The Profile Findability Log, Six Weekly Readings to Keep, Revert, or Retest
Run one LinkedIn profile edit through six weekly readings and give it a written keep, revert, or retest verdict against a noise band set from your own baseline.
Key takeaways
- Search Appearances counts every search that returned your profile, by recruiters or anyone else, so a rising count proves nothing until the share of recruiter titles in the searcher list rises with it.
- Nobody has published how much weekly Search Appearances swing with no edits, so the noise band has to be your own baseline mean plus or minus the largest week-to-week swing you recorded before editing.
- In Refolk's index, 84% of a UK sample of recruiter-titled profiles hold the title Talent Acquisition Partner, against 20% in the US, so a UK job seeker who counts only Recruiter titles will misread an edit that is working.
- Refolk's index shows 6.47 data analyst profiles per recruiter-titled profile in the UK against 2.29 in the US, which is why a noise threshold from one country cannot be reused in another.
- Edit right after a weekly reading, log the next reading as partial because re-indexing is claimed to take 24 to 72 hours, and judge only on the two clean readings after that.
- At 0 to 3 a week, inbound recruiter messages are too sparse to decide a test, so decide on appearances and searcher titles and use messages only to confirm.
This is the working log for anyone who changed their LinkedIn headline or title fields to get found by recruiters for a target role and now needs to know whether the edit did anything. It follows one case, a UK data analyst aiming at fintech roles, through six weekly readings and ends with a written keep, revert, or retest verdict. The bar for "it worked" comes from your own two-week baseline, not from a percentage taken from someone else's profile.
What does LinkedIn Search Appearances actually count?
Search Appearances counts how many times your profile appeared in someone's search results over the past week. It does not count clicks, profile views, or messages, and it includes any search, by a recruiter or by anyone else.
Third-party guides agree on both points. I could not find a LinkedIn Help article that defines the metric, so treat this definition as well supported but not official.
The "anyone else" part is why this log exists. Classmates, peers in the same role, vendors and founders all run searches. A headline edit can raise your count by pulling in more of them while recruiters see you no more often than before.
What you see next to the count depends on your account, and LinkedIn has not published how free and Premium accounts differ. Its Premium help page says Premium viewer insights include how viewers found you (search, Feed, and so on), which profile sections matched their search keywords, and the number of viewers in the past 90 days with the change from last week. Third-party sources disagree about free accounts. One says you need Premium to see keywords. Another says the free view shows the company, job title and keyword data of people who found you. One third-party guide says free accounts reach the page at linkedin.com/analytics/search-appearances. Open it on your own account and note which breakdowns appear before you plan a log column you may not be able to fill.
The four readings you can log every week
| Reading | What it proves | When it lies |
|---|---|---|
| Search appearances | You were returned in searches | Peer searches count the same as recruiter searches |
| Searcher titles and companies | Who runs the searches that return you | A short list shifts a lot when one person changes |
| Profile views | Someone clicked through | Free accounts see only the last 5 named viewers over 90 days |
| Inbound recruiter messages | A recruiter judged you a fit | At 0 to 3 a week, too few to pass any threshold |
Premium keeps viewer history for longer. LinkedIn's help page says viewer data can be filtered up to 365 days for Premium Business and Premium Career, and Premium also reports how many viewers found you through search. Nobody has published conversion rates from appearance to view or from view to message. Don't grade your funnel against a benchmark. Grade each week against your own earlier weeks.
Why the noise threshold must come from your own baseline
The threshold has to come from you for two reasons. Nobody has published how much weekly Search Appearances move when you change nothing. And the pool of people searching around a role differs too much between markets for one number to work everywhere.
The best public description says the count resets weekly and "can swing widely with no change on your part," because it also depends on how many people happened to search your terms that week. Ranking advice says to change one variable and watch Search Appearances for 7 to 14 days. None of the ranking pages I reviewed sets a noise band (the range of weekly movement you see with no edits). Without one, a before-and-after story compares two single weeks and has no way to tell a real result from an ordinary swing.
Refolk's index shows how different the conditions are for the same role in two English-speaking markets.
| Measure (Refolk's index) | United States | United Kingdom |
|---|---|---|
| Profiles matching "Data Analyst" | 64,792 | 16,373 |
| Technical Recruiter + Talent Acquisition Partner profiles | 28,289 | 2,530 |
| Analysts per recruiter-titled profile (derived) | 2.29 | 6.47 |
The US has 3.96 times as many analysts as the UK, but 11.18 times as many recruiter-titled profiles. Two caveats. First, the recruiter counts are not limited to data hiring, so read the ratio as a measure of how crowded the market is, not as a forecast of how many searches you will get. Second, the ratio says that baselines will differ between markets, not which way your own swing will go. Your baseline answers that in two weeks.
