebook, TikTok
Review window: 2017-05-06 to 2024-05-06
Review depth: Full-depth — trigger: duty (3) "maintain client confidentiality under our MSA terms"
Policy basis: default six categories

2. IDENTITY ATTRIBUTION
| Platform  | Handle/URL                          | Status         | Anchors matched |
| LinkedIn  | linkedin.com/in/danielokafor-leeds  | Confirmed      | 5 (name, Brightwell, Corvane, Sheffield, Leeds) |
| TikTok    | tiktok.com/@dokafor_sells           | Confirmed      | 4 (name, Brightwell, Leeds, link to Confirmed LinkedIn) |
| Instagram | instagram.com/dan.okafor            | Confirmed      | 3 (name, Leeds, Brightwell event tag) |
| Facebook  | facebook.com/daniel.okafor.5821     | Not attributed | 1 (name; profile lists Bristol) |

3. FINDINGS
Finding 1 — Severity 2 — Category D
  Platform/URL: tiktok.com/@dokafor_sells/video/7299481730215574817
  Posted: 2023-11-09   Accessed: 2024-05-06
  Content: Video titled "How I closed £340k with Pennine Foods". At 0:22 a
  slide shows "Pennine Foods — £340,000 — 3-yr term — renewal 01/02/2026".
  Narration: "Biggest deal of my year, and here's the exact pricing play."
  Relevance to role: Duty (3) requires client confidentiality under MSA
  terms; duty (2) covers renewal and pricing negotiation.
  Status: Confirmed

Finding 2 — Severity 2 — Category F
  Platform/URL: corvane.co.uk/news/2019/07/corvane-wins-northern-retail-contract
  Posted: 2019-07-18   Accessed: 2024-05-06
  Content: Corvane Ltd press release: "...the contract was secured by Sales
  Executive Daniel Okafor, working alongside Head of Sales Priya Sandhu."
  CV states "Head of Sales, Corvane Ltd, Jan 2017 – Feb 2020."
  Relevance to role: Duty (1) portfolio responsibility at senior level;
  stated prior sales leadership is a basis for the seniority of this role.
  Status: Confirmed

Finding 3 — Severity 1 — Category F
  Platform/URL: linkedin.com/in/danielokafor-leeds
  Posted: profile as of 2024-05-06   Accessed: 2024-05-06
  Content: LinkedIn lists Corvane Ltd "Jan 2017 – Aug 2019"; CV lists
  "Jan 2017 – Feb 2020". Difference: 6 months. No independent source found.
  Relevance to role: Accuracy of employment history submitted for this
  application.
  Status: Confirmed (inconsistency for candidate to clarify)

4. CREDENTIAL CONSISTENCY
| CV claim                                   | Public source                                                   | Match / Discrepancy                          | Source URL |
| Head of Sales, Corvane Ltd, Jan 2017–Feb 2020 | Corvane press release 2019-07-18: "Sales Executive"; same release names a different Head of Sales | Discrepancy: title 3 levels above source (Finding 2) | corvane.co.uk/news/2019/07/corvane-wins-northern-retail-contract |
| Corvane end date Feb 2020                  | Candidate's LinkedIn: Aug 2019                                  | Inconsistency: 6 months (Finding 3)          | linkedin.com/in/danielokafor-leeds |
| Account Manager, Brightwell Logistics, Mar 2020–present | Brightwell team page: "Daniel Okafor, Account Manager"  | Match                                        | brightwell-logistics.co.uk/our-team |
| BA Business Management, Univ. of Sheffield, 2016 | University news, 2016 graduation list                     | Match                                        | sheffield.ac.uk/news/graduation-2016-management-school |

5. EXCLUSIONS
Information outside job-relevant categories was observed and excluded.
Items outside review window: 14

6. SUGGESTED CLARIFYING QUESTIONS FOR HR
- [Finding 1] Can you tell us about the TikTok video of 9 November 2023
  featuring Pennine Foods? Was the client's or Brightwell's approval given
  to share the contract value and renewal date?
- [Finding 2] Please describe your role and title at Corvane Ltd, including
  any change of title between 2017 and your departure, and who you reported to.
- [Finding 3] Your CV and LinkedIn give different end dates at Corvane
  (February 2020 and August 2019). Which is correct, and what were you doing
  between leaving Corvane and joining Brightwell in March 2020?

