What are the biggest energy loads in a CPG plant?

On most consumer packaged goods lines the largest loads are refrigeration compressors for cold rooms and freezers, compressed air for the packaging hall, and steam or hot water for CIP and cook or pasteurize steps. The fillers and case packers that operators watch most closely are usually steadier and smaller than the utilities running behind the walls.

Why can a utility bill not tell me which line or SKU used the energy?

The utility meter reports one number for the whole plant over a whole month, with no attribution to a specific line, shift, or SKU. To split energy by product you have to measure at the drive and the PLC and stamp it against run state and case counts, which the bill and an hourly paper round sheet cannot do.

Do I need new sensors to start energy monitoring in CPG manufacturing?

Usually not for the major loads. Variable frequency drives, compressors, chillers, and boilers already report power, load, and firing rate to their controllers. The gap is that these signals live in separate controllers and are not joined to run data or the bill, so most of the work is reading what the machines already publish rather than adding hardware.

How does machine-level energy data reduce the demand charge?

Utilities usually bill a peak-demand ratchet on the highest fifteen-minute interval in a month, and in a plant that peak is often set by several assets starting at once after sanitation. When the historian shows exactly which startups coincided, you can stagger them by a few minutes to shave the ratchet without slowing any cases, and the saving shows up on the next bill.

How is energy turned into a cost-per-case number?

Once energy is measured by asset and by interval, and the line data provides cases, SKU, and shift on the same clock, you divide energy by cases to get cost per case. That number lets a plant compare SKUs and sequences, so scheduling can favor the runs that spend less energy when the order book allows.

Energy Monitoring for CPG Facilities

Most CPG plants meter energy at the utility bill, which is the one place it can never tell you which line, SKU, or shift actually spent it.

Where the energy actually goes on a CPG line

Energy monitoring manufacturing CPG facilities is harder than it looks, because the meter that finance sees is the utility meter at the fence and it reports one number for the whole plant. On most consumer packaged goods lines the load is not sitting where operators look. The filler and the case packer draw steady, predictable power, but the money is usually spent upstream and behind the walls: refrigeration compressors holding cold rooms and spiral freezers, the boiler making steam for cook kettles and pasteurizing, and the air compressors feeding every blow-off, reject flapper, and pneumatic clamp on the packaging line.

Compressed air is the clearest example. A packaging hall can leak twenty to thirty percent of its generated air through worn seals and open blow-offs, and the compressor answers that leak by running longer at the meter. Nobody logs it, because a hiss at a fitting does not show up on a case count. Refrigeration behaves the same way. A defrost cycle that runs too often, or a door left open on a chilled staging room during a long changeover, pulls the compressors harder for the rest of the shift and shows up only as a slightly higher bill three weeks later.

What energy monitoring manufacturing CPG plants usually miss on paper

The gap is not measurement, it is attribution. A plant can have a good SCADA screen and a diligent shift log and still not be able to answer a simple question: how much energy did we spend to make yesterday’s run of the twelve-ounce SKU versus the family-size one. The utility bill cannot answer it. The paper round sheet where an operator writes down amps or a boiler reading once an hour cannot answer it either, because it captures a snapshot, not the run, and it is copied by hand into a binder that no one queries.

That matters more in CPG than in most sectors because the demand charge and the power factor penalty often sit inside the bill quietly. Utilities usually bill a peak-demand ratchet based on the highest fifteen-minute interval in the month. In a plant where the freezer compressors, the boiler feed, and a startup surge all land in the same window, one bad Monday morning can set the demand charge for the entire billing period. When that number is a line item on a bill and not an event tied to a machine, it never gets fixed, because no one can point to the fifteen minutes that caused it.

Compressed air leaks. Generated air is one of the most expensive utilities per unit of work, and the loss is invisible on paper because a leak never registers as downtime or a bad case.
Refrigeration duty. Cold rooms, blast freezers, and glycol loops respond to door discipline, defrost timing, and ambient load, none of which a case count reveals.
CIP and steam. Clean-in-place cycles and cook or pasteurize steps burn hot water and steam that get charged to the plant, not to the SKU or the changeover that triggered them.
Peak demand. A single coincident startup surge can set a monthly ratchet, and on a paper log that fifteen-minute window is simply gone.
Idle running. Conveyors, pumps, and air handlers left energized through breaks, sanitation, and short stops draw steady power with zero cases to show for it.

