Applicant flow log

Who keeps one, after the 2025 changes

For decades the applicant flow log was mainly a federal contractor document, built to support affirmative action plans under Executive Order 11246. That order was revoked in January 2025, and the OFCCP stopped enforcing its race and sex affirmative action requirements. Two contractor programs remain in force: Section 503 of the Rehabilitation Act (disability) and VEVRAA (protected veterans). VEVRAA still requires contractors to invite applicants to self-identify as protected veterans and to keep yearly data on applicants and hires for three years. OFCCP rules effective September 2026 removed the Section 503 disability self-identification invitation and disability data collection, though written affirmative action programs for covered contractors remain. Confirm current rules with counsel.

The log is also useful well beyond contractors. Title VII still prohibits selection procedures that cause disparate impact without being job related, and the Uniform Guidelines on Employee Selection Procedures (1978) expect covered employers to keep data showing whether their selection steps affect groups differently. Federal enforcement priorities on disparate impact shifted in 2025, but the statute did not, and private plaintiffs and state agencies can still bring claims. EEOC rules generally require keeping hiring records for at least one year. The glossary entry on the applicant flow log covers the definitions in more detail.

Defining who counts as an applicant

The hardest part of the log is not the spreadsheet; it is deciding who belongs in it. A requisition for 40 customer service agents might draw 2,500 clicks, 1,400 started applications and 900 completed ones. Counting every click inflates the pool with people who never really applied.

The long-standing OFCCP approach for online hiring counts someone as an applicant when they applied electronically for a specific position, were actually considered, met the basic qualifications and did not withdraw before an offer. Many employers apply a similar definition voluntarily because it is defensible. Basic qualifications must be objective and set before review: "available for the 6-week day-shift training class" works; "good communicator" does not.

If you cannot review everyone, write down a neutral method, such as the first 300 complete applications or a random sample, and apply it every time. Collect self-identification with your EEO self-identification form and keep it away from decision-makers.

Worked example: checking a stage for adverse impact

A regional bank's operations center fills 40 seats. 600 people meet the basic qualifications: 360 women and 240 men. All take a typing and data-entry assessment. 216 women pass (60%) and 120 men pass (50%).

The impact ratio is the lower rate divided by the higher: 50% ÷ 60% = 83%. That is above the four-fifths rule threshold of 80%, so on this measure alone the assessment does not show adverse impact by sex.

Now the interview stage. Of 336 who passed, 100 are interviewed: 70 of the 216 women (32.4%) and 30 of the 120 men (25%). Ratio: 25% ÷ 32.4% = 77%. Below 80%. That does not prove discrimination, especially with small numbers, but it is the signal to look at how interview slots were offered, whether scheduling favored some shifts, and whether a statistical test confirms the difference. Run the same checks by race and ethnicity. Doing this per stage, not just at hire, is the point of the log; a final hire ratio can look fine while one step is screening a group out.

High-volume hiring and your ATS

At a few hundred hires a year, a manual log is possible. At a few thousand, it has to come from the recruitment tracker or ATS. Check that your system timestamps each stage, forces a disposition code when a candidate is closed, keeps self-ID data in a separate restricted table, and can export stage counts by group. Build a monthly stage-by-stage report rather than waiting for a requisition to close; evergreen requisitions for agents or warehouse associates may never close.

Automated screening deserves a closer look. If a knockout question, chatbot screen or AI scoring step removes candidates, it is a selection procedure and belongs in your stage analysis. New York City's Local Law 144 already requires bias audits for certain automated hiring tools, and other jurisdictions have similar rules. Requirements vary by state, city and country.

Common mistakes

• No written basic qualifications, so the applicant pool cannot be defined after the fact.
• Dispositions left blank, or a single catch-all code like "not selected" for every rejection.
• Recruiters able to see self-ID answers in the candidate profile.
• Analyzing only hires, never the screen, assessment and interview stages.
• Deleting data from cancelled or evergreen requisitions before the retention period ends.