Who counts as a recruiter in your searcher-title list?
A recruiter, for this log, is anyone holding a title that recruiters in your market actually use, and those titles change by country. In Refolk's index samples, "Talent Acquisition Partner" is 84% of UK recruiter-titled profiles but 20% of US ones. Build a list of recruiter titles for your country before your first baseline reading.
| Title (25-profile samples, Refolk's index) | United States | United Kingdom |
|---|---|---|
| "Technical Recruiter" | 16 (64%) | 3 (12%) |
| "Talent Acquisition Partner" | 5 (20%) | 21 (84%) |
| Other titles | 4 | 1 |
A UK job seeker who scans the searcher list for the word "Recruiter" would catch only 3 of the 25 profiles in the UK sample. A US job seeker doing the same catches the 16 Technical Recruiters and misses the 5 Talent Acquisition Partners. The error runs the same way in both countries: an edit that is working looks as if peers are driving it.
The figure this list feeds is recruiter share: the number of recruiter-titled entries in the week's searcher list divided by the number of entries shown that week. It is the second axis of every verdict. What it proves: the edit changed who finds you, not just how often. When it lies: if the list shows only a few entries, one person can move the share sharply. Write the denominator down every week so you can see when that is happening.
The six-reading log, step by step
The log has two baseline readings, one edit made right after the second, one partial reading, two clean readings, and a verdict. A sixth reading confirms a revert or opens the next edit. Every reading happens on the same weekday.
One edit, six weekly readings
- Freeze the profile and snapshot itWrite down the exact headline, current title, experience titles and skills, check that the experience titles already point at the target role, then make no edits. Label the snapshot v0 and date it.
- List the recruiter titles for your marketWrite the recruiter titles used in your country, starting with Talent Acquisition Partner and Technical Recruiter, and add any recruiter-style title from your own searcher list.
- Take a two-week baseline on a fixed weekdayRecord search appearances, searcher titles and companies, profile views, and inbound recruiter messages for R1 and R2, counting recruiter-titled entries against entries shown.
- Set the noise band from your baselineThe band is the baseline mean plus or minus the largest week-to-week swing. If R1 and R2 are equal or nearly so, take a third baseline reading first.
- Make one edit right after a readingChange one field on the same day as a reading and record the old wording, the new wording, and the date.
- Log the next reading as partialRecord it in full, mark it partial because of the re-indexing lag, and do not judge the edit on it.
- Take two clean readingsRecord the next two weekly readings with no other profile changes. These are the only readings the verdict uses.
- Write a keep, revert, or retest verdictCompare both clean counts with the band and the recruiter share with the baseline share, then write the verdict and a one-sentence reason.
- Confirm a revert or start the next editAfter a revert, take one more reading to check the count comes back toward the band. After a keep, the kept version becomes the new frozen snapshot.
Six readings for one edit
- R1 baselineRecord all four readings on a fixed weekday
- R2 baseline, then editRecord, set the noise band, make one edit the same day
- R3 partialLog it but don't judge it, because the re-index lag splits the week
- R4 cleanThe first reading that counts toward the verdict
- R5 clean, verdictSecond clean reading, then write keep, revert, or retest
- R6 confirmAfter a revert, check that the count falls back toward the band
Why two baseline weeks
Sources disagree on baseline length. One says to note your metrics for 7 days. Another compares the next 2 weeks with the prior 2 weeks. This log uses two weeks because one week gives you no swing at all, and a band needs at least one swing.
How to set the band
This method is my own, not taken from a published source. The band is the baseline mean plus or minus the largest week-to-week swing. A clean reading counts as a move only if it lands outside the band. With two baseline readings you have exactly one swing. If R1 and R2 come out equal or nearly so, the band is so narrow that almost any change clears it, so take a third baseline reading before you edit.
Why R3 is only partial
LinkedIn has not published how long re-indexing takes. Third parties claim 24 to 48 hours, others 48 to 72, and they say the count itself updates weekly. Either way, the first week after an edit includes a day or more of the old headline. Some advice says to edit on a Monday and check on Wednesday, but that reads a weekly metric two days into a new week and mostly measures the old version.
Why the verdict waits until R5
Advice on the measurement window ranges from 7 to 14 days to 2 to 3 weeks. One partial reading plus two clean ones puts the verdict three weeks after the edit, at the long end of both ranges, on purpose.
