The Cost of the Run — ATM Cash & Service Routing

Is an extra stop cheaper
than the emergency run it prevents?

A cash run priced on the bank's own tariff, in four named terms, on a calendar that walks a business week at a time — beside the one cost boundary in this market that is drawn on a map and has a published price.

California reference implementation · 5,377 published cash points · 617 assertions green On-prem · no keyed provider · no route, no reorder, nothing called optimal

© 2026 Tabaqat · Built on Strata — sovereign geospatial applications. The office estate and the Federal Reserve rate card are published and probed; cash levels, run programme, carriers and the whole tariff are GENERATED (seed 20260818) and illustrative — not any institution's.

The decision this drives

Change the replenishment policy, or reopen the carrier contract.

Not is today's run right — that is a different decision, taken by a different person against a different clock. This app is read at policy review and at contract renewal: monthly, quarterly, or the week before the cash-in-transit contract is re-let. Its output is an argument about money, not a morning's dispatch pack.

Which is why the manifest is a stop list in manifest order, never a league table, why nothing on screen is labelled optimal, and why the app never reorders a stop. It computes no route and uses no road network — a public router would transmit the customer's stop sequence, which is to say where cash will be and when, to a third party.

Buyer: head of ATM operations Reads over the shoulder: the CIT contract owner · the cash planner · whoever defends the line in next year's budget

3.25×the published spread on one line item — armoured courier and cash management at $2,667 to $8,667 per ATM per year in a fleet-level ten-year breakdown

Nobody publishes where inside that band a given estate actually sits, because the answer is a function of one bank's contract and one bank's run programme.

The same source names what is not in the budget: "Everything else is either spread across other budgets, written off as one-off events, or never measured at all."

The finding that shaped the app

The cost the estate names as its biggest problem is the one it has handed to the counterparty to score.

74%named transportation and cash-handling costs the biggest challenge — the top answer, ahead of ATM out of cash at 63%

62%have outsourced ATM cash management outright — the top response to the outsourcing question

48%take the KPIs from the vendor: "Vendor creates KPIs and/or reports periodically"

19%"I don't know / I don't have any reporting" — and 14% of those who outsourced use no KPI at all

Asked which cash-management KPIs they use, 47% answered out of cash and only 32% answered costs. The base the industry's savings percentages are quoted against is, by the industry's own survey, unmeasured.

The source: ATM Total Cost of Ownership Guide, Networld Media Group — over 325 financial institutions, IADs and ATM providers across the US, Europe, Latin America, Africa and Asia Pacific. Dated 2017 and quoted as the industry's own measurement of its own practice, not as current market share.

The market, page-verified 2026-08-30

Everyone optimises the order. Nobody prices the plan.

Product	What it publishes, verbatim	What it does not do	Prices a plan not yet ordered?
Sesami — SES Cash Cycle Optimizer	"25% Proven Reduction Of CIT Costs" · "25% Reduction Of Typical Inventory Holding Costs"	no tariff, no cost per stop, no cost per delivered unit, no map	No
Morphis — Reverse Vendor Invoice™	"constantly monitoring and comparing vendor activity to the underlying contract"	checks the bill that arrived; no per-stop or per-delivered-unit price, no map	No — it looks backward
Diebold Nixdorf · Auriga · NCR Atleos · CMS	"a minimum savings of 15%" · "10 to 25%" · "over 200 user-adjustable variables" · "25-30%"	a percentage off a total, on the vendor's own claim; no rate card, no map	No
Cash Management.iQ	"view cash points on a map with status updates"	the only map in the set — and it is a status map, not a cost map	No
NamSys Cirreon · G+D · MorphisACM	GPS trucks, stop reordering from the depot, carrier margin	the other side of the invoice — the carrier's product	No
This app	a currency figure per delivered unit with the tariff terms visible behind it	does not forecast cash, does not reconcile an invoice, does not dispatch	Yes — that is the whole product

A percentage cannot be checked; a rate card can. And the one published instance of the method working credits 14% off CIT cost to a bank's own analysis of trips-per-ATM against stock balance — no vendor product named.

Why the card has four terms and never one number

A one-sided cost-per-run figure recommends in one direction.

serve

The tariff cost of the run: per stop, per bag, per hour of crew time at the stop, per transaction, basis points of value carried — with the minimum charge, the contract discount and the surcharge applied in that order.

carry

The cost of the cash the run leaves in the machine, for the days it sits there, at a rate the reader states and can change. This is the term whose absence recommends always fewer, larger runs.

emergency

The out-of-scope premium as its own billed line, because that is how it appears on a real invoice — separately itemised beside the routine armoured-carrier fee, never blended into it.

cross-shipping

A generated bundle volume at a published Federal Reserve rate. The two are never summed into one unlabelled figure, and the line says on itself which half is which.

