What Real Research Buys You

Surface research is what most candidates do: read the homepage, skim the careers page, glance at recent press, maybe read a Glassdoor review or two. It takes 20 minutes and produces interchangeable sentences ('you are a leader in your space', 'I love your mission'). It is enough to not get caught flat-footed; it is not enough to score above average.

Deep research is what strong candidates do: read four or five engineering blog posts carefully, watch a recent conference talk by someone on the team, find adjacent job descriptions for the same kind of role at the same company, read the most recent earnings call (if public), and skim a representative sample of employee reviews looking for patterns rather than one-off complaints. It takes 60 to 90 minutes and produces three to five facts about the company that the average candidate does not have.

The payoff for the extra hour is large. Deep research changes:

• The 'why this company' answer. From generic to specific, from interchangeable to memorable.
• The 'tell me about yourself' future beat. From 'I am excited about your team' to 'the cell-based architecture migration you wrote about last quarter is exactly the class of problem I want to work on'.
• Your story selection. Instead of telling your default story bank in the same order as you tell every loop, you pick the stories that map to the values this specific company cares about most.
• Your questions for the interviewer. From 'what is the team culture like' to 'I noticed in the eng blog post on X that you used Y; can you walk me through how that decision is being maintained'.
• Your offer-stage decision. Often forgotten, but real: research that reveals red flags before the offer saves you a year of regret.

One hour of extra research compounds across every conversation in the loop. It is among the highest-leverage prep moves in the entire process.

Surface Sources vs. Deep Sources

For any company you interview at, sources fall into two categories.

Surface sources (low signal, high availability):

• Homepage and product pages
• About page and careers page mission statement
• Recent press releases written by the company
• LinkedIn About section
• The job description you applied to

Deep sources (higher signal, lower availability):

• Engineering blog posts written by their actual engineers
• Public retros, postmortems, or RFC repositories
• Conference talks by their engineers (search YouTube for talks at conferences in their domain)
• GitHub repos owned by the company, and how active they are
• Adjacent job descriptions: senior, staff, and principal roles for related teams
• Recent earnings calls (if the company is public)
• Glassdoor and Blind reviews read in aggregate (look for patterns, ignore individual venting)
• Recent press written about the company (decisions, layoffs, pivots, leadership changes)
• The CEO's, CTO's, or VP's recent X/Twitter or LinkedIn posts
• Any podcast appearances by leadership in the past year
• Articles by engineers who have left the company (more honest than current-employee marketing)

The rule: every company-specific claim you make in an interview should be sourced from a deep source, not a surface one. If your only source is the homepage, your answer will sound like the homepage.

What Deep Sources Actually Reveal

Different deep sources reveal different things. Use them targeted, not as a fishing expedition.

Engineering blog posts reveal taste. Not just what the company built, but how they talk about the trade-offs. A company whose blog posts honestly discuss what did not work, what they would do differently, and what they are still uncertain about is signalling an engineering culture that rewards intellectual honesty. A company whose blog posts only describe successes signals the opposite.

Look for specifics. A post that says 'we picked X because of Y reason, and on day three of rollout we hit Z problem we did not predict, and here is how we handled it' tells you the team thinks like real engineers. A post that says 'we are excited to announce our new platform that empowers users at scale' tells you the post is marketing.

Public RFCs and retros reveal decision processes. Not all companies publish these. The ones that do are signalling that they expect engineers to write things down and that decisions are auditable rather than tribal. Read two or three of their published RFCs and watch how the alternatives are presented, how disagreement is captured, and how decisions are concluded. This is the closest thing to seeing the actual culture from the outside.

Conference talks reveal pride. Engineers give conference talks about the work they want to be associated with. A company that sends three engineers to give talks per year about distinct problems is signalling depth and engineering investment. A company whose conference presence is one keynote by the CTO is signalling that engineering is not the foreground.

Adjacent JDs reveal priorities. The senior, staff, and principal JDs for adjacent teams reveal what the company is actively building toward. The verbs they use, the technologies they highlight, the explicit responsibilities listed: these tell you what the company is actually willing to pay for, which is the most honest version of what they value.

Glassdoor in aggregate reveals friction. Individual reviews are noisy. Patterns are not. If 40% of recent reviews mention 'reorgs' or 'unclear strategy' or 'micromanagement', those words are doing real work. Read 30 reviews, not 3. Look for words and phrases that recur. Discard the angriest and most positive outliers; trust the middle.

Earnings calls reveal pressure. For public companies, the most recent two earnings calls tell you what the CFO is being asked about and what the CEO is being asked to defend. If the calls are dominated by margin pressure, every interview is going to subtly grade for cost discipline. If the calls are dominated by AI strategy, every team is going to be partly oriented around that.

Recent press reveals strategic moves. A layoff, a leadership change, a major pivot, a public miss on a launch: each of these tells you something about the company's current state that the marketing pages will not. A candidate who has read the recent press and walks in with no sense of the strategic context is at a disadvantage they could have closed in 15 minutes.