Reading | Date | Version in force | Status | Search appearances | Top searcher titles | Recruiter-titled / entries shown | Top searcher companies | Profile views | Recruiters among named viewers | Inbound recruiter messages | Note R1 | | v0 | baseline | | | / | | | | | R2 | | v0 | baseline, edit to v1 after | | | / | | | | | R3 | | v0 then v1 | partial | | | / | | | | | R4 | | v1 | clean | | | / | | | | | R5 | | v1 | clean, verdict | | | / | | | | | R6 | | per verdict | confirm | | | / | | | | | Noise band: mean ____ plus or minus largest swing ____ = ____ to ____ Recruiter titles counted: ____________________
Paste into a spreadsheet; leave a cell blank rather than guess, and keep the recruiter-title list fixed for the whole cycle.
The worked log: one UK data analyst, reading by reading
The case passes through every decision point in order: building the recruiter-title list, freezing aligned fields, one headline edit, a count that clears the band for the wrong people, a revert, and a narrower retest. It is a composite case built to hit every one of those points. I don't print weekly counts for it, because one profile's numbers would become a false benchmark for every reader in a different market.
The setup: a data analyst in the UK whose current position title is already "Data Analyst" at a payments company. Their headline, v0, is a descriptive line: "Turning payments data into decisions". The target is data analyst roles at UK fintech companies.
| Reading | Version in force | Status | Decision at this point |
|---|---|---|---|
| R1 | v0 | Baseline | Record four readings, make no edits |
| R2 | v0 | Baseline | Set the band, edit to v1 the same day |
| R3 | v0 then v1 | Partial | Log it, don't judge it |
| R4 | v1 | Clean | First reading that counts |
| R5 | v1 | Clean | Verdict on v1 |
| R6 | v0 restored | Confirm | Check the count comes back |
Decision 1, before R1: what counts as a recruiter
The first wrong turn comes before any edit: filtering the searcher list for "Recruiter". In the UK sample, that word matches 3 of 25 recruiter-titled profiles. The case's list starts with Talent Acquisition Partner and Technical Recruiter, and any other recruiter-style title that appears in R1 or R2 is added.
Decision 2, before R1: what to freeze
Semantic matching reads the whole profile, not just the headline. Coverage of LinkedIn's Hiring Assistant says it translates job descriptions and intake notes into role qualifications and matches them to candidate profiles. It also lets recruiters name a "perfect" candidate and find similar profiles. So the experience titles have to point at the same role as the headline before the freeze, not partway through the test. The resume Refolk drafts from your own history for a target role is a useful reference here: line up the profile's experience titles with it first, then freeze.
Decision 3, after R2: what to change
Edit v1 moves the title string to the front: "Data Analyst | SQL | Power BI". The temptation is to rewrite the About section or reorder skills on the same day. Don't. One source says plainly not to update other profile sections at the same time, and a second change makes the verdict unreadable. The old and new wording and the date go into the log.
Decision 4, at R4 and R5: when the count rises for the wrong people
This is the fork that matters most. Suppose both clean readings land above the band, but the recruiter share stays flat against the baseline, and the top searcher companies are other analysts' employers and early-stage firms rather than fintech employers. The count moved, but the people finding you didn't change. Refolk's index shows why a generic title string tends to produce this.
| "Data Analyst" title query (Refolk's index) | United States | United Kingdom |
|---|---|---|
| Profiles matched | 64,792 | 16,373 |
| Exact current title, 25-profile sample | 14 (56%) | 21 (84%) |
| Other current titles in sample | 11 (44%) | 4 (16%) |
A plain title string matches people who hold other current titles, such as founders, CTOs, and VPs of marketing, and those people both show up in adjacent searches and run them. The UK sample collides less than the US one (4 of 25 against 11 of 25), so a US reader making the same edit should expect this outcome more often. The verdict on v1 is revert, written with the reason: the count rose, the recruiter share did not.
Decision 5, at R6: confirming the revert and opening v2
The revert to v0 is made right after R5. R6 carries the same re-indexing lag, so read it with care. If R6 has already fallen back toward the band, that is good evidence the edit caused the rise and not the market. If it is still high, take R7 before you blame the edit. The next cycle tests v2, which keeps the title first but narrows it to the target: "Data Analyst for fintech | payments and risk reporting | SQL, Power BI". If R6 and R7 sit well away from the old mean, take a fresh baseline before v2. Otherwise the original band still holds.
How do you turn two clean readings into keep, revert, or retest?