The carrying rate is a named, dated, on-screen, editable input — 3.625 % / yr, captioned FOMC statement, 17 June 2026 — target range 3-1/2 to 3-3/4 percent. A nonsense value is refused rather than silently applied.

The habit it corrects is documented: a practitioner in the 2017 survey, "The cost of cash is close to 0 percent, so the fewer replenishments, the better." The habit outlived the rates. The app never hides the input that flips its own recommendation.

The application, first paint

A time player. Cover the bottom bar and the app stops working.

ATM Cash & Service Routing at first paint: the manifest panel on the left, the full-bleed San Diego sub-zone map with numbered stops and straight connectors, the four-term cost card on the right, and the replenishment calendar across the bottom.

Shipped build, 2026-08-30 — a browser screenshot taken by the automated driver, not a mock-up.

The calendar is the mechanism, not decoration. Cross-shipping is measured over a business week, billed over a quarter, against an allowance that expires unused. The money in this problem is defined in time.
Press ▸ and everything moves at once. Cash points repaint to their days-of-cash state, the manifest repopulates, the exemption gauge accrues toward 875 and snaps back at the boundary, and both sides of the cost card reprice.
Two entry points, one record. A manifest row or a map cash point adopts the same site — fly, flash ring, popup, cash curve, detail. A second click releases.
Move a stop between the plans and both sides reprice, with the sign of the delta flipping visibly.
The distance basis is a field, not a footnote. DIST_BASIS reaches the popup, the legend and every export row.

44 behaviours specified and tested

The number that is new

Cost per delivered unit — and the run it avoids.

Cents per $100 of cash delivered, which is deliberately the same axis as basis points, so it sits directly beside the published cross-shipping figures of 4.0 bp on $20s and 8.0 bp on $10s. Both plans, four terms each, and the signed delta — on one card, from one tariff, at one moment.

week of 14 Sep 2026 · proposed 24 stops vs committed 16
proposed total $14,718 · committed total $13,286
the proposed plan costs +$1,432 more in absolute dollars
…and delivers $1.122 m against $775 k, avoiding one emergency run
= 131.18¢ against 171.43¢ per $100 delivered — 40.25¢ cheaper

The total and the unit price disagree, and that is the entire point. A planner reading the total buys the wrong policy. Every term is derived from the tariff table, so changing one rate moves every number on screen in one frame.

Cost — proposed vs committedgenerated tariff

proposedcommitted servetariff$4,652$3,020 carrycash held @ 3.625 % / yr$701$656 emergencyout-of-scope premium, billed separately$285$570 cross-shipgenerated volume at a PUBLISHED rate$9,080$9,040

cost / delivered unit131.18¢ per $100 delivered — committed 171.43¢−40.25¢ Δ total vs committed+$1,432 24 stops vs 16 · 1 emergency run vs 2 · straight-line total 768.8 km

The card as shipped, on the week of 14 Sep 2026. The two readings are printed side by side and never conflated — a saving quoted only as a total, and a saving quoted only per unit, recommend opposite policies.

Why this is a map at all

A cost boundary with a published price and a quarterly clock.

The Federal Reserve's currency recirculation policy charges $8.00 per bundle above a de minimis of 875 bundles per quarter, per Federal Reserve zone or sub-zone, in the $10 and $20 denominations only, when an institution deposits fit currency and orders the same denomination within the same Monday–Friday business week. "Any unused exemption expires at the end of the quarter and is not transferable."

Twelve sub-zones are designated, and one of them splits California: San Diego carries its own 875-bundle allowance, monitored separately from the rest of the state. It is public, priced, per-institution, and calendrical and geographic at once — and not one product in the market study surfaces it.

29 Jun · $028 Sep · $16,2005 Oct · $0 — the quarter turns

427published cash points inside the San Diego sub-zone — a 1:1 correspondence with San Diego County, zero leakage in either direction, across 77 ZIPs

76,125bundles of exemption held inside California each quarter by the 29 institutions operating in all three cash zones — three separate 875-bundle allowances each, expiring unused

42institutions hold at least one sub-zone office; 36 of them also sit in the Los Angeles zone, and therefore hold two separate allowances

The refusals, and each one is measured

Where the denominator does not exist, the app prints a dash.

The app scoped to zone unstated: every cost cell reads $0, the cost per delivered unit reads an em dash, and the exemption gauge is greyed out because that denominator is not published.

Scope the estate to zone unstated and the exemption gauge greys out, the manifest empties, and the reading refuses to compute. A cross-shipping bill is a real invoice; a guessed zone is a confident wrong number.