Identifying What This Company Cares About More Than the Average Company

Every company says they value craftsmanship, customer focus, integrity, and a bias for action. The interesting question is not what they value; it is what they value more than other companies in their tier value it.

Three heuristics for finding the differentiated values:

The repetition heuristic. Read three or four artefacts from different parts of the company (a blog post, an exec talk, an adjacent JD, a recent press release). Note any value-language word or phrase that appears in all of them. The values that recur across artefacts are the ones the company actually means, not the ones they list. If 'reversibility' or 'long-term thinking' or 'rigor' shows up in three places, it is real. If it appears once on the careers page and nowhere else, it is filler.

The cost heuristic. What does this company give up to defend its value? A company that loudly values 'speed of execution' but takes nine months to ship a feature is signalling a value they aspire to, not one they live. A company that loudly values 'engineering quality' and has no on-call rotation, no postmortems, and no public retro practice is signalling the same. Look for the values the company has paid a real cost to maintain. Those are the values the interviewer is genuinely grading on.

The contrast heuristic. What does this company explicitly distance itself from? Read leadership commentary for the things they say their competitors get wrong. 'We do not believe in moving fast and breaking things' tells you the company is grading for craftsmanship. 'We do not believe in over-engineering' tells you the company is grading for shipping velocity. The contrasts the company chooses to draw are the values they distinguish themselves on.

The goal of these heuristics is to leave with a list of two or three differentiated values, not the full list of every value on the careers page. Two or three values you can map your stories to is more useful than ten you cannot.

Translating Values Into Stories: The Right Direction

The core craft move is to translate company values into stories from your own experience, rather than translating your stories into company-shaped stories.

The failure mode is rewriting your story to fit. 'Let me tell you a story about long-term thinking' followed by a story that is mostly about debugging a production issue, with the words 'long-term thinking' jammed in. The interviewer hears the seam.

The right move is to start with stories that are honestly about the company's differentiated values, and to lead with the language the company uses for them. If the company values 'rigor', and you have a story where rigor is genuinely the through-line, frame the story with the word 'rigor' in the opening sentence. If the value is 'reversibility', frame the story around the reversibility of the decision you made. The story does not change; the foreground does.

If you do not have a story that honestly demonstrates a company value, do not invent one. Either pick a different value where you have a strong story, or be honest about the gap. Inventing fake-fit stories is the most reliable way to fail a behavioral round, because the interviewer is trained to spot the seam between the story and the framing.

A practical exercise: take your top eight banked stories from story-banking. For each, write down the two or three values it most genuinely demonstrates. Then, for any company you interview at, you can match their two or three differentiated values to the subset of your bank that authentically demonstrates each.

The 60-Minute Prep Template

This is the template you run before any onsite. Block 60 minutes. Time-box each section.

Minutes 0 to 10: Engineering blog (deep)

Pick the three most recent engineering blog posts. For each, write down:

• One technical decision they describe
• One trade-off they surface honestly
• One sentence about what kind of engineer would have written that post

If the engineering blog has no recent posts (last six months), that is itself a signal: either the company does not invest in engineering communication, or the team has been in heads-down mode. Note which.

Minutes 10 to 20: Adjacent JDs and team descriptions

Find two senior or staff JDs for related teams (not the role you are interviewing for). For each, note:

• The verbs in the responsibilities section (do they say 'implement', 'own', 'define', or 'set strategy'?)
• The technologies and product surfaces mentioned
• The values phrases that appear (and which ones repeat across both JDs)

The verbs and the repeated values are your single richest source of how the company actually talks about senior work.

Minutes 20 to 30: Recent press and leadership commentary

Search for the company name in the past three months of news. Note:

• Any reorgs, layoffs, or leadership changes
• Any major product launches and how they were received
• Any public missteps and how the company responded
• Any public commentary by the CEO or CTO in podcasts, X/Twitter, or interviews

This tells you the strategic context the company is in right now, which the careers page will not.

Minutes 30 to 40: Glassdoor and Blind in aggregate

Read 20 to 30 recent reviews. Do not focus on individual ones. Look for:

• Words and phrases that appear in 5 or more reviews (real signal)
• Patterns about specific teams, managers, or org units (more relevant if it overlaps with the team you are interviewing with)
• The split between positive and negative themes (a healthy company has both; a company with only one tone is suspicious in either direction)

Resist the temptation to spend more than 10 minutes here. The signal compounds quickly and then plateaus.

Minutes 40 to 50: Identify the differentiated values

Using the repetition heuristic, the cost heuristic, and the contrast heuristic from earlier in this lesson, write down two or three values the company seems to care about more than the average company in its tier. Use the company's own language. If they call it 'rigor', write 'rigor', not 'attention to detail'.

Minutes 50 to 60: Map your stories

For each of the two or three differentiated values, pick which of your banked stories most authentically demonstrates it. If you do not have a strong match for one of the values, note the gap and decide whether to pick a thinner story or be honest if asked.

