Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/MajoRodri/HRIA/llms.txt

Use this file to discover all available pages before exploring further.

The applies column appears to be a direct measure of job demand — how many candidates applied to a given posting. It is not. It is a measure of one specific application mechanism: LinkedIn Easy Apply, a feature that allows candidates to submit their LinkedIn profile as an application without leaving the platform. Postings that route applicants to an external URL — the employer’s own ATS, a third-party application portal, or a company careers page — record applies = 0 or null regardless of how many candidates actually applied through that external link. The applies column is not a count of applications. It is a count of a specific subset of applications, structurally set to zero for a significant portion of postings.

​
Two Application Types

LinkedIn postings use one of two application_type values that determine how the applies column behaves:
Application TypeBehaviorapplies Column
SIMPLE_OPTLinkedIn Easy Apply — candidate submits profile in-appCounted (non-zero where applications received)
COMPLEX_OPTRedirects to external URL — candidate applies on company websiteAlways 0 or null — LinkedIn has no visibility into external submissions
A posting from a company with a robust Workday or Greenhouse ATS that routes all applications externally will record applies = 0 even if 500 candidates submitted applications through the company’s career portal. A competing posting using Easy Apply will show applies = 47 from a smaller candidate pool. The Easy Apply posting appears far more “popular” by the applies metric alone — but this comparison is meaningless because the two postings use incomparable application mechanisms.

​
The Numbers

MetricValue
applies column null rate~80.6%
Median applies among non-null values4
Application type: SIMPLE_OPT (Easy Apply)Minority of postings
Application type: COMPLEX_OPT (external link)Majority of postings by many large employers
The 80.6% null rate on applies reflects the structural reality that most large employers — who are over-represented in this dataset — use their own ATS systems and external application links rather than LinkedIn Easy Apply. Easy Apply is used more frequently by smaller companies and individual recruiters who lack enterprise ATS infrastructure. The median applies of 4 among non-null values is strikingly low. This figure does not mean the average LinkedIn posting receives 4 applications — it means the average Easy Apply posting receives 4 in-platform submissions, which represents only the fraction of a posting’s total application volume that came through Easy Apply specifically.

​
Impact on Analysis

​
Views vs. Applies Correlation

Any correlation analysis between views and applies is structurally compromised by this asymmetry. A posting that receives 10,000 views and routes applicants externally will show applies = 0, breaking any linear relationship between visibility and measured applications. Phase 4, Visualization 5 (views vs. applications scatter plot) must be interpreted with this caveat front-of-mind: the scatter pattern reflects the Easy-Apply-vs-external split as much as any genuine engagement relationship.

​
Demand Signal Distortion

Jobs with external application links appear to have zero demand in the applies column, even if they are the most competitive postings in the dataset. This distortion means:
  • Easy Apply postings look more popular than external-link postings by this metric alone
  • Large enterprise employers (who predominantly use external ATS) appear to have low application volumes
  • Niche or startup postings using Easy Apply may appear disproportionately “hot” compared to equivalent enterprise roles
  • Any model that uses applies as a demand signal will encode this structural incomparability as a real demand difference

​
Imputed Values

In HRIA analyses where missing applies values are imputed (e.g., median imputation with value = 4), the imputed values represent Easy Apply application counts — not total application volume. Presenting these imputed values as “applications received” would be misleading. The true application volume for external-link postings is unknown and likely significantly higher than 4.
Imputed applies values (median = 4) should be treated as a floor estimate — the minimum observable Easy Apply activity — not as a representative measure of actual candidate interest. For postings using external application links, actual application volume may be 10–100x higher than any imputed figure.

​
Comparing Application Types

# Compare Easy Apply vs external link postings
app_type = df.groupby('application_type').agg(
    count=('job_id', 'count'),
    median_applies=('applies', 'median'),
    median_views=('views', 'median')
)
print(app_type)
Running this comparison reveals the structural split in the dataset between SIMPLE_OPT and COMPLEX_OPT postings and quantifies the difference in observed applies rates — confirming the incomparability of the two groups and the magnitude of undercounting in external-link postings.

​
Mitigation Strategies

Use views as the primary engagement metric for all postings. Views are recorded by LinkedIn for both Easy Apply and external-link postings, making it the only engagement signal that is comparably measured across the full dataset.
StrategyDescriptionWhen to Apply
Segment by application_typeAnalyze SIMPLE_OPT and COMPLEX_OPT postings separately; never combine raw applies countsAll applies-based analysis
Use views as primary metricViews are counted for both application types and provide a comparable engagement signalDemand and competition analysis
Treat applies as lower boundReport applies values as “at least X Easy Apply submissions” rather than “X applications received”Client-facing reports
Filter before correlation analysisWhen analyzing views-to-applies conversion, filter to SIMPLE_OPT postings onlyCorrelation and funnel analysis
Acknowledge in visualizationsAdd dataset footnotes to any views vs. applies scatter plot noting the COMPLEX_OPT applies = 0 structural issuePhase 4 visualizations

Build docs developers (and LLMs) love

Get started for freeTalk to us