Time-to-hire measures how fast you got a signature. It says nothing about the thing you actually bought: working output. A team can celebrate closing a candidate in three weeks and then watch that person spend two months waiting for access, decoding an ambiguous scope, and rebuilding context nobody wrote down — while the roadmap item they were hired for sits exactly where it was. The metric that captures what you actually care about is time-to-value: the number of days from 'we have a defined need' to 'this person has shipped a first increment that matters.' Put that number at the center and every talent decision — how you hire, how you onboard, even which engagement model you choose — starts pointing the same direction.
Defining time-to-value concretely
TTV needs hard edges or it becomes another vibes metric. The clock starts when the need is written down — a brief that names the outcome, the data involved, and what 'good' looks like. Not when someone first said 'we should hire for this' in a meeting; when the brief exists. The clock stops when the person has shipped a first increment that matters: code in production, a working evaluation harness the team now uses, a model integration handling real traffic — something a stakeholder can point at. It explicitly does not stop at the start date (that's time-to-hire wearing a costume) and it doesn't wait for 'fully ramped' (unmeasurable, perpetually deferred). First shipped increment is the honest checkpoint: early enough to be a fast feedback loop, real enough that it can't be gamed with a slide deck.
- Clock starts: a written brief exists — outcome, context, definition of done for the first increment.
- Clock stops: the first increment is live — merged, deployed, in use by someone other than its author.
- Not the start date: a signed contract that's followed by three idle weeks is a hiring win and a value failure.
- Not 'fully productive': pick the first shipped thing, because it's the only ramp milestone you can't argue about.
What stretches it: the three segments
TTV decomposes into three segments, and naming them separately matters because they have different owners and different fixes. Segment one is hiring: brief to person secured. Segment two is onboarding: secured to genuinely able to work — access granted, environment running, context transferred. Segment three is scope: able to work to knowing exactly what to ship first. Most organizations measure segment one obsessively and treat two and three as unavoidable weather. They aren't. An engineer who waits eight days for repository and data access, then spends two weeks discovering that the brief's 'build the retrieval pipeline' means three different things to three stakeholders, is losing weeks that never appear in any hiring report.
| Segment | Clock | Where the days leak | Cheapest fix |
|---|---|---|---|
| Hiring | Brief → person secured | Sequential interview rounds, scheduling gaps, offer approvals | Parallel steps, pre-cleared offer bands, decision deadlines |
| Onboarding | Secured → able to work | Access requests in queues, no environment docs, context held in heads | Pre-provisioned access on day one; a one-page context doc written before arrival |
| Scope | Able to work → knows what to ship first | Vague briefs, competing stakeholder definitions, no named first deliverable | Define the first increment in the brief itself, before the search starts |
The cheap half of the problem
The uncomfortable observation for anyone who has just invested in a faster hiring process: segments two and three are usually cheaper to fix and just as large. Cutting a week from hiring takes process redesign and organizational will. Cutting a week from onboarding takes a checklist: accounts requested the day the offer is signed, a scoped dataset ready, a one-page architecture note, a named person who answers questions in hours not days. Cutting a week from scope ambiguity takes one decision: name the first deliverable in the brief, before you've met a single candidate. 'Ship an evaluation harness for the support-bot flow, running against 50 real tickets' is a first increment; 'own our LLM quality efforts' is a season of clarification meetings. Teams that do this consistently report the same pattern: the fix cost nothing and nobody could explain why it hadn't always been done.
How engagement models compare on TTV — honestly
Once TTV is the ruling metric, the engagement-model question stops being ideological. Different models genuinely differ on each segment, and the comparison deserves honesty rather than advocacy. An embedded or augmented engineer from a vetted bench compresses segment one dramatically — days to secure rather than weeks — because the vetting happened before your need existed. A permanent hire runs the full search but, done well, may carry deeper long-term context. The honest caveats cut both ways: an embedded engineer still pays the full segment-two and segment-three cost if your access and scope are a mess — the model fixes the hiring segment, not your onboarding. And a permanent hire's longer TTV amortizes if the need lasts years; a 60-day TTV against a three-year tenure is noise, while the same 60 days against a five-month project is fatal.
| Model | Hiring segment | Onboarding segment | Scope segment | Honest overall read |
|---|---|---|---|---|
| Embedded / augmented (vetted bench) | Days — vetting pre-done | Same as anyone; often faster if the partner has an onboarding playbook | Same as anyone — your brief quality decides | Fastest to first value; the fit for defined outcomes and urgent windows |
| Permanent hire (own search) | Weeks to months — the full funnel | Same leaks, plus notice periods before day one | Same — scope clarity is yours to provide either way | Slowest to first value; amortizes if the need is truly multi-year |
| Freelance marketplace | Days to post, but vetting burden is yours — real segment is longer than it looks | Often longest: least context, least commitment to your stack | Highest risk of scope drift without daily management | Fast on paper; TTV highly variable with vetting rigor you supply yourself |
Instrumenting TTV without bureaucracy
The failure mode of any good metric is the reporting apparatus that grows around it. TTV needs exactly four timestamps per role: brief written, person secured, access complete (they could genuinely start work), first increment shipped. One spreadsheet row. The three segment durations fall out by subtraction, and after a handful of hires you know precisely which segment eats your time — which is the entire point, because now the fix has an address. Resist the upgrade path: no weighted ramp scores, no productivity surveys, no TTV review meetings. Review the numbers quarterly, fix the worst segment, and remeasure. If instrumenting the metric takes longer than reading this article, the instrumentation is the new waste.
- Four timestamps: brief written, person secured, access complete, first increment shipped.
- Three durations by subtraction — each one names its own owner: hiring process, IT/onboarding, brief author.
- Quarterly review, one fix per cycle, remeasure. No standing meeting, no tooling purchase.
- Compare engagement models on your own TTV data after a few engagements, not on anyone's marketing claims — including ours.
