The Why-This-Company Answer Trace, One Employer From Filings to 90 Seconds

You have a screen with a specific company next week, and you need an answer to "why do you want to work here" that names something real about them and ties it to work you have actually done. This guide follows one candidate and one public employer from the annual report, the earnings call, the posting and the team's blog, through a dated claim list, to a timed answer. It keeps the four drafts that failed along the way. By the end you will have a 60 to 90 second answer you can audit line by line, plus a recruiter cut and a hiring-manager cut.

What makes a why-this-company answer specific enough to pass

A specific answer has two traces. Every claim about the employer points to a dated public source, and every claim about you points to a line on your resume. If either trace is missing, the answer reads as praise. Praise is the thing hiring managers say they reject.

The evidence that generic answers cost you is fairly consistent, although some of it is old or secondhand:

• In an older CareerBuilder survey of hiring managers, reported by HRKatha, 45% said candidates knew nothing about the job or company.
• Resume Now surveyed 925 HR workers. 78% said personalised details signal genuine interest and fit, and 36% named generic content as a top reason for rejecting a resume.
• A Gartner survey of 3,290 candidates found 39% used AI in their last application, and 29% generated assessment answers with it.
• Greenhouse found 65% of hiring managers say they have caught applicants using AI deceptively, and AI-generated scripts made up 32% of those cases.

The AI angle is less clear than it sounds. In Tilburg's second experiment, recruiters did no better than chance at identifying AI-written cover letters. Yet 62% of hiring managers in the Resume Now survey say uncustomised AI-generated resumes are more likely to be rejected. Put those two findings together and the conclusion is simple: interviewers cannot reliably spot the tool, but they do penalise the missing specifics. A source-traced answer fixes the missing specifics no matter how you drafted it.

I could find no published survey that counts AI detection in spoken interview answers, and no published scorecard wording for generic praise. Both are not established. What is published is how structured interviews work. Google's re:Work guide describes interviewers who "ask a set of planned, rigorous, and relevant interview questions and use a scoring guide". One published sample rubric scores on a four-point scale: Poor, Mixed, Good, Excellent. An answer you can audit gives the scorer something concrete to mark.

The worked example: one candidate, one listed employer, four documents

The example is a backend engineer with a first-round screen next week for a platform team at a publicly listed software company. I keep the employer unnamed and only paraphrase its filing language, because the point here is the decisions, and those carry over to your own case. The specific risk text does not carry over. You will read your own employer's text yourself.

The candidate's resume has four lines that matter here:

• R1. Owned the migration of the billing service from a self-hosted queue to a managed one; wrote the rollback plan and the on-call runbook.
• R2. Built cost dashboards that let each service team see its own cloud spend.
• R3. Mentored junior engineers through their first on-call rotation.
• R4. Part of the team that launched the company's mobile app.

R4 is already a warning sign, and I come back to it in the failure modes. The candidate pulled four documents:

1. The latest 10-K, from EDGAR.
2. The prior-year 10-K, from EDGAR.
3. The latest earnings call transcript. If an 8-K exhibit exists, use that over the IR page.
4. The posting, saved with the date it was captured, because postings get edited.

Then the team's engineering blog, which the candidate reads in the filter step.

Why the 8-K exhibit matters: companies decide for themselves whether to post a transcript. The SEC recommends a replay or transcript but does not require one. Archives also disappear. NorthWestern Energy's filing says its archived webcast "will be available for one year". Sturm, Ruger attached a call transcript as an 8-K exhibit but reserved the right to discontinue that availability at any time. When a transcript is filed as an exhibit, as Broadridge did with its February 8, 2017 call as Exhibit 99.1, it stays in EDGAR. So cite the exhibit where one exists. Use EDGAR full-text search, or the company's filing index, to check whether recent 8-Ks carry a transcript.

How to read a 10-K and earnings call for interview prep

Read the 10-K for change, not for content. A risk factor that is new or rewritten since last year is a dated, checkable claim that a mission statement can never be. Three sections do almost all the work:

The SEC's investor bulletin on reading a 10-K describes Item 1A as the most significant risks to the company and notes that companies generally list them in order of importance. It describes MD&A as management telling its story "in its own words". Since the 2020 Regulation S-K amendments, filers must group risk factors under headings and add a summary if the section runs past 15 pages, which makes the diff faster.

On time: a complete 10-K read takes 4 to 8 hours for a company the size of Apple or Amazon. Experienced analysts skim one in 60 to 90 minutes by focusing on MD&A, the key statements and new risk factors. You only need the skim. Sources disagree on order. One says to read MD&A first, another says to start with Item 1 and Item 1A. I start with Item 1 so I know which segment the team belongs to before I read the risks.

Practitioners explain why the diff matters. Risk factors are legal drafting, so a new one means "someone inside the company decided this year that the exposure was material enough to disclose." Reused risks tell you nothing about this year.

In the example, the candidate's diff surfaced three items, all paraphrased here:

• E1. A rewritten Item 1A risk about depending on third-party cloud providers for availability and cost.
• E2. A newly added Item 1A risk about regulation of AI features.
• E3. In the latest call's prepared remarks, management returned to infrastructure efficiency, and an analyst asked about it again in Q&A.

