The Hidden Operating Costs of Supermarkets

What it costs to keep a system running

A refrigerated case costs something every hour it runs, and most of that cost never appears on a line called "refrigeration". Electricity sits in the power bill beside lighting and air-conditioning. Refrigerant and call-outs sit in repairs. Spoiled stock sits in markdowns and shrink. Wages for the technician sit in maintenance. Part 2 asked where an operator's next dollar should go. This part asks what the dollars already committed to existing equipment are costing, and what an operator would have to measure to tell whether continuing to run it is still the better course.

We have kept to the limits we set for the series. Everything here comes from public sources. We have not seen, asked for or estimated any store's energy, maintenance, sales or shrinkage figures. We make no Fiji savings estimates and no financial projections, and where a number comes from overseas we say so and leave it overseas. We do not treat age as a fault: a well-kept older system that performs adequately is a legitimate way to run a store.

The draft separates what Fiji sources document from what overseas technical evidence suggests. "Reported" means an organisation or newspaper said it. "Our reading" is interpretation. The last section lists every source and how we read it, and what changed from Draft 01.

One result first. On public evidence, our first hypothesis, that older refrigeration, lighting and building systems raise energy and maintenance costs, is still Inconclusive. That means the evidence is insufficient, not that there is no difference. Nothing found for this part moves it, and the other three hypotheses are unchanged.

Where the cost shows up

Four costs make up the continuing cost of a refrigeration system. They land in different parts of a store's accounts, which is a large part of why they are easy to miss.

Two things follow. First, our reading: no single line in the accounts shows what it costs to keep a system in service, so an operator who reads only the repair bill will undercount. Second, the Fiji column is mostly empty. We found no published Fiji figure for any of the four. That does not show none exists, only that our public search did not reach one, and it is why section 5 is about what an operator would measure.

A note on support. Reuters, citing Danfoss, reported on 2 November that certain older controller models are end-of-life and no longer receive firmware updates (Evidence Record E21 to E23). That is a vendor statement about specific products in a different market, and we do not repeat the incident account. The general point needs no incident: whether the maker, or a local agent, still supplies spares and support for a system's controller is a fact to check system by system, and it belongs in the cost of keeping that system running.

What Fiji documents: electricity and refrigerant rules

This section covers only Fiji sources and is dated 3 November 2026. Neither the electricity position nor our reading of the refrigerant law is legal or regulatory advice.

Electricity

The January 2026 increase did not take effect, and the tariff is now deferred. In order:

• The December 2025 tariff notice (EFL's public notice, cited in Part 2) set out new tiered commercial rates from 1 January 2026. Fiji Times headlines in our search results record an approval "effective January 2026" and then a suspension for public consultation. We read the headlines only. On 13 January 2026 the Fiji Commerce and Employers Federation asked the FCCC to set aside its 19 December 2025 determination and restart the review (Fiji Sun, 14 January).
• FCCC, 1 May 2026. The review was deferred to 31 July 2026, and the current structure, set under the authorisation of 1 October 2019, "remains unchanged during this period". Read in the FCCC's own PDF.
• FCCC, 21 to 26 May 2026. An interim fuel surcharge of 5.91 cents per kWh applies to existing tariffs from 26 May 2026, across all customer categories (FBC News, 21 and 26 May). FBC's 26 May report says it came into effect "yesterday", which would be 25 May. The FCCC's 3 November release gives 26 May, and we use that.
• FCCC, 3 November 2026. The proposed electricity tariff is deferred to 31 March 2027, with public notice to be given before any decision to implement it. The interim surcharge stays at 5.91 cents per kWh until December 2026. The FCCC gave as reasons that it did not want consumers and businesses to carry the new tariff and the surcharge at the same time, that fuel prices remain unstable, and that the continued surcharge allows a reconciliation of over- or under-recovery. It also cited falling hydro dam levels, dry conditions in the Western Division and a forecast "super El Niño", which mean more fuel is burned. EFL has asked for a further increase in the surcharge, which the FCCC is assessing. The current rate stays unchanged meanwhile. We read this in the FCCC's media release PDF.