The LA / SF split is not published — 4,950 of 5,377 offices. A nearest-office proxy exists and looks respectable (2,737 / 2,213, only 64 inside a 25-mile ambiguity band). The app does not use it.
There is no ATM estate, deliberately. No public US register exists — FDIC returns {"total":0} for any ATM service type, OpenStreetMap carries a machine reference on 4 of 1,896 California nodes. The lane renders empty with its citation.
Straight-line is quantified, not disclaimed. A real 25-leg run measured ×1.257 against driven distance, per leg 1.10–1.71. A per-km term priced on straight line under-bills by 20.4% — which is why the tariff prices stops, bags, hours, transactions and value instead.
Refusing to reorder is printed as a cost. A solver re-sequenced six stops for 42.8% less distance. The app prints that as forgone, so the refusal is a stated trade rather than a hidden limitation.

Proof, not promises

Built, driven and measured — 2026-08-30.

617assertions green — 297 live · 186 offline · 134 in real headless Chrome

44behaviours specified, each mapped to the suite that exercises it

4.72worst measured contrast ratio anywhere, fill against halo — every informational token clears 4.5 in both themes

5,377published cash points on the map, derived from the frozen register — no count is hardcoded anywhere

A screenshot caught what 435 assertions could not

CARTO's raster CDN now composites "API KEY REQUIRED" across every tile while answering HTTP 200 with real map content. A guard asserting the basemap was keyless was green while the shipped map was branded end to end. The set moved to OpenFreeMap / Versatiles / OSM / OpenTopoMap — and the finding is a core one, inherited by every app built on this template.

The numbers found their own bug

The cross-shipping term came out $0.00 in all 18 weeks because the generated bundle volume was a flat per-institution draw. The fee is assessed on an institution's whole activity in the zone, so the volume now scales with its measured office count — and the reset became something you watch rather than something a caption claims.

The emergency term priced the state, not the run

It counted every run-out in the whole 5,377-office estate against a 24-stop plan — $203,490 against $6,039 of serve cost, drowning the other three terms by two orders of magnitude. It is now scoped to the week's candidate pool: the sites the planner is actually deciding over.

Reading the figures: the office estate (FDIC BankFind, index locations_20260828090007), the San Diego sub-zone polygon and the Federal Reserve rate card are measured and published. Cash levels, the 48-site contract estate, run manifests, carriers, the whole tariff and cross-shipping volumes are GENERATED (seed 20260818), stamped on every screen, in every popup and in every export — and never summed into a measured figure.

Stated as boundaries, not caveats

What this application will not do.

An analytics layer over the bank's own contract, never a system of record. Not a dispatch system, and never authoritative over the carrier's operational schedule.
No route, no road network, no reorder, nothing called optimal. A self-hosted routing engine inside the perimeter is a named optional upgrade that swaps the basis and changes the label with it.
On-prem, behind the perimeter and SSO. No keyed provider, no external call it cannot make from inside a bank's network, and every public call at build time.
No deep link and no share URL — refused, not omitted. A cash-run scope must not reach a URL, a browser history or a referer header.

Read-only. Commit plan needs a writable, authenticated ESRI backend; without one the plan is held locally and the app says so. It never appears to have written.
Refresh, not streaming, and team-level views rather than per-user row entitlements.
It prices a replenish-and-collect estate. Cash recycling — in use by 42% of the surveyed deployers — collapses two terms into one and is a named future term, not a silent assumption.
English only, LTR throughout, and the tariff is the most invented object here: its structure is the published shape of a CIT contract; its numbers are ours.

Every one of these is written into the delivered application's own README, and the market mismatch with it: this is a demonstrable build in a market whose data can be verified — California — not a delivery to a named account. The Fed cross-shipping rule has no Gulf analogue. What ports is the shape; what does not port is every number.

Point it at your contract

One tariff.
One run programme.

Your carrier rate card with its minimum charge, discounts and surcharges; your site list with cash levels and cadence; your emergency-callout history; and a carrying rate your treasury will stand behind. Two weeks, and this card prices your contract instead of a generated one.

Swappable by configuration, not by code: the estate, the tariff terms, the currency, and the zone geography wherever a central bank draws one. The four cost terms are a typology the app renders, not a list of US numbers it ships.

tabaqat.net → Solutions → Financial Services info@tabaqat.net

© 2026 Tabaqat · Built on Strata. Reference implementation over the California office estate and the published San Diego cash sub-zone; cash, runs, carriers and the tariff GENERATED (seed 20260818). Analytics only — not a system of record, not a dispatch system, and not the carrier's schedule.

1 / 12