Board-level detail · financial appendix

The upgrade as a capital project: NPV, IRR, and where the case breaks.

This is the depth behind the one-line version on the Case page. It treats the ~$1.4 million all-in Tarana wireless upgrade the way a CFO or commissioner would — a discounted-cash-flow (DCF) analysis with NPV, IRR, a 10-year pro-forma, break-even, and sensitivity. Every operating input is drawn from the District's audited 2024 financials (WA State Auditor) and its own Tarana cost workbook — not our estimates. The capex is not the ~$1.2M vendor sticker: it's the all-in funded figure — sticker plus sales/use tax, tower labor, and contingency (the risk model's funded P80; see how we modeled risk) — because the sticker omits all three. The model is open and reproducible (sources/analysis/financial_appendix.py); the figures below are read straight from its output. Where a choice could flatter the result, we made the conservative one and say so.

$255K

NPV at a 4% discount rate (the District's own ~3% cost of capital, rounded up)

7.5%

IRR — roughly 2.5× the District's cost of capital, even on the all-in capex

8.3 yr

Discounted payback (7 yr undiscounted), inside the gear's ~10-yr life

1.18×

Profitability index — $1.18 of value per $1 invested

Base case: all-in capex of $1,383,192 (funded P80: hardware sticker + 8% sales/use tax + tower labor + Class-4 contingency), the 878 customers the upgrade actually serves, $40/mo proposed wholesale rate, 45% cash (EBITDA-style) margin, 10-year gear life. PRIMARY INPUTS

Read these two assumptions before the numbers above — they're what the headline rests on, and an opposing analyst will name them first.

The rate is the proposed $40/mo, not the ~$31 collected today. $40 is the District's proposed single-tier rate (~28% above the current blended ~$31). At today's $31 the same model gives –$105K NPV, 2.5% IRR, 8.8-yr payback — on the all-in capex basis, today's rate no longer clears the 4% hurdle on its own. The proposed $40 rate is load-bearing, and we say so rather than let you find it.
The headline treats the served revenue as fully at-risk without the upgrade. The +$255K NPV equals the project's cash vs. zero — i.e. it assumes the congested, end-of-life line would otherwise be lost. That's the District's own “severely congested” premise, but it's a judgment: if the line would instead just decline, the decision NPV is –$725K at a −10%/yr fade, and it only clears its cost of capital if the un-upgraded line would shed more than ~41%/yr (section 04).

Board summaryThe metrics a utility board reads first.

The same case, translated into the vocabulary a commissioner or a bond analyst actually uses — debt-service coverage, break-even take rate, and rate impact. Every figure is read straight from the same model (financial_appendix.py).

6.5×

Debt-service coverage — the telecom segment's audited ~$808K/yr surplus is 6.5× the new debt service. For context, peers report system DSCRs of 5.5–6.7× (Pend Oreille, Mason 3). On the upgrade's own incremental cash alone, worst-year coverage runs ~0.4× in the deliberately pessimistic scenario to ~1.2–1.4× in the balanced and optimistic ones (see the three scenarios) — the segment-level coverage here is the covenant a bond analyst actually tests.

$124K/yr

Debt service if bonded at 4% over 15 yr — a conservative rate; the District's actual 2020 bond cost was ~2.9%, so real service runs lower

741 of 878

Break-even take rate — customers needed for NPV = 0. The District already serves 878, so it clears break-even today — though the cushion is thinner on the all-in capex basis (section 05)

The two questions a board asks first:

“Does this hit electric ratepayers?” — No, $0. The telecom segment is ring-fenced by statute (RCW 54.16.330: separate accounting + revenue dedicated to the telecom line) and runs its own ~$808K/yr surplus. The upgrade is funded from telecom's own cash or bonded against telecom revenue — never electric rates.

“What if we win a grant?” None is committed to this refresh — but the precedent is strong (the 2010 build was ~60% grant-funded, and WA BEAD's single largest award, $332M, went to a fixed-wireless provider). A grant covering 25% of the capex shortens simple payback to ~5.3 yr; 50%, to ~3.6 yr. Disclosed as upside, not banked into the base.

01 — How to read thisThe honest version, including where it's weak.

A capital project is worth doing when the cash it returns, discounted to today, exceeds what it costs — a positive NPV — and when its IRR beats the cost of the money. On the base case it clears both comfortably. But DCF is only as good as its counterfactual, so we put the weak point up front in section 04, and in section 06 we show what this commercial lens leaves out entirely.

Conservative choices we made (so a skeptic doesn't have to find them):

Capex charged all-in at the funded P80 — the vendor's ~$1.2M hardware sticker plus 8% sales/use tax, tower labor, and a Class-4 contingency ($1,383,192 total) — rather than the bare sticker, which omits all three.
Discount rate set at 4%, above the District's actual ~2.9% bond cost — a higher hurdle.
Scoped to only the 878 customers the upgrade serves today — not the whole wireless line.
Cash margin centered at 45%, with no credit for the growth headroom the budget already funds.
Once set, the rate grows only 1.5%/yr (well below commercial escalation), the gear is worth $0 at year 10 (no salvage), and no subscriber growth is assumed.

(The base rate level — the proposed $40 — is the one un-conservative input, which is why it's flagged in the box above and shown at today's $31.)

02 — Base caseNPV by discount rate.

