TI Playbook/Field Library/Pipeline Health Check

Is the pipeline actually healthy?

"We have 200 candidates in the pipeline" tells leadership nothing. This is the quarterly health check that tells them whether they can fill what's open - from sources that won't dry up, without candidates stalling out.

I'm selling mode Guide 3.2 Dashboard format Quarterly review

The quarterly pipeline review

Once a quarter, leadership wants to know if the talent pipeline is in good shape. The temptation is to answer with volume - candidate counts, reqs filled. But volume hides the things that actually break a pipeline.

This health check replaces "how many candidates" with five signals that predict whether you'll hit hiring plan next quarter. The output is a dashboard leadership can read in two minutes and a verdict they can act on.

Health is coverage, diversity, and flow - not volume

A pipeline of 200 candidates can be unhealthy and a pipeline of 60 can be in great shape. What separates them is whether the candidates are spread across your open roles, sourced from channels that won't collapse, and moving through stages instead of aging out.

The five signals below are what a strong health check measures. Each maps to a panel on the dashboard.

Signal 1

Coverage vs open headcount

Qualified candidates per open req, measured against a role-specific target. Thin pools like chiplet need a different bar than verification.

Breaks when: a critical role sits below 2 per req
Signal 2

Depth by role

Coverage broken out per role, not averaged. A healthy blended number can hide one role that's starving.

Breaks when: the average looks fine but one role is red
Signal 3

Source concentration

How much of the pipeline comes from a single source. Over-reliance on one competitor or channel is fragility, not strength.

Breaks when: one source exceeds ~50% of inflow
Signal 4

Time-in-stage / aging

How much of the pipeline has gone untouched too long. Aging candidates are effectively gone, even if they're still in the system.

Breaks when: 40%+ untouched in 90+ days
Signal 5

Stage conversion / flow

Where candidates drop off. A single stage with a conversion cliff is where your pipeline is actually leaking.

Breaks when: one transition falls below ~30%
Together

One red signal is a warning

Two or more red signals at once is a pipeline that will miss plan. The dashboard exists to catch the combination early.

What the health check looks like

Toggle between a healthy quarter and an at-risk quarter to see what each signal looks like in good and bad shape. The data is a worked example for a semiconductor engineering org.

Q3 Pipeline Health · Eng Hiring · sample data

At-risk quarter

Overall pipeline health
At risk — 3 of 5 signals red. Will miss plan without intervention.
Coverage by role — candidates per open req (▏ = target)
Source mix — where the pipeline comes from
Stage flow — where candidates drop off

What each panel is telling you

Coverage tiles

Read the reds first

A green blended coverage number means nothing if chiplet sits at 1.2 per req. Always scan the per-role bars before quoting the average.

Source mix

Concentration is hidden risk

If 60% of the pipeline comes from one competitor and they freeze hiring or get acquired, your inflow collapses overnight. Diversify before it's forced.

Aging

Stale equals gone

A candidate untouched for 90 days has almost certainly moved on or cooled off. Counting them as pipeline inflates the picture and delays the real problem.

Stage flow

Find the cliff, fix the cliff

One transition with a conversion drop is where the leak is. An onsite-to-offer cliff usually points at calibration or comp, not sourcing.

Six steps to a quarterly dashboard

Step 1

Set role-specific coverage targets

Don't use one bar for every role. Thin pools (chiplet, RF/analog) might target 3 per req; deeper pools (verification) might target 4-5. Set the target from the supply snapshot, not a gut number.

Do this: Anchor each target to the role's pool size from Guide 01. Document why each target is what it is.
Step 2

Pull coverage per role, never blended only

Compute candidates-per-req for each role separately. Keep the blended number for the headline, but the per-role breakdown is what surfaces the starving role.

Do this: Color each role green / amber / rust against its own target, not the average.
Step 3

Map source concentration

Tag every pipeline candidate by source and compute the top source's share. Flag anything over ~50% from one channel or competitor as a concentration risk, even if volume looks healthy.

Do this: Name the dominant source explicitly. "61% from one competitor" lands harder than a generic bar.
Step 4

Measure aging, then discount it

Calculate the share of pipeline untouched in 90+ days. Then re-run coverage excluding aged candidates - the "real" coverage is almost always lower than the headline.

Do this: Show both numbers. The gap between raw and active coverage is the credibility of the review.
Step 5

Trace stage-to-stage conversion

Compute conversion at every transition. Flag the single biggest drop. That's your bottleneck - and it usually has a specific, fixable cause.

Do this: Pair the bottleneck with a hypothesis. Onsite-to-offer cliff → likely comp or calibration. Screen-to-HM cliff → likely a misaligned bar.
Step 6

Write the verdict and the one move

End with a single health verdict and the one intervention that matters most this quarter. Leadership doesn't want five action items - they want to know the one thing to fix.

Do this: "Pipeline at risk; the one move is diversifying chiplet sourcing off our single competitor source." One verdict, one move.
On continuous measurement
The decision cadence has shifted from annual to continuous. A pipeline reviewed once a year is a pipeline you find out is broken a year too late.

The quarterly review one-pager

Copy this scaffold for the written summary that accompanies the dashboard.

Quarterly pipeline review
PIPELINE HEALTH CHECK — [QUARTER] — ENG HIRING
Prepared by: [name] · [date]

VERDICT
[Healthy / At watch / At risk] — [X] of 5 signals red.
The one move this quarter: [single most important intervention].

THE FIVE SIGNALS
1. Coverage (blended):   [X] per req   — [G/A/R]
2. Depth by role:        [worst role] at [X]/req vs [target] — [G/A/R]
3. Source concentration: [X]% from [dominant source] — [G/A/R]
4. Aging:                [X]% untouched 90+ days — [G/A/R]
   Active coverage (ex-aged): [X] per req
5. Stage flow:           bottleneck at [stage]→[stage] ([X]%) — [G/A/R]

WHAT'S WORKING
- [healthy signal + why]

WHAT'S AT RISK
- [red signal + the cause + the fix]

THE ONE MOVE
[The single intervention, who owns it, by when.]

NEXT REVIEW: [date]

The marks of a health check leadership acts on

Where AI helps, and where it can't

Let AI compute the metrics and flag the reds. Coverage ratios, source concentration, aging percentages, stage conversion - this is exactly the structured math AI is reliable at.
Let AI draft the bottleneck hypothesis. Ask it what typically causes a drop at a given stage - it'll surface the usual suspects to investigate.
Don't let AI set the coverage targets. The right target per role comes from the actual pool size and your hiring plan, not a model's generic benchmark. A wrong target makes every status light wrong.
Don't trust AI to know if a candidate is really "active." Aging logic depends on your ATS hygiene. If last-touch dates are unreliable, the aging signal is garbage in, garbage out - a human has to vouch for the data quality.
Don't let AI pick the one move. The intervention is a judgment call about what's tractable this quarter given your team and constraints. AI can list options; the prioritization is yours.