The Tracked Search Standard, and the Minimum You Must Record

This is a standard for one job: keeping a job search organised well enough that you always know your next move and never let a live application die silently. It is for someone applying across many companies at once, where the count of open threads has passed what memory can hold. What follows is a gradeable definition of a search under control, plus the checklist to verify it, so two people would grade the same log the same way.

Most trackers you can download fail because they treat completeness as the goal. They pile on columns, every cell gets filled, and the sheet looks organised right up until a live application goes cold because nobody set a follow-up date. This standard runs the other way. It names the smallest field set that earns its place, fixes the status vocabulary and the rule that nothing sits without a next action, and turns the whole log into funnel math that diagnoses where you stall.

What counts as a search under control

A search is under control when every live application has a Status, a named Next Action, and a future-dated due date, and when you can compute your own stage-to-stage conversion rates from the log. That is the entire definition. Everything else is convenience.

Note the word "gradeable." A standard that two people read differently is not a standard. So the test is mechanical: filter your log to live rows (anything not Rejected, Offer, or closed), and check two things. First, does every one of those rows carry a next action with a date in the future? Second, can you produce an application-to-callback percentage and a callback-to-interview percentage for a defined period? If both are yes, your search is under control by this standard. If either is no, it is not, regardless of how full the sheet looks.

The reason to draw the line at the workflow fields and not at column count is documented. In practice the biggest leak in a search is upstream: across public benchmark reporting, 97 percent of applicants are eliminated before they ever speak with a human. Most of the analysis columns people add - salary history, benefits notes, interviewer bios - describe events that never happen for the majority of rows. The three fields that decide whether you act on the rows that are still alive are worth more than twenty that archive the ones that are not.

Completeness is what a stale tracker looks like. Control is a next action with a date on every live row.

The minimum field set, grouped by the job each field does

Record the smallest set that lets you find a role, audit what you sent, drive your next action, and improve over time. Published floors range from a seven-field "minimum viable tracker" to a ten-column set; the useful move is not to argue the number but to group fields by the job they do and keep only what serves one.

There are four jobs a field can do:

• Identity fields help you find the role again: Company, Role Title, Source.
• Submission fields let you audit what you sent: Date Applied, and optionally the resume version.
• Workflow fields drive your next action: Status, Next Action, Next Action Date.
• Analysis fields let you improve: enough stage data to compute rates, plus a Contact where a human exists.

If a column does not fall into one of those four jobs, it is an orphan. Delete it. The test for keeping a field is simple: name the decision it changes. A field that changes no decision is overhead that makes you stop updating the sheet.

The one identity field people skip and shouldn't is Contact. A row that names a real recruiter or hiring manager is a row you can follow up on directly instead of shouting into a portal. Finding that human is its own task, and it is exactly what Refolk is for: describe the person you need at a target company and get a name and a path, rather than refreshing an application-status page that was built for the employer's convenience.

The canonical status vocabulary and its legal transitions

Adopt one fixed, short list of statuses that you control, and record the employer's raw label separately in Notes. There is a de-facto default pipeline - Saved, Applied, Interview, Offer, Rejected - but no published standard, and employers do not standardise their own labels either. The load-bearing design rule is to keep your stage separate from theirs: set your working stage from the latest real event, and let one pipeline span every employer.

Here is a working vocabulary. Each status has to be able to name a next action, or it does not belong on the list.

The transitions that are legal

No source publishes a state machine for job search statuses, so this is a design decision, and I will state it plainly as one. The rule that makes transitions legal: a status may only change when a real event moves the row forward, backward to Closed, or laterally to a follow-up state. Concretely:

• Any status may transition to Rejected or No Response. Rows die at any stage.
• Applied may go to Follow-Up Sent, then to Screen Scheduled, or to Closed.
• Screen Scheduled advances to Onsite or closes; it does not skip back to Applied.
• Onsite advances to Offer or closes.
• Offer is terminal for tracking; the decision belongs to a separate process.

What is illegal is a jump with no event behind it - moving a row to "Screen Scheduled" because you feel optimistic, or leaving it at "Applied" after a recruiter has clearly emailed you. The stage reflects the last thing that actually happened, nothing more.

The follow-up clock: when a row is due or stale

A live application with no reply becomes due to follow up at 7 to 14 days after submission, and stale once it passes that window unactioned. This interval is the single most agreed-upon number in the public advice, and it is where trackers earn their keep, because a follow-up date is the thing that keeps an application from dying in silence.

The evidence is consistent. General guidance puts the window at 7 to 10 days after submitting. Survey data lands in the same place from two independent samples.

Two boundaries matter. On the early side, following up within 24 to 48 hours of applying reads as impatient and tends to hurt candidacies, so the floor is day 7 unless the posting says otherwise. On the senior side, stretch the window to two or three weeks for executive and leadership roles, where hiring moves slower.

You also need a ghosted rule so dead rows leave the live set. Define it once and apply it mechanically: no response after a set number of days post-interview marks the row Closed (no response) while you stay polite. Pick your number - fourteen days post-interview is a defensible default - and let the log enforce it rather than your mood.