7. LIMITATIONS
- Facebook: the only same-name account found (Bristol) matched 1 anchor and
  was not attributed; no Facebook content was reviewed.
- Instagram Stories and Highlights older than 24 hours were not available
  and were not reviewed.
- Finding 2 relies on a single employer press release; Corvane Ltd was not
  contacted. Employment verification through the screening vendor would
  confirm title and dates.
- Retention: no employer retention period supplied; default deletion of
  report and evidence log on 2024-11-06 or 6 months after the hiring
  decision, whichever is later.

Quality bar

• Section 2 shows an anchor count of 3 or more for every account whose content appears in Section 3. The count is 4 or more where Section 7 flags common-name or low-anchor.
• No Section 2 row counts photo resemblance as more than 1 anchor.
• No word in the report names or implies race, religion, age, disability, health, pregnancy, family, sexual orientation, gender identity, national origin, marital status, union activity, political affiliation or protected off-duty conduct.
• Section 5 contains the fixed exclusion sentence word for word and no other description of excluded material.
• Every finding has a URL, post date, access date, category letter A–F and severity 1–3, or the label Context unclear with no severity.
• Every Content line is 40 words or fewer.
• Every Relevance line quotes or names a numbered duty from the role description.
• Every Severity 2 or 3 Category F finding cites a source the candidate does not control.
• Findings appear in severity order, highest first, then most recent first.
• Section 6 has one question per finding, each tagged with its finding number, and no question asks about excluded material.
• The report contains no hire or no-hire recommendation, score, rating or trait adjective about the candidate.
• The header and Section 1 contain no unfilled brackets. Consent name and date, jurisdiction, review window and review depth are all stated.
• No reviewed content is dated before the window start date or before the candidate's 18th birthday.
• The report body is 900 words or fewer, excluding tables.

Failure modes

Wrong person's posts in the report. Cause: attribution on name plus photo resemblance, or on a tipster's handle. Check: every reported account in Section 2 shows 3 or more anchors (4 where flagged), photo counted at most once, tip counted as zero.

Report reads as a character verdict. Cause: interpretive words crept into Content or Relevance lines ("reckless", "boastful", "poor judgement"). Check: search the draft for adjectives applied to the candidate and replace each with the dated, observable post.

Protected information leaks through a quote. Cause: a verbatim quote or description incidentally shows religion, health, family or location of worship. Check: reread every Content line against the Step 5 list. Paraphrase the qualifying element only. If the finding cannot stand without the protected element, drop it and add the Step 5 line to Section 7.

Trivial findings inflate concern. Cause: swearing, nights out, political posts or complaints about an old boss logged as findings. Check: every finding maps to exactly one of A–F and its Relevance line names a numbered duty. Delete any that fail either test.

LinkedIn shorthand treated as a credential lie. Cause: CV compared only with the candidate's own LinkedIn and graded Severity 2. Check: every Category F finding at Severity 2 or above lists a source URL outside the candidate's own accounts. Otherwise regrade to Severity 1, "inconsistency for candidate to clarify".

Severity creep from old content. Cause: a single post 4–6 years old graded at full severity because it was striking. Check: every finding older than 3 years shows an age-discount line, or a stated reason the discount does not apply (Category A, live D, or a repeat within 24 months).

Probable account content slips in. Cause: the analyst opened a Probable account "just to check" and a finding followed. Check: no Section 3 URL belongs to a handle marked Probable or Not attributed in Section 2.

Window or age boundary breached. Cause: an old post surfaced by a tip was reviewed anyway. Check: every post date in Section 3 falls on or after the Section 1 start date and after the under-18 cutoff.

License

MIT