Judge two things: whether both clean readings fall outside the noise band, and whether the recruiter share rose against the baseline share. Messages confirm a verdict but never decide one.
Count against band, recruiter share against baseline
| Both clean counts vs band | Recruiter share vs baseline | Verdict |
|---|---|---|
| Above | Up | Keep |
| Above | Flat or down | Revert: wrong people |
| Inside, or split | Up | Retest: two more clean readings |
| Inside | Flat | Check alignment, then revert or retest |
| Below | Any | Revert, then confirm at R6 |
Messages trail behind appearances because recruiters filter before they write. Recruiter seats must keep InMail response at 13% or above across a minimum of 100 InMails sent in every 14-day period. In early results, Hiring Assistant cut profiles viewed per InMail accepted by 62%. Recruiters therefore message fewer people who fit better. More appearances won't bring more messages unless fit improves too.
If several cycles show the count and recruiter share rising while messages stay flat, the gap is more likely fit than findability. That is the point to stop editing the headline and check fit against real postings. Refolk scores how well you fit each posting you apply to, and it is a quicker way to find the gap than another headline cycle.
Edit: [field] changed from "[old wording]" to "[new wording]" on [date], right after R[n]. Baseline: mean [x], largest swing [s], band [x - s] to [x + s]. Clean readings: R[n+2] = [count], R[n+3] = [count]. Against band: above / inside / split / below. Recruiter share: baseline [a] of [b] shown, clean [c] of [d] shown. Up / flat / down. Top searcher companies match target employers: yes / partly / no. Inbound recruiter messages (confirm only): [count] over the cycle. Verdict: KEEP / REVERT / RETEST. Reason, one sentence: Next action and date:
Fill one per cycle and paste it under the R5 row; the reason must name both axes.
How the findability log goes wrong
Most failures are false positives: a count that rose for the wrong people, rose because of noise, or rose because two things changed at once. Each one below comes with the check that catches it.
- Appearances rise, but for the wrong people. The count includes any search, so a jump from classmates or vendors looks like a win. Check: did recruiter share rise, not just the total?
- Reading recruiter titles too narrowly. In the UK sample, 84% of recruiter-titled profiles are Talent Acquisition Partners. Check: build the per-country title list before the baseline.
- Logging a noise swing as a win. The count can swing widely with no change on your part. Check: a move smaller than your baseline's largest swing is a retest, not a keep.
- Editing mid-week. The weekly window then covers two headline versions. Check: edit right after a reading and treat the next reading as partial.
- Changing several fields at once. If the headline and title change together, you can't tell which edit caused the result. Check: one field per cycle, logged with its date.
- Assuming the headline is the only thing AI search reads. Semantic matching reads the whole profile, so a headline-only edit can look like a null result while the experience titles keep pulling in the old role. Check: the headline and the experience titles point at the same target role.
- Using InMail counts as the test signal. At 0 to 3 a week, messages are too sparse to cross any threshold. Check: decide on appearances and recruiter share, and use messages only to confirm.
- Broad title keywords pulling in the wrong roles. 44% of the US "Data Analyst" title matches hold other current titles. Check: compare the top searcher companies and titles with your target employers.
A count that rose for the wrong people is not a win. It is a louder version of the wrong headline.
Where the evidence is thin
Several things this log depends on are not published. Each one comes with a way to check it on your own account.
- The metric's official definition. I could not find a LinkedIn Help page for Search Appearances. Check: read the description on your own analytics page and note it in the log header.
- The free and Premium split. It is unpublished and third-party sources contradict each other. Check: list the breakdowns your account shows before R1.
- Re-indexing time. Claims of 24 to 48 and 48 to 72 hours come from third parties. LinkedIn's only related statement is that outside search engines can take several weeks to pick up changes, which it does not control. Check: the partial-reading rule covers both claims.
- Weekly variance with no edits. Nobody has published it. Check: your baseline is the only valid measure.
- Visible headline length. The 220-character limit is well established, and LinkedIn raised it from 120. Estimates of what shows in search results range from 45 to 50 characters on mobile to about 120 on desktop. A claim of about 80 characters in the Recruiter Lite list view is unverified, and nobody has published what Hiring Assistant shortlists show. Check: search your own name on phone and desktop and note where the headline cuts off.
- Boolean search. One source, citing LinkedIn Help, says Recruiter supports full Boolean across titles, companies, skills and keywords, and that AI-Assisted Search writes Boolean behind plain-language queries. Whether Recruiter Lite still supports Boolean is disputed. Check: keep the exact target title string intact in the headline either way.