The transcript is a recency check, not a primary source of team detail. Nobody has published whether the prepared remarks or the Q&A more often name team-level priorities, so record which part the item came from and do not assume either is better.

The next fork is the one most candidates skip: is the item new, and does this team own it?

The posting said the team owns "reliability and cost of the cloud platform", and a recent post on the team's blog, written by one of its engineers, described moving services to managed queues. That makes E1 a lead claim and E3 a supporting one. E2 is new but company-level. It goes to the top-left quadrant, and the third discarded draft below shows what happens when you lead with it anyway. The blog post becomes E4, a team-owned claim with its own byline and date.

The procedure, from filings to a timed answer

The procedure is seven steps and takes about 3 to 4 hours. That is far more than the 20 to 30 minutes or 60 to 90 minutes that ranking pages suggest, and the extra time goes into the diff and the filter, which is where the specifics come from.

The ledger is the working document. The answer is written from the ledger and from nothing else.

CLAIM LEDGER | Employer: [name] | Screen date: [date] | Team: [team name from posting]

KEPT
ID | Employer claim (paraphrased) | Source doc + section | Date | Link | Team owns it? (posting line or blog byline) | Paired resume line
E1 | [rewritten risk or trend] | [10-K, Item 1A] | [filing date] | [EDGAR link] | [quote the posting line] | [R#]
E2 | [call remark] | [8-K exhibit, prepared remarks or Q&A] | [call date] | [link] | [posting line] | [R#]
E3 | [team blog post] | [blog, author] | [post date] | [link] | [byline is on the team] | [R#]

DISCARDED
ID | Claim | Reason: mission-parrot / stale / team does not own / uncheckable figure / boilerplate / "we" story

Step 4 is where the candidate side usually falls apart, because a resume rewritten many times stops matching what you actually did. If your resume exists in several versions, Refolk builds it from your own work history and scores how well you fit the posting. You get first-person lines you can pair against each claim, and you see which of the posting's requirements have no line behind them. In the example, E1 paired with R1 and R2, E3 paired with R2, and the team's blog post paired with R1. R3 had nothing to pair with, so it stayed out.

How this goes wrong: the four discarded drafts

Most first drafts fail in one of four predictable ways, plus two quieter ones. Find yours in the drafts below before you write a second one. Each of these drafts was a real stage in the example, and each fails a specific check.

Draft 1: the mission-parrot

"I've always admired your mission of helping businesses make better decisions, and your culture of innovation really resonates with me. I'd love to be part of a team that's changing the industry."

This answer would work for any employer, which is exactly the personalisation gap the Resume Now respondents describe. It also has zero traces. Check: put a competitor's name in. If the answer still reads true, discard it.

If a competitor's name fits your answer just as well, the answer was never about this employer.

Draft 2: the stale-news draft

"I was excited by the product launch I read about, and I'd like to help scale it."

The source was a press article that predates the latest call, and on that call management had already moved on to infrastructure efficiency. The interviewer hears that the candidate's research stopped a quarter ago. Check: is there a newer filing or call that overrides this claim? If the source is an IR webcast, is it still live, given archives like NorthWestern Energy's that last one year?

Draft 3: the product the team does not own

"Your latest 10-K added a new risk on AI regulation, and I'd love to help you navigate that."

This is the most tempting failure, because the claim is real, dated and new. But it is a company-level risk, and the platform team's posting says nothing about it. A hiring manager for platform reliability cannot do anything with this sentence, and it suggests the candidate wants a different job. Check: the posting's responsibilities and the bylines on the team's blog. If neither touches it, the claim is context at most.

Draft 4: the figure the interviewer cannot check

"I read that you're growing far faster than the market, and I want to be somewhere with that momentum."

The growth claim came from a secondhand article and did not trace back to the filing. If the interviewer knows the real number, the candidate loses credibility. If the interviewer does not, the sentence adds nothing. The private-company version of this mistake is quoting a Form D "amount" as money raised, covered below. Check: cite only figures you can point to in a filed document, with its date.

The two quieter failures

• "We" stories. R4, "part of the team that launched the mobile app", tempted the candidate into "we shipped". Amazon tells candidates to "focus on 'I' instead of 'we'" when describing achievements. Check: every candidate claim maps to a resume line you personally owned.
• Boilerplate risk taken as a priority. In practice, many Item 1A risks are reused year after year with light edits. A risk you did not diff could be ten years old.

How long the answer should run

Aim for 60 to 90 seconds, which is where most guides converge. Kelly's guide says that is "long enough to make one specific connection between you and the company, and short enough for the interviewer to follow up", and that past two minutes you are "likely rambling."

One guide puts 45 to 75 seconds at about 150 to 200 spoken words. That is roughly 2.7 words a second, so by my arithmetic 60 to 90 seconds comes to about 160 to 240 words. Another suggests about 15 seconds on why you chose the company, 45 on connecting your experience to its needs, and 20 on how you will contribute. The example used that split. Here is the version the candidate timed at just over a minute:

"Two things in your own documents pulled me toward this team. In this year's annual report, the risk factor on depending on third-party cloud providers was rewritten, and on the latest earnings call management came back to infrastructure efficiency in the prepared remarks. The posting says this team owns reliability and cost of the cloud platform, so that is where those words turn into work.