The 3 November FBC News report (5:04 pm) and the Fijivillage report of the same day say the same. Both are secondary reporting, and we treat the FCCC release as the determination itself. The FCCC's wording is the authority, not the newspaper summaries, and its page does not show the release's formal legal status beyond the media release.

Our reading: what an operator pays today is the 2019 structure plus the interim surcharge. Both can change, and a further surcharge request is pending. We have not verified the commercial band rates in force from an FCCC schedule, so we give no cents-per-unit figure for a supermarket. The FCCC's own advice to consumers, that every unit saved reduces the fuel burned, is a statement and not a finding about any store. For an operator, the practical consequence is that kWh used is the quantity within its control, and the price per kWh is not. Correction to Part 2: its statement that the commercial tariff moves from 42.95 to 63.74 cents a unit from 1 January 2026 has been corrected in Part 2 v0.2, v0.3 and Evidence Record E14 and D6, with a dated correction note in each.

Refrigerant rules

We read the Ozone Depleting Substances Act 1998 (the principal Act) and the Ozone Depleting Substances (Amendment) Act 2020 on PacLII, each through our fetch tool, which returns summaries and quotations and not a page-by-page reading. We have not found or read any regulations made under the principal Act.

The 2020 amending Act. As we read it, section 14(2) as replaced says no person shall import, export or manufacture a controlled substance in bulk listed in Part 2 of Schedule 1, which covers HCFCs, "with effect from 1 January 2030". Section 14(7A) does the same for HCFC chillers from 1 January 2021, and section 14(7B) bars manufacture of HFC substances from 1 January 2021. The Director may grant exemptions, to be determined under the Montreal Protocol (section 14A). The restrictions concern import, export and manufacture. The amending Act contains no provision on servicing, on continued use of existing equipment, on refrigerant already in the country, or on transition for existing stock. Section 1(2) says it commences on a date or dates appointed by the Minister in the Gazette. We have not checked the Gazette, so we do not know when each provision took effect.

The principal Act. It does not stay silent on handling. As we read it:

• Section 14 sets the phase-out dates and prohibitions, including for air-conditioners, refrigerators and compressors.
• Section 15 requires workshops to use an approved recovery and recycling unit.
• Section 16 requires permits to import, sell, store or process controlled substances.
• Section 17 requires a licence to recycle, recover or capture them.
• Section 18 requires, where possible, that a controlled substance be recovered before disposal; says that when equipment is serviced or repaired the substance is replaced with a suitable substitute where practicable; and makes it an offence to discharge or release a controlled substance.

The Act sets no leak-check schedule, repair deadline, charge limit or operating standard for equipment already in use, and it does not require existing equipment to be retired. The servicing duty in section 18 is one of substitution where practicable.

The 1 August 2025 equipment ban. The Fiji Times reported on 23 July 2025 that the Ministry of Environment and Climate Change would ban the import, export and manufacture of equipment pre-charged with HCFCs from 1 August 2025. We have not located the legal instrument behind that date, and the amending Act text we read does not mention it. The article does not address installed equipment or servicing.

Our reading: nothing we read requires an operator to retire working equipment, and nothing we read ends servicing or the use of refrigerant already held or reclaimed in Fiji on 1 January 2030. The 2030 date concerns bulk import, export and manufacture, and is subject to exemptions. What it may do to the price and availability of virgin HCFC gas for existing systems is a question we cannot answer from these texts, and it depends on stocks, reclaimed gas, exemptions and regulations we have not read. We do not predict it. The handling duties are relevant to the cost of keeping a system in service in one respect: recovery, licensed servicing and the offence of releasing a controlled substance bear on how a leak is handled. How they apply to a leaking operating system is a legal question we have not explored. We also do not know which refrigerants Fiji supermarkets use.

What overseas technical evidence adds

This comes from outside Fiji, mostly the United States and Europe, in cooler and drier climates. It describes mechanisms and ranges. It is not a measurement of any Fiji store.

Electricity. The 2005 US EPA report prepared by ICF estimated refrigeration at 30 to 50 per cent of total store energy, up to 65 per cent for small supermarkets (Evidence Record E6), and calls its own sources "varying or even conflicting