Skip to content

Availability Funnel Dashboard

Monitoring dashboard — instantiates Effective-Input Delivery Assurance

Shows supply narrowing stage by stage into the fraction actually usable at the point of action, and tracks the response and off-target signals it produces over time.

A live monitoring surface that renders the delivery path as a narrowing funnel — supplied at the top, usable-at-the-point-of-action at the bottom — and keeps the drop between them, and the response it produces, in continuous view. Its defining move is that it is the watch, not the measurement or the accounting: it takes whatever the assays and audits produce and aggregates it into one continuously-updated effective-availability ratio and its trend, so a widening gap or a rising off-target signal is caught as it moves rather than at the next quarterly review. It does not generate the per-stage numbers and it does not explain the loss — it makes the shrinkage and its consequences legible at a glance and over time, to a human who can act.

Example

A municipal water utility pumps roughly 100 ML/day of treated water into its network. Nominal reporting stops there — "we supplied 100." The Availability Funnel Dashboard renders the rest of the path as a taper: pressure-zone losses, main breaks, background leakage, unbilled authorized use, and apparent losses (meter under-registration, unauthorized connections), down to the metered consumption that actually reaches taps — about 78 ML/day, an effective-availability ratio of ≈0.78 system-wide. The funnel is drawn per district-metered area, and one old district reads only ≈0.55. A response lane tracks customer-reported low pressure; a side-effect lane lights up on abnormal night-flow (the signature of a hidden leak).

A control-room operator watches district DMA-12's ratio slide from ≈0.62 to ≈0.49 over a week and dispatches leak detection before it becomes a burst main. The dashboard did not measure the leak or price its cost — it made the slide visible, segmented and in time to act, while a single system-wide "78 ML/day delivered" would have hidden it entirely.

How it works

  • One canvas, top-to-bottom funnel. Each stage is a band whose width is the amount surviving it; the eye reads the loss directly as the taper.
  • Foreground the ratio and its trend, not the supplied number. The headline is usable ÷ supplied over time, sliced by district, cohort, or route — because the aggregate hides the local collapse.
  • Add a response lane and an off-target lane. The monitor watches consequences — did the intended effect occur, and is the loss showing up as harm or waste somewhere — not just throughput.
  • Alarm on the derivative. A falling ratio or a rising off-target signal trips before the absolute number looks bad.

Tuning parameters

  • Refresh cadence / latency — real-time telemetry versus a daily rollup; faster catches transients but amplifies noise and cost.
  • Segmentation granularity — one system funnel versus per-district or per-cohort funnels; finer finds the localized collapse but dilutes each panel's signal.
  • Alarm basis — threshold on the level versus on the rate of change; rate-of-change catches slides early but false-alarms on noise.
  • Which lanes are shown — ratio only, or ratio plus response plus off-target; more lanes give more context and more clutter.
  • Baseline / target overlay — which "good" line to draw against; a stale target makes a real slide look normal.

When it helps, and when it misleads

Its strength is turning an invisible, gradual shrinkage into something a human watches and reacts to, and co-locating the response and off-target lanes stops the classic error of celebrating throughput while the usable fraction quietly falls.

A dashboard shows only what it is fed: an un-instrumented loss stage is an invisible taper, so the funnel can look healthy because the leak sits upstream of the first meter. Its classic misuse is that the displayed ratio becomes the target and gets gamed — operators improve the number on the screen (re-baselining, quietly excluding an inconvenient stage) rather than the water reaching taps, exactly the trap Goodhart's law names.[1] The discipline that guards against it is to periodically re-validate that the displayed stages still map to reality (via an assay or audit) and to treat the response lane, not the ratio, as the ground truth.

How it implements the components

The Availability Funnel Dashboard fills the monitoring side of the archetype — the components that surface and track, not the ones that measure or account:

  • effective_availability_ratio — its headline: usable-at-the-point-of-action ÷ supplied, trended and segmented so a falling fraction is visible early.
  • response_monitoring — a dedicated lane tracks whether the intended downstream response actually occurred, over time.
  • side_effect_signal — an off-target lane surfaces the harm or waste the loss is producing elsewhere, alongside the ratio.

It does not produce the per-stage numbers it charts (those come from Stagewise Availability Assay) or book and explain the loss (that is First-Pass Loss Audit); this mechanism displays and watches, it does not measure or account.

Notes

A dashboard is a monitor, not a controller — it says "the usable fraction is falling here," never "do X about it." It is worth pairing with a mechanism that acts on what it reveals: Route–Form–Timing Optimization or a dosing rule that turns the observed slide into a correction. Where it overlaps a Sankey Loss Map, the difference is time: the Sankey is a static snapshot of one accounting; the funnel dashboard is the same shape kept live.

References

[1] Goodhart's law — "when a measure becomes a target, it ceases to be a good measure" (after economist Charles Goodhart). A funnel dashboard is especially exposed because the displayed ratio is easy to improve cosmetically — by re-baselining or excluding an inconvenient stage — without moving the usable amount at the point of action. ↩