That is the work I have been doing. I owned the move of our billing service from a self-hosted queue to a managed one, and I wrote the rollback plan and the on-call runbook. Then I built cost dashboards that let each service team see its own cloud spend, which ended the argument over whose bill it was.

Your team's post on moving to managed queues read like my last year, a few steps further along. I'd like to bring the cost-visibility side to that, and learn how you run reliability at your scale."

Every employer sentence maps to E1, E3 or the blog post. Every candidate sentence maps to R1 or R2. Nothing in it comes from outside the ledger.

[0 to 15s] WHY THIS COMPANY
"In [document] from [date], [paraphrased new or changed item]. The posting says this team owns [posting line], so that is where it lands."

[15 to 60s] MY MATCHING WORK
"That is the work I have been doing. I [resume line R#, first person]. Then I [resume line R#, first person], which [result you can defend]."

[60 to 80s] WHAT I WOULD CONTRIBUTE
"Your team's [blog post or talk, date] read like [connection]. I'd like to bring [specific skill] to that, and learn [genuine question about the team]."

The recruiter cut and the hiring-manager cut

Write two versions from the same ledger. The recruiter cut leads with the match to the posting and keeps one employer claim. The hiring-manager cut keeps the filing detail and ends on a question about the team. No published source varies length or content between a recruiter screen and a hiring-manager round, so this is my practice, not an established norm.

The reason for two cuts is who actually runs the screen. In Refolk's index of professional profiles, the people who screen and the people who own teams are distributed very differently by market:

Recruiters here means people titled Technical Recruiter or Talent Acquisition Partner. The ratios are derived from the counts.

The step from that ratio to "less team knowledge in the recruiter seat" is my inference, not a measured fact. Still, it fits a second pattern. The top current employers in Refolk's US recruiter sample include K2 Partnering Solutions and Experis, which are agencies. Some screens are run by people who do not work for the employer at all, and they may score you against the posting rather than the team's roadmap. Check your recruiter's current employer before the call. If it is an agency, lead with the posting match.

Title conventions also differ by market, which matters if you try to look up your screener:

This comes from a top-25 sample in Refolk's index, so it is directional only. A UK candidate who searches "technical recruiter" will mostly miss the person who is screening them.

In the example, the two cuts looked like this:

• Recruiter cut. One sentence on the posting's "reliability and cost of the cloud platform" line plus the rewritten cloud-provider risk (E1), then R1 and R2, then the contribution line. The filing diff detail and the blog reference come out.
• Hiring-manager cut. The full answer above, plus a closing question about how the team weighs reliability against cost now that management has raised infrastructure efficiency on the call.

The hiring-manager cut also has to fit structured loops. AWS says that "in every interview, regardless of the role or level, candidates are assessed on their alignment with these principles," and a secondary source reports that each Amazon interviewer is assigned specific Leadership Principles to evaluate. How Amazon officially scores "why Amazon" is not established. For any employer with published principles or competencies, label each ledger claim with the principle it touches, so the interviewer can score the story as well as the motive.

One practical point: the recruiter has the resume you sent, not your master copy. If you tailor each application, the recruiter cut should quote lines from the version that went in. Refolk keeps the tailored version for each posting next to the posting itself, which makes that check quick.

The fallback source set when the employer is private

A private employer has no 10-K, so swap in the public records that do exist and treat each one more cautiously:

1. A Form D, if one exists. It must be filed within 15 days of the first sale, which makes it one of the earliest public signals of a raise. Form D filings sit in EDGAR, which needs no API key or sign-up.
2. The posting and its history. Stands in for Item 1 when you are working out what the team owns.
3. The team's own blog, with bylines. Still the best team-ownership evidence.
4. Press releases and leadership interviews. Stand in for MD&A, with no legal drafting behind them.

Nobody has publicly established how long a blog post, press release, posting or leadership interview stays valid. You need a rule of your own, written down. My house rule, labelled as mine: a private-company source counts only if nothing newer from the employer contradicts it, and I re-open every link within 48 hours of the screen. If a newer post or posting edit changes the story, the older claim moves to Discarded with the reason "stale".

The private version of the swap test is stricter, because private sources are thinner. If your only employer claim is a funding round, the answer is still about money and not about the team. Find a team-owned claim or say less.

Before the screen: the audit and keeping it current

Audit the answer the day before, line by line against the ledger. Do not rely on memory. Re-check sources that can move: postings get edited, IR archives expire, and a new filing or call can land between your research and your screen.

Keep the ledger after the screen. If you reach the hiring-manager round weeks later, run the diff again only against what has come out since: a new 10-Q, a new call, a posting edit. Usually that takes minutes, not hours. When you apply to a second employer, reuse the procedure, not the claims. The resume lines carry over, but every employer claim has to be found again from that company's own documents.