Context-Aware Video Onboarding

The problem

You hired a SaaS Account Executive in February. The plan said four months to first quota hit. It is now mid-June. They have not hit it. Their manager swears they are smart and motivated. The LMS dashboard says they completed 96% of the onboarding curriculum in their first two weeks. The CSM you hired the same week is in roughly the same place — managing a thin book, escalating things she should be handling herself, two months behind where you projected. The backend engineer hired the same month finally shipped his first feature solo last week, four months in instead of three. All three watched the same 90-minute company orientation video on day one. None of them learned how to actually do their job from it. They picked it up by Slack-DMing their manager twenty times a week, which is exactly what the orientation video was supposed to prevent.

The math is grim and nobody runs it. A $120K AE ramping for six months instead of three-and-a-half is roughly $50K of fully-loaded burn that should not have happened. Multiply by the number of AEs you hire per year. Add the CSMs at $24K of unproductive overhang each. Add the engineers at $36K-$54K. A 50-person company hiring 12 people a year is leaking $300K-$500K of margin to ramp inefficiency that the P&L will never show as a line item. It shows up as missed quota, longer ticket queues, slower feature velocity, and the 14% voluntary turnover in the first 90 days that everyone shrugs at because nobody connects the dots back to the onboarding video.

What most people get wrong

They buy an LMS and hope the platform will do the thinking. Lessonly, Trainual, 360Learning, Articulate Rise — they are competent platforms, and every head of People I have met has at least one of them in the stack. The platform is downstream. The platform delivers whatever curriculum you put into it. If what you put in is a 90-minute company-history video and a generic tools tour, the most polished LMS in the world will deliver that curriculum and your hires will still ramp at the industry-average four-to-six months. The platform is not the system. The system is the role map, the decision inventory, the manager interview, and the 4-7 minute micro-modules. Platforms run a system; they do not produce one.

They hire a video agency and end up with beautiful content that ages out in 18 months. I have audited four onboarding programs in the last two years that paid $80K-$200K to a corporate-video agency. The output is gorgeous. Studio lighting, motion graphics, a voice-over actor. The output also describes a product that has since changed, references a tool the team migrated off, and assumes an org chart that has been restructured twice. The team will not refresh the content because the per-minute production cost is prohibitive. So the beautiful video plays for two years past its sell-by date and quietly teaches the wrong answers. Loom-quality video recorded by the senior teammate in 90 minutes and edited in Descript ages gracefully and refreshes cheaply. Studio-quality video does not.

They wave at the existing Loom library and say “we already have videos.” Most engineering and CS teams have an unindexed Loom library with 40-100 walkthrough videos recorded by senior teammates over the last two years. The videos are individually useful. The library is not a curriculum. A new hire pointed at it has no idea which video to watch when, which decisions it covers, which ones are still accurate, and which ones describe systems that have been replaced. The Loom library is source material. It is not a finished training program. The work is structuring the material into role maps and sequenced modules with comprehension checks, not pretending the dump is the finished thing.

A working approach

The book proposes a six-step build sequence. Each step is light enough to ship in a focused day or two, which is the entire reason this model beats the nine-month big-bang rebuild that most L&D leaders sketch in their heads:

STEP 1 — Role map (60-90 min per role)
  Definition of competence, current ramp benchmark, fully-loaded monthly cost
STEP 2 — Decision inventory (8-15 decisions per role)
  Trigger scenario, options, right answer, why, likelihood, failure cost
STEP 3 — Manager interview (30 minutes)
  Default decisions, edge cases, escalation map, the one thing
STEP 4 — Micro-modules (4-7 minutes each)
  Hook, wrong path, pattern interrupt, right answer, why, escalation, quiz card
STEP 5 — Comprehension check
  Three-question quiz + 48-hour manager check (green / yellow / red)
STEP 6 — Metrics dashboard
  TTQA (Time to Quota), TTC (Time to Competence), FFR (First-30 Failure Rate)

The role map is the chapter most heads of People skip because they think they already have one. They do not. The exercise: write down the definition of competence for the role in one sentence (“can ship a feature to production without senior review” / “manages a book of 8+ accounts without manager involvement”). Note the current ramp benchmark and the target ramp after the rebuild. Note the fully-loaded monthly cost of an unproductive hire — for a $120K AE that is roughly $20K/month. Note expected hires per year. The numbers force honesty about what the rebuild is actually worth. The companion template in the bonus folder (role-mapping-template.md) is the structured version with sequencing cells for weeks 1-4 and a source-material inventory.