Measuring from the machine, not the utility bill

The change is to measure energy where the work happens, at the drive and the PLC, and to time-stamp it against what the line was actually doing. Most of the data already exists. Variable frequency drives report power and current. Compressors and chillers expose run state and load. Boilers report firing rate and feedwater. The machine controllers already know when a line was running, when it was in changeover, and when it was down for sanitation. The problem is that these signals live in separate controllers and historians that do not talk to each other or to the bill.

When energy is read at the PLC over the protocol the machine already speaks and stamped with the same clock as the line’s run state, kilowatts stop being a monthly total and become a curve you can lay against the schedule. Now the leak shows up as compressor power that stays high while the line is down. The demand spike shows up as three assets starting inside the same minute. The wasteful SKU shows up as more kilowatt-hours per thousand cases than the one that runs beside it. None of that requires a new sensor network in most plants. It requires reading what the drives and controllers already publish and joining it to the run data.

Turning kilowatts into cost per case

A CEO or COO does not act on a kilowatt curve, they act on cost per case and on the demand line of the bill. The value of machine-level energy monitoring is that it lets you divide the two things you can now both see at the same resolution: energy by asset and by interval on one side, cases and SKU and shift on the other. That division is the number that changes behavior. When a family-size run costs meaningfully more energy per case than a standard run, scheduling and staging can be arranged so the freezer and boiler are not fighting a cold start every changeover.

It also reframes the demand charge as an operations decision rather than a fixed cost. If the historian shows that the monthly peak is being set by simultaneous startups after sanitation, staggering those startups by a few minutes across the line can shave the ratchet without slowing a single case. That is a real dollar figure, visible on the next bill, traced to a specific fifteen-minute window that a person chose to change. The point of energy monitoring in CPG manufacturing is not a dashboard, it is a short list of decisions that each carry a name and a number.

What good looks like on the floor

A plant that has this working does not talk about energy in the abstract. The shift lead can see that line three’s air compressor ran at seventy percent load through a two-hour changeover, which means it was feeding a leak, and a maintenance tag gets written against a specific asset rather than a vague note to walk the line someday. Finance can see that the demand charge came from a single Monday startup and can ask a concrete question about sequencing. The plant manager can compare two SKUs on energy per case and put the cheaper sequence on the schedule when the order book allows it.

Getting there usually means getting the plant off paper first, because energy data that lives on an hourly round sheet cannot be joined to anything. The signals have to be live, machine-timed, and in one place with the case counts. Once that data layer exists, energy is just one more thing the plant can finally see clearly, alongside downtime, quality, and changeover.

Where Harmony fits

Harmony is an AI-native operating system for American manufacturing that gets plants off paper and spreadsheets and ready for AI, and energy is a natural first thing to make visible once the data is live. Harmony connects at the PLC, Allen-Bradley and Rockwell, Siemens, Omron, Mitsubishi, over OPC UA or whatever the machine already speaks, so compressor load, drive power, and boiler firing are read from the line rather than copied off an hourly round sheet. It unifies that machine data with your software systems and the paper the crew still fills out into one live data layer, which is what lets a kilowatt reading sit next to a case count and a SKU on the same clock. Getting there is the same work as moving to paperless manufacturing software, and it applies directly to a consumer packaged goods operation where refrigeration, compressed air, and CIP steam drive most of the bill. Then Harmony layers AI on top for search, scheduling, predictive maintenance, and back-office automation across finance and procurement, and the AI proposes while a person approves, because a decision that changes a demand charge should have a human name on it. We are software and hardware agnostic, and our published pilot is $15–20K one-time over 4–6 weeks with forward-deployed engineers on-site and working software by week three. Customers include Mossberg, MoonPie, and CLS.