The procedure: from empty sheet to named bottleneck

Run this sequence once to set up, then loop the recurring steps weekly. The setup is about twenty-five minutes total; the weekly loop is about half an hour. The order below is deliberate: log first, act on due dates second, compute rates last. Some sources front-load source tagging to prove what works later, and that is a fair emphasis, but it does not change the sequence.

Turning the log into a diagnosis: the funnel math

Compute a conversion rate for each consecutive pair of stages, then compare each to a reference band; the stage with the lowest rate relative to benchmark is your primary bottleneck. This is what turns the log from a diary into a diagnosis. The per-stage formula is the count at the later stage divided by the count at the earlier stage, times 100 - for example, interviews scheduled divided by total applications over a set period.

Do it over a fixed window so the denominator is stable. Two rates carry most of the diagnostic weight for a volume searcher:

• Application-to-callback: callbacks divided by applications. This is your top-of-funnel rate and, for most searchers, where the constraint lives.
• Callback-to-interview: interviews divided by callbacks. This tests whether your screens convert.

Now read the number against a candidate-facing band. Be careful which band you use. Employer-funnel figures like the 3 percent applicant-to-interview aggregate or Jobvite's 8.4 percent are per-posting rates across all applicants, including unqualified mass applicants, so they flatter the individual. Use the candidate tiers instead.

The diagnostic reading is direct. If application-to-screen is very low, your targeting or criteria are off. If screen-to-interview is low, your screening presentation is too broad. If interview-to-offer is the constraint, interview quality needs work. Given that interview-to-offer now runs above 36 percent while application-to-interview sits near 8.4 percent, the constraint for most searchers is the top of the funnel, so polishing interview technique yields less than fixing who you apply to.

There is a real leverage point hiding in who you are as an applicant, and the index numbers make it concrete. In Refolk's index of professional profiles, 55,949 US Software Engineers list Python as a skill against just 602 who list Rust - roughly a 93x gap in pool size. A scarce-skill applicant faces a far smaller competitor pool, which mechanically lifts callback odds. That is often what the Healthy tier reflects: niche roles where the skill is in genuinely short supply.

If your funnel math says the leak is at the top, the fix is often not a better resume but a better-targeted list of who to apply to and who to follow up with. This is where finding the actual human beats working a portal.

How this goes wrong: failure modes and false positives

The most valuable part of a standard is knowing when a log that looks fine is not. Each failure below has a false positive - the thing that makes it look healthy - and a mechanical check you can run.

• Completeness mistaken for control. A twenty-column sheet looks organised while rows go stale. False positive: every cell filled. Check: filter live rows for a missing future-dated next action; any hit is a fail.
• Portal status treated as signal. The employer's portal is built for the employer, so its words are vague and slow. Check: is each stage set from a real event or a portal word? Only events count.
• Status inflation. Fifteen custom statuses look thorough while hiding indecision. Check: can each status name a next action? If not, delete it.
• Batching updates. Logging weekly loses details and misses follow-up windows. Check: compare Date Applied to when the row was logged; large gaps predict missed follow-ups.
• Following up too fast. A 24-to-48-hour nudge reads as impatient and hurts. False positive: high activity, low callbacks. Check: no follow-up before day 7 unless the posting says otherwise.
• Reading employer benchmarks as personal rates. The 3 percent and 8.4 percent figures are employer aggregates. Check: compute your own numerator and denominator before comparing to any band.
• Bottleneck misdiagnosis from thin data. Two interviews is not a rate. Check: require a minimum count, roughly 30 or more applications in a stage, before declaring a bottleneck.

A copy-paste starter you can adopt as policy

Below is the minimum column set and the status list, ready to paste into a sheet. It is the smallest thing that passes the grade, and you can add analysis columns later once you have enough rows to compute rates.

Company | Role Title | Source | Date Applied | Contact | Status | Next Action | Next Action Date | Notes
Statuses (fixed): Saved | Applied | Follow-Up Sent | Screen Scheduled | Onsite | Offer | Rejected | No Response
Ghosted rule: no reply 14 days post-interview -> Closed (No Response)
Follow-up clock: no reply by day 7-14 -> due (senior roles day 14-21)

You can adopt the same rule set as team policy in a two-person job-search pod or a career-services cohort, because every clause is checkable. When a coach reviews a member's log, they run the same filter and read the same bands, and they reach the same verdict.

Keeping the standard current as your search moves

The log is not a document you finish; it is a control loop you run until you sign an offer. Two things keep it honest over a long search: the weekly review that clears the live set, and the monthly funnel read that tells you whether to change what you are doing rather than just doing more of it.

Re-check the follow-up window against the role level as your search shifts. If you move from mid-level to senior applications, the due date stretches from 7-14 days to 2-3 weeks, and applying the shorter clock will make you look impatient. Re-check your bottleneck each time you cross the 30-application mark in a stage, because the constraint moves as the funnel narrows - the leak that was targeting last month can become screening this month once your targeting improves.

Finally, treat the top-of-funnel constraint as an addressable problem, not a fixed fact about you. When the math says callbacks are the leak, the highest-yield changes are narrowing your target list and reaching the human who decides, not rewriting your resume for the fifth time. Describe the people you need at your target companies, get names and warm-intro paths, and put them in the Contact field so your follow-ups land on a person. That is the difference between a log that records a stalled search and one that moves it.