The decision inventory is the chapter that does the real heavy lifting. A decision point passes three tests: there are genuinely multiple reasonable options, a wrong choice has real consequences, and the right choice requires context the hire does not yet have. “How do I log into Slack” is not a decision point — it is a setup step. “A customer just asked for a refund five days outside our policy window — do I approve, escalate, or decline” is. A complete decision inventory for a single role lands at 12-18 rows. You prioritize the high-likelihood, high-failure-cost rows for the v1 curriculum (typically 8-10 of the 18) and backlog the rest.

Role mapping

The role-mapping conversation is best run with two senior people in the role rather than one. A single person will miss the decisions they consider obvious — the autopilot stuff they have been doing for two years without conscious thought. Two people together catch each other’s blind spots. The mechanic: ask them to walk through a typical week and listen for the word “obviously” or “of course.” Each “obviously” is a decision that is obvious to them and not obvious to a new hire. A 4-hour session with two senior people typically surfaces 14-18 decisions, which is more than enough for the v1 curriculum.

The three sources of decisions worth mapping: the senior person on the team (the head-knowledge source), the last three new-hire Slack DMs to the manager (the questions they could not figure out are the decisions you have not taught), and the last three “why did they do that” conversations (each one is a decision point that was never addressed). Together these three sources cover roughly 80% of the real decisions a new hire will hit. The other 20% you discover during the pilot retrospective and add to v2.

Manager interview SOP

The 30-minute manager interview is the highest-leverage technique in the entire book. A single conversation typically generates the scaffold for 8-12 decision-point modules — enough to build a v1 curriculum for one role. The full script is in the bonus folder (manager-interview-script.md) with a sample of two real (anonymized) interview outputs. The structure:

1-3   Setup and framing (you record, they review draft)
4-12  Default decisions ("walk me through a typical week")
13-20 Edge cases ("what do new hires get wrong most often")
21-26 Escalation map ("when X breaks, who hears it first")
27-30 The one thing ("one thing you would teach on day one")

Three discipline rules during the interview. First, get comfortable with silence — senior teammates often pause 5-10 seconds before answering substantive questions, and that silence is them retrieving something they have not consciously thought about. If you fill the silence with a clarifying question, you cut off the retrieval. Second, anchor in present tense and real examples (“walk me through the last time you made that call”) rather than aspirational ones (“what should the process be”). Third, do not skip the closing question. The “one thing” answer becomes the opening line of your first module and is the most quotable moment in the entire curriculum.

The post-interview workflow: transcribe with Otter.ai or Whisper, paste the transcript into Claude or ChatGPT with a prompt asking it to extract decision points structured as ID / name / scenario / right-answer / why, edit the draft inventory with the senior teammate over 20 minutes, ship it as the scaffold for the v1 modules. The AI drafts roughly 3x faster than starting blank. The senior teammate’s edit pass is still required for accuracy — the model gets product names, escalation paths, and team-specific jargon wrong consistently.

Video structure: 5-7 minute branching modules

The 90-minute monolith is dead. Anyone who watches video on the internet in 2026 knows the format — YouTube on 6-12 minute videos, TikTok on 30-60 second clips, podcast chapters on 5-10 minute segments. The 60-minute corporate orientation video has a documented drop-off point at 18 minutes. People stop watching before the second module ends. The 4-7 minute decision-point module is the format the curriculum is built around.

The anatomy of a single module:

Beat 1  Hook (20-30s):              "You're about to..." — the scenario
Beat 2  Wrong path (30-45s):        The intuitive but wrong answer
Beat 3  Pattern interrupt (15-30s): "Here's why that's wrong at our company"
Beat 4  Right answer (90-120s):     The actual right choice and the steps
Beat 5  The why (45-60s):           The reasoning that justifies the answer
Beat 6  Escalation (15-30s):        "If you're still unsure, do X"
Beat 7  Quiz card (20-30s):         1-3 questions to confirm retention

Total: roughly 4 minutes 15 seconds to 5 minutes 45 seconds per module. Reserve the branching scenario format — where the hire chooses an option and sees the consequence before being shown the right answer — for the 3-5 highest-stakes decisions per role. The cognitive science is well-established: information you predicted incorrectly sticks better than information you were told passively. ATD’s adaptive-learning study reported 31% higher 30-day retention on branching versus passive video for identical content. Branching costs more to build, so spend it where the wrong choice would cost real money or trust (discount approval, deploy timing, refund decisions outside policy), and keep simple modules for the rest.

Production quality matters but the trap is over-production. The middle path that works: senior teammate records in Loom or Tella with a good external mic ($80-150) and decent lighting (a window or a $60 ring light), light edit in Descript to cut filler words and add captions. Total per-module cost is 2-4 hours of someone’s time and zero external spend. The result looks intentional without being expensive, and you can refresh modules quarterly without flinching at the cost. Synthesia and HeyGen are tempting for the speed but hires can tell the speaker is synthetic within 30 seconds — reserve AI avatars for procedural how-to modules where personality does not matter, and keep your senior teammate’s face on the high-stakes decision modules where trust transfers.