The deliverable from this 60 minutes is a one-page sheet you can review on the morning of the onsite:

• Three facts you learned from deep sources (one for the 'why this company' answer)
• Two or three differentiated values in the company's own language
• The story from your bank that maps to each value
• One thoughtful question you can ask the interviewer based on what you read

That one-page sheet is your prep asset for the entire loop.

Worked Example: The 60-Minute Sheet for a Hypothetical Onsite

For illustration, here is what the deliverable looks like after a real 60-minute prep session for a hypothetical fintech-infrastructure company. (Names are placeholders so that the example shows the shape rather than recommending a specific company.)

[ Three facts from deep sources ]
  1. March eng blog: rebuilt settlement pipeline as event-sourced;
     honestly discussed operational cost of replay and consistency edge cases.
  2. Senior platform JD verbs: 'define', 'establish patterns', 'multi-quarter
     strategy'. Staff bar is real; not just a senior-with-bigger-projects role.
  3. Recent CFO commentary on earnings: doubling enterprise headcount; team
     I am interviewing with is the platform underneath that.

[ Three differentiated values ]
  1. Rigor about trade-offs (appears in two blog posts, the staff JD, and a
     CTO podcast).
  2. Long-term thinking on data integrity (the eng blog post is unusual in
     its industry for honestly discussing the cost of replay).
  3. Cross-team leverage (staff JD explicitly: 'multi-team patterns adopted
     more than once').

[ Story mapping ]
  Rigor about trade-offs        -> Payments DB migration (read replica chosen
                                   over sharding; trade-offs explicit).
  Long-term thinking on data    -> Refactor of event log we did to make replay
                                   safe (genuine fit).
  Cross-team leverage           -> Canary playbook adopted by two other teams
                                   the next two quarters.

[ One thoughtful question ]
  In the March eng post you described the operational cost of replay as a
  real ongoing cost. How is the team thinking about that cost six months in?
  Is it lower than predicted, higher, or about right?

This is the asset. It fits on one page. It took 60 minutes. It will improve every conversation in the loop.

Reading Engineering Blogs Strategically

The engineering blog is the single richest source for most onsites, so it is worth a focused note on how to read it.

First, filter by recency. Posts older than 18 months tell you about historical culture, not current. The last 12 months of posts is what matters.

Second, read the trade-off paragraph, not the headline. Most posts have one paragraph somewhere in the middle that says 'here is what we considered, here is what we picked, here is what is still hard'. That paragraph is the highest-signal content. The intro and conclusion are usually marketing.

Third, note the writer's level of detail. Are the numbers concrete? Are the failure modes named? Is the post written in first-person plural ('we considered, we chose, we hit') or in passive marketing voice ('it was determined that')? The voice tells you the writer's posture toward the reader.

Fourth, track who writes the posts. If most posts are by managers or by 'the engineering team', the company tends not to give individual engineers a public voice. If posts are by named ICs at various levels, the company invests in IC visibility, which usually correlates with stronger IC career paths.

Fifth, look for the absent topics. What does this company not write about? A company with no posts on reliability, on-call, or postmortems is signalling either that those areas are mature and uninteresting (sometimes true) or that they are not invested in (more often true). The omissions tell you almost as much as the inclusions.

What to Do When Research Reveals Red Flags

Deep research sometimes surfaces concerning signal: a recent leadership departure, a Glassdoor pattern about burnout, a public misstep handled badly, a layoff six weeks ago. The candidate's question is what to do with that information.

Three options, in order of preference:

1. Address it in your questions to the interviewer. Most loops include a 'do you have any questions for me' segment. Use one of your questions to ask about the red flag, framed neutrally. 'I noticed there was a reorg in the platform org last quarter. How is the team feeling about it now?' or 'I read about the production incident in March. Can you walk me through what changed in your engineering practice as a result?' These questions land well because they show you have done the work. They also give the company a chance to either reassure you or to confirm the concern.

2. Adjust your level of investment in the loop. If multiple red flags accumulate, a sensible response is to take the loop seriously but with a more conservative posture: ask harder questions, push back on hand-wavy answers, treat the interview as a two-way evaluation. You are gathering information, not just performing.

3. Walk away if the red flag is dispositive. This is the rarest and most useful application of deep research. If the company has just laid off 30% of engineering, or if the Glassdoor pattern shows a recent collapse in employee sentiment, or if the leadership commentary suggests the strategy is incoherent, you may correctly decide that this is not a loop worth running. The 60 minutes of research has saved you the 8 hours of onsite plus the eventual decision regret. That is a return.

Do not weaponise red flags in the interview itself. The interviewer is generally not in a position to defend their company's macro decisions, and pressing them in that direction is unkind and unproductive. Hold the information for the offer-stage decision, where it belongs.

Bridge to the Next Lesson

This lesson taught you to do the deep research that grounds every company-specific answer. The next lesson, Handling Curveball & Hypothetical Questions, addresses a different category of risk: what to do when the interviewer asks something you did not prepare for. Research helps with the prepared questions; the next lesson is about the unprepared ones. Together they cover the two failure modes of behavioral interviews: insufficient prep on the question you should have prepared for, and brittle delivery on the question you could not have prepared for.