Introduction #

When this guide fits: You manage a large industrial account where the utility bill mixes kWh energy, interval demand (kW or kVA), reactive charges, and sometimes time-of-use blocks—and you need to rank levers by dollars, not by vendor slogans.

When it is not suitable: You are structuring a utility-scale PPA, behind-the-meter storage arbitrage, or carbon offset procurement—those need financial and interconnection models outside this operations audit.

Verified workflow: 2026-05-25 — interval data and tariff lines from a metal fabrication campus (Great Lakes tariff, anonymized). Replace rates with your PDF; keep the sequence: bill decomposition → peak forensics → no-Capex controls → Capex only with meter proof.

Annual cost stack before optimization: demand dominatesEnergyDemandReactiveFixedDemand ≈ 71% ofvariable stack ($)

Case snapshot — bill anatomy (FY2025 baseline) #

Tariff line (monthly avg) Quantity Rate (example) $/month
Energy (kWh) 1,980,000 kWh $0.0468 / kWh $92,664
Demand (kW, 15-min max) 18,200 kW $12.10 / kW $220,220
Reactive (kVARh excess) 3,600 kVARh $0.82 / kVARh $2,952
PF penalty (displacement) tariff rider $3,100
Fixed customer charge 1 $4,800 $4,800
Variable subtotal $318,936

Insight: Demand + reactive + PF71% of variable spend. A lighting-only project cannot fix a Monday 07:12 melt-furnace + chiller coincidence.

Use the Industrial Energy Estimator to re-run kWh and demand $ sensitivity after you substitute your quantities and rates.

Step 1 — Forensics on the five peaks that set demand #

Utility interval export (March 2026) — top 15-minute windows:

Rank Timestamp (local) Billed kW Dominant loads (operator log)
1 Mon 07:12 18,200 Two arc furnaces + central chiller
2 Wed 14:38 17,650 Furnace + trim press bank
3 Tue 06:55 17,100 Morning preheat + dust collection
4 Thu 11:20 16,880 Heat treat + compressed air recovery
5 Fri 15:05 16,420 Afternoon batch + HVAC economizer failed open

Action: Assign each row to an owner (melt shop, utilities, HVAC). No Capex until logs explain why loads coincide.

Submeter anchors we installed for the audit:

Feeder Share of peak kW (est.) Meter ID
Melt shop incomer 48% MS-01
Chiller plant 22% CH-PLT
Compressed air 14% CA-HDR
Trim / finishing 11% TF-02

Cross-check total plant kW with the Factory Load Calculator when adding or retiring lines.

Step 2 — No-Capex wins executed (Q1 2026) #

Lever Change Metered effect
Start sequencing Delay second furnace tap by 4 min after chiller start Peak #1 −1,050 kW
Chiller supply reset +2 °F supply when load < 70% Peak #3 −280 kW
Economizer repair Stuck OA damper on AHU-4 Peak #5 −190 kW
Night compressed air Shut tertiary booster 22:00–05:00 kWh −38,000/mo
PF bank tuning See companion case PF penalty −$3,100/mo

Companion reactive work: Factory Power Factor Fix (468 kW case).

Combined demand reduction (same production volume): 18,200 → 16,680 kW (−1,520 kW).

Monthly demand savings = 1,520 kW × $12.10/kW = $18,392
Annualized ≈ $220,704 (before tax/fees)

Step 3 — Bill reconciliation after controls (April 2026) #

Line Before After Δ $/month
Energy kWh $92,664 $89,120 −$3,544
Demand kW $220,220 $201,828 −$18,392
Reactive + PF $6,052 $2,940 −$3,112
Variable subtotal $318,936 $293,888 −$25,048

Simple annualized variable savings: ≈ $300,600. Internal labor + controls work ≈ $158,000~6.3 month payback on OpEx/Capex blend (excludes furnace projects still in business case queue).

Model your tariff mix in the Energy Estimator before approving storage or solar Capex.

Step 4 — What we deferred (and why) #

Proposal Vendor claim Audit decision
2 MWh battery peak shave −2 MW peaks Hold — sequencing not exhausted; duty cycle unclear
LED relamp only −15% bill Partial — kWh yes; demand unchanged
New 1,200 HP compressor Faster recovery Reject — leaks + schedule fix cut header kW 11%
Random PF caps −10% bill Reject — harmonic study required (common mistakes)

HVAC and motors — ranked by $/month at this site #

System $ lever type First engineering move
Melt + dust collection Demand Sequence taps; do not add capacity
Central chillers Demand + kWh Reset + staging; see HVAC Load vs Capacity
Trim motors kWh VFD on variable torque; check harmonics
Lighting kWh Occupancy verified; not primary demand fix

Browse the Power calculator hub for kW ↔ kVA ↔ amps when the tariff mixes units.

Measurement discipline (so finance trusts engineering) #

  1. Align site meter definitions with the utility settlement CT—arguing two PF values wastes months.
  2. Store 15-minute CSV exports for 24 months; tariff riders change.
  3. One KPI per submeter on the shift board (example: kW per ton poured).
  4. Require two weeks before/after interval files for any vendor savings claim ≥ 8%.

Deeper audit method: How to Conduct an Industrial Energy Audit. Tariff math primer: How to Calculate Industrial Energy Costs.

Next steps you should take #

  1. Rebuild the bill table with your tariff PDF—rank lines by annual dollars.
  2. Export top 10 15-minute peaks; join operator logs within ±5 minutes.
  3. Implement one sequencing or setpoint fix; re-export interval data after 30 days.
  4. Only then open Capex tickets (battery, chiller, major VFD) with meter proof.
  5. Request indexing in Search Console after measurable updates (this page 2026-05-25).
Which tariff line should we optimize first?

Sort every variable line by annual dollars using your PDF. On many large industrials, interval demand wins—attack simultaneity before relamping alone.

Why did demand stay high after an LED retrofit?

LEDs cut kWh, not necessarily the 15-minute kW peak when furnaces, chillers, or compressors still align. Plot peaks, do not average monthly kWh only.

Will VFDs always lower the bill?

Strong on variable-torque loads; weak if harmonics add filters/reactors without planning. Submeter the feeder before/after.

Should we buy battery peak shaving before fixing starts?

Usually no—sequencing and controls are cheaper and define the real peak duty cycle batteries must serve.

How do kW and kVA demand differ on the bill?

Some tariffs bill kVA intervals. Use the kW to kVA calculator and the PF case study if reactive power inflates apparent demand.