Comprehension checks and the 48-hour manager loop

Watching is not knowing. A hire can play the video muted in a background tab and click “mark complete.” The completion event is not evidence of learning. The honest evidence is whether the hire can answer the question the module was supposed to teach.

Two checks. First, a three-question quiz at the end of each module — two multiple-choice that directly test the decision (the right answer is the one the module taught; the wrong answers are the plausible-but-wrong choices the module addressed) and one open-ended “what would you do” that takes the hire 90 seconds to answer and the manager 60-120 seconds to skim later. Same-surface quizzes (Articulate Rise, Lessonly, Storyline) have noticeably higher completion than separate forms because clicking out costs 20-30% of hires. If you are running the v1 on Notion + Loom, Tally pages embedded into the module page get you close to same-surface for free.

Second, the 48-hour manager check. Twenty-four to forty-eight hours after the hire completes a module, their manager sees a card in their queue: “Sarah completed the module on refund decisions yesterday. Spend 5 minutes asking her how she would handle a refund request and confirm she has the answer.” The manager has a 2-minute scenario conversation, marks it green / yellow / red, and moves on. If yellow, the system reassigns the module and flags it for a follow-up the next day. The 48-hour window is calibrated to the decay curve on new information: a hire who watched a module Monday morning and is asked about it Wednesday morning either still has it (good signal) or has lost it (caught early, can re-teach before they hit the real-world scenario). Waiting a week means the hire has already either hit the scenario in the wild (too late) or lost the information entirely.

The single most common failure of the loop is the manager skipping the check — they are busy, the check looks optional, they mark it green without doing it. The fix is cultural: the head of People makes it clear the manager is accountable for new-hire ramp, and the check is non-optional. The data backs this up cleanly. Managers who reliably do the check have hires who reliably ramp on time. Managers who skip it have hires who do not. Recurring tasks in Asana, Linear, or ClickUp that auto-assign 48 hours after module completion are the forcing function that makes the loop survive.

Metrics: TTQA, TTC, FFR

Video completion rate is a vanity metric. Eighty-five percent of hires finished the curriculum tells you nothing about whether they can do the job. The metrics that matter are about whether the hire can actually perform the role at the expected level by the expected date. Three, in priority order:

Time to Quota Achievement (TTQA) is the cleanest measure for revenue-producing roles (AE, SDR, BDR, some CSMs). Number of days from start date to the first month in which the hire hits their assigned quota. The hire either hits the number or does not — no judgment call, no subjectivity. Industry benchmarks in 2026: SDR/BDR 45-75 days, Account Executive 4-6 months, Inside / SMB AE 75-110 days, Enterprise AE (>$100K ACV) 6-9 months. A well-built role-specific curriculum compresses these 30-50%. An AE benchmark of 5 months drops to 3.0-3.5.

Time to Competence (TTC) is the equivalent for non-revenue roles. Number of days from start date to the date the manager confirms the hire can perform the core role activities without supervision. Competence is defined per role and is binary at the activity level — engineer “can ship a feature to production without senior review,” CSM “manages a book of 8+ accounts without manager involvement,” Tier 2 support engineer “handles tier-2 ticket volume without escalating to engineering.” The criteria are written in advance so the call is not subjective in the moment. Benchmarks: backend engineer 90-120 days, frontend engineer 60-90, DevOps/SRE 120-150, Tier 1 support 30-45, CSM 60-90, PM 90-120. A role-specific curriculum typically compresses these 30-40% as well.

First-30 Failure Rate (FFR) is the inverse view: how often does a new hire make a material error in their first 30 days that the team has to clean up. Most mid-market teams without structured onboarding report 1.8-3.1 material errors per new hire in the first month. A team running a role-specific curriculum gets to 0.4-0.5. The difference is not the hires — it is whether they were taught the decisions before they had to make them.

The 30/60/90 ramp dashboard is what the head of People takes to the executive team. It is what justifies the L&D investment and the maintenance budget. A cohort that hits the targets is evidence the program is producing real outcomes. A cohort that misses them tells you what to fix in the next iteration. The book includes a sample dashboard layout in Chapter 7 and a ROI worksheet in the bonus folder (ramp-time-roi-calculator.csv) with 26 roles pre-loaded — the file uses fully-loaded monthly cost and current vs. target ramp weeks to compute annual savings per role. An organization hiring two AEs per year saves $810K. An engineering team hiring two backend engineers and one senior engineer saves $246K. The numbers are conservative and they add up fast.