NPV stays positive at every hurdle a public utility would defensibly apply (3–6%); it reaches zero at the IRR of 7.5%.

Discount rate	NPV	Read
3%	$341,955	below the District's bond cost
4%	$255,217	base case (conservative)
5%	$174,786	a stiff hurdle
6%	$100,108	a very stiff hurdle
7.5% (IRR)	$0	the return that exactly breaks even

03 — SensitivityWhat actually moves the answer.

One-at-a-time, each input swung across its plausible range (margin and customer count from the model's documented bounds; capex across the Class-4 span the risk model documents; the rest from reasonable spans). Base NPV is $255K. Charging the capex all-in buys a thinner cushion than the old sticker-price view: the pessimistic end of every lever except the discount rate now dips below zero — the honest cost of counting tax, labor, and contingency up front rather than discovering them later. The base case stays positive throughout.

Bars shaded red cross below $0 at their pessimistic end — on the all-in capex basis that's every lever except the discount rate. Ranges: cash margin 28–58% and customers 685–1,035 are the cost-recovery model's documented bounds; ARPU $31–$45/mo, gear life 8–12 yr, discount 3–6%, capex −10%/+40% (the AACE Class-4 accuracy span the risk model documents).

04 — The weak point, stated plainlyIt all hinges on the counterfactual.

The base case treats the served-customer cash stream as at risk without the upgrade — justified by the District's own memo that the network is “severely congested with no possibility for expansion” and by the age of the 2010-era gear. A fair skeptic asks: what if the line would largely survive anyway? So rather than assert a decline rate, we solve for it.

Break-even counterfactual. The upgrade clears its 4% cost of capital as long as, without it, the line would lose more than ~41% of its subscriber base a year (about a 40%/yr cash-flow fade) — i.e. the un-refreshed line would have to collapse, not merely fade, over just a few years. For capacity-capped, end-of-life fixed-wireless gear that the District itself calls maxed out, a collapse is a plausible path — but it is a steep hurdle and a judgment, and moving to the all-in capex basis made it steeper. We flag it as the load-bearing assumption rather than burying it. HONEST CAVEAT

05 — Break-even margins of safetyHow wrong can the inputs be?

Holding the discount rate at 4%, each input can move a long way toward the pessimistic before NPV reaches zero:

38%

cash margin break-even
(base 45% — a 7-pt cushion)

$34

ARPU break-even, /mo
(base $40 — the rate could fall ~16%)

741

customer break-even
(base 878 — could lose ~16% and still pay off)

06 — What this DCF leaves outWhy the commercial number understates the case.

The analysis above is deliberately narrow: it counts only the wireless segment's own cash. For a public utility that omits most of the value:

Rate protection for ratepayers. The network's main job is to hold down what ~13,400 households pay for internet — a benefit that lands in residents' pockets, not the segment's revenue line, so DCF never sees it.
Option value & irreversibility. Keep the asset and the District can still adjust later; let it go and a rebuild runs on the order of ~$40,000 per home. Standard infrastructure doctrine (World Bank / RAND) favors the reversible option under uncertainty — value a pure DCF doesn't price.
“Do nothing” isn't a flat baseline. The 2010 plant is obsolescing; the realistic no-upgrade path is decline and eventual failure, not a stable annuity.
Growth headroom already paid for. The upgrade budget funds radios for ~1,035 customers — about 18% above today's 878 — so new sign-ups add revenue against an already-committed cost.

So treat section 02 as the narrow commercial case, not a floor: whether that number itself is positive depends on the counterfactual (section 04), and it omits the public-benefit value above — which doesn't. The point isn't that the segment IRR is large; it's that a public utility shouldn't decide a 25-year ratepayer asset on segment cash alone — and the narrow number clears its hurdle only if the no-upgrade path is a genuine collapse, which is exactly why we don't rest the case on it.

07 — Worst case & reproducibility

Stack every pessimistic assumption at once — NCI exits (685 customers), a 28% margin, the legacy $31 ARPU, an 8-year gear life, a 6% discount rate, and a 40% capex overrun on top of the already-loaded all-in base — and the project goes negative at –$1,472K. That is the honest downside, and it requires essentially everything to break the wrong way simultaneously. The bounded loss on a ~$1.4M one-time spend is itself the point: the cost of being wrong here is recoverable; the cost of letting the asset go is not.

Reproduce it. The DCF model is sources/analysis/financial_appendix.py (built on the cost-recovery recovery_model.py); both read the District's Tarana workbook and the SAO 2024 audit. Run it and you get the JSON this page renders. Inputs, code, and sources are all on the Sources page.

Primary inputs: WA State Auditor's Office, Okanogan PUD 2024 financial statements (telecom segment revenue, operating income, depreciation, plant); Okanogan PUD “Tarana Upgrade costs” workbook (per-site hardware, $787/radio, 878 customers, 1,035-radio plan); proposed $40/250 Mbps single-tier wholesale rate (District Tarana presentation). Cost of capital: District 2020 revenue-bond true-interest-cost ~2.9%. Capex basis: the risk model's base-configuration funded P80 ($1,383,192 = hardware sticker + 8% WA sales/use tax + tower labor + AACE Class-4 contingency; see how we modeled risk and the model documentation). DCF, break-even, and sensitivity computed in financial_appendix.py. Figures are management-style estimates for decision support, not audited projections.

← Back to the plain-language Case