- Funnel conversion rates. None are published for job seekers. Check: compare your weeks with each other, never with a benchmark.
Keeping the log current after the first verdict
Keep reading on the same weekday and run one cycle at a time: an edit, a partial reading, two clean readings, and a verdict, which comes to about four weekly readings per edit. A kept version becomes the new frozen snapshot, and its clean readings become the reference for the next cycle.
Take a fresh baseline whenever the conditions underneath the log change: a new target role, a move to another country's market, a long pause in the log, or a reverted profile whose count doesn't come back toward its old band. Rebuild the recruiter-title list when you change country, using the market mix as the starting set. When findability is settled and the question becomes whether you fit the roles that now find you, the work moves to the resume and the posting. Refolk tailors the resume to each posting you apply to, so the materials keep pace with the profile.
Before you write the verdict
- The v0 snapshot is dated and records the headline, current title, experience titles and skills word for word.
- The recruiter-title list for your country was frozen before R1 and includes Talent Acquisition Partner and Technical Recruiter.
- R1 and R2 were taken on the same weekday, and the band is written down as a range.
- Exactly one field changed this cycle, on a reading day, with old and new wording logged.
- R3 is marked partial and was not used in the verdict.
- Both R4 and R5 were compared with the band, and recruiter share was compared with the baseline share using recorded denominators.
- Top searcher companies were checked against your target employers.
- Inbound messages were used only to confirm, not to decide.
- The verdict names keep, revert, or retest, gives a one-sentence reason, and dates the next action.
Questions job seekers ask
How long should I wait after changing my LinkedIn headline before judging it?
Wait for two clean weekly readings after one partial reading, which means about three weeks from the edit. LinkedIn has not published its re-indexing time; third parties claim 24 to 72 hours, so the first week after an edit still carries part of the old headline. Advice ranges from 7 to 14 days to 2 to 3 weeks. Three weeks sits at the long end of both on purpose, because a weekly metric needs whole weeks.
Why did my LinkedIn search appearances drop after a headline change?
It may be noise, or the edit may really have cost you searches. The count resets weekly and can swing widely with no change on your part, because it depends on how many people searched your terms that week. Compare the drop against the largest swing in your own baseline. If both clean readings fall below the band, revert and take one more reading to confirm the count recovers.
What do the searcher titles in LinkedIn Search Appearances mean?
They show the job titles of the people whose searches returned your profile, which tells you who is finding you, not just how often. That matters because the count includes any search, by recruiters or anyone else. Track the share of recruiter titles, using the titles recruiters actually hold in your country. In Refolk's UK sample, Talent Acquisition Partner outnumbers Technical Recruiter 21 to 3.
Can I A/B test my LinkedIn headline using recruiter messages?
No, not as the deciding signal. Most job seekers get 0 to 3 inbound recruiter messages a week, which is too few to separate a real change from chance. Recruiters also filter hard before messaging: seats must keep InMail response at 13% or above across 100 sends per 14 days. Decide on appearances and recruiter share, and use messages only to confirm a verdict.
Do I need LinkedIn Premium to run a findability log?
Not for the core of it, but check what your own account shows before you start. The weekly appearance count and some searcher breakdown are visible without Premium, though third-party sources disagree on exactly which breakdowns free accounts get, and LinkedIn has not published the split. Premium adds keyword matches, found-through-search counts, and viewer history up to 365 days instead of the last 5 named viewers.
How many characters of my LinkedIn headline do recruiters see in search results?
Nobody has established it. The 220-character limit is well documented, but estimates of what shows in search results range from 45 to 50 characters on mobile to about 120 on desktop, and a claim of about 80 in Recruiter Lite is unverified. Nobody has published what Hiring Assistant shortlists show. Search your own name on your phone and desktop and note where the headline cuts off.
Put this to work
Paste your career in once. Every application after that is written for you.
Drop a resume or a LinkedIn URL. I rank the live openings against it, rewrite the resume and write a cover letter for the best of them, and fill in the employer's form when you press the button. You read, you decide what goes out.
01Drop your resume
A PDF or a LinkedIn URL. About a minute, once.
02I rank the openings
Every weekday morning, the live catalog scored against your history. Up to 20 worth your time, not two hundred links.
03Each one is written up
Resume rewritten for the posting, a cover letter, a fit score. Press send, or let me fill in the form.
- New matches ranked and written before you are up.
- Every bullet stays inside what your history supports. Nothing invented.
- Queued, submitted, interviewing, offer: one screen, not a spreadsheet.
Free, no card. Up to 25 applications sent a week, and nothing is sent until you say so.