The 2-week pilot launch

The reason most onboarding rebuilds fail is they try to do everything at once. The Heads of People who try to rebuild for every role spend three months scoping, produce a 40-page plan, get partial sign-off, start building, and the product changes underneath them. The curriculum that ships nine months later is already 30% out of date and was never tested with real hires before launch.

The 2-week pilot avoids all of that. Pick one role. Build the curriculum for that role only. Run it with the next 1-3 hires into that role. Iterate based on what happened. Then expand to a second role with everything learned. By the time the original critics ask “where is the program for my team,” you have proof from three roles that the model works.

Week 1 is the build week. Monday: pick the role with an upcoming hire in the next 4 weeks, an articulate senior teammate, and a measurable ramp benchmark. Monday afternoon: run the 30-minute manager interview. Tuesday: convert the transcript to the structured decision inventory. Wednesday: write the 8-10 module scripts (roughly 600-900 words each for a 5-minute module). Thursday: record the modules in Loom or Tella, one at a time, with 15 minutes between each. Friday: light edit in Descript, assemble into the role-curriculum hub (Notion page or LMS course), build the quizzes (Tally or Google Forms for v1), schedule the recurring 48-hour manager-check tasks.

Week 2 is when the curriculum meets a real hire. The hire goes through standard company orientation in the morning (parental leave, benefits — the universal stuff stays generic). In the afternoon, they get the role-picker pre-roll and the first 2-3 modules. Manager checks fire 48 hours later. By Wednesday they are doing real role work and the curriculum gets its first real-world stress test. Friday day 10: the pilot retrospective with the hire and the manager. Three questions: what did the modules get right, what was missing or unclear, what is the one thing that would have made this week easier. The retrospective notes become the v2 backlog. Common v2 changes: 1-2 missing decisions, 1-2 modules that need a clearer right-answer explanation, possibly a sequencing change.

Total elapsed time: 14 days from kickoff to first hire onboarded. Versus 4-6 months for a typical big-bang rebuild. The pilot does not give you the final program — it gives you v1 plus a v2 backlog plus enough early signal to decide whether the model is working before you spend another two weeks on the next role. The book closes with the curriculum calendar that scales the model: one new role every 2-4 weeks, with iteration on previously-launched curricula running in parallel. A 200-person company with eight distinct roles can have a complete curriculum stack in 4-6 months following the cadence — half the time a big-bang rebuild typically takes, with every successive role building on the templates and patterns from the previous ones.

Included with the book

• 30-Day Role Mapping Worksheet (markdown) — the structured template covering role definition, decision inventory, 30-day sequencing, source-material inventory, branching-scenario candidates, pilot cohort tracker, retrospective notes, and the quarterly review card. Copy once per role, fill in, keep the completed worksheets in your team workspace as the canonical source of role-curriculum decisions.

• The 30-Minute Manager Interview Script (markdown) — the structured interview script with lead questions, follow-up probes, pacing guidance, two sample (anonymized) outputs from real interviews (a CSM role and a backend-engineer role), common interview mistakes, and the pair-interview variation for high-variance roles.

• Ramp-Time ROI Calculator (CSV) — 26 roles pre-loaded with current ramp benchmark, target ramp after the rebuild, fully-loaded monthly cost, expected monthly hires, and computed annual savings. Open in Sheets or Excel, edit your role mix, and the dashboard recomputes the rebuild’s financial case.

Get the full picture

Readers of this also chose

Questions readers ask

Do I need an LMS to run this?

No. The v1 can run on Notion + Loom + Tally for $0/month. The LMS conversation belongs after you have shipped 2-3 role curricula and have evidence the model works. Most mid-market companies that already have an LMS (Lessonly, Trainual, Articulate Rise, 360Learning) can run the model inside it - the platform is downstream of the curriculum design.

What if I need a refund?

Checkout runs on Lemon Squeezy. The standard refund window applies. You keep the PDF either way.

How big does the team need to be for this to be worth it?

The 2-week pilot is worth running at any size that hires more than 3 people a year into the same role. Below that, ad-hoc one-on-one onboarding is fine. Above that, the math on ramp-time savings starts to compound fast - a 50-person company hiring 12 people a year typically recovers $300K-$500K of margin per year.

How long does it take to roll out across the whole company?

One role at a time, 2-4 weeks per role, with iteration on previously-launched curricula running in parallel. A 200-person company with 8 distinct roles can have a complete curriculum stack in 4-6 months. The cadence is what makes it work - big-bang rebuilds at 9-12 months typically ship outdated content.

What if my senior teammates are not great on camera?

Most are not. The Loom-quality bar is intentional: it removes the on-camera-performance anxiety that studio production introduces. A senior teammate explaining a decision they make every week records well even if they would be terrible at a TED talk. Light edit in Descript cuts filler words and fixes pacing. The result reads as authentic, which is exactly what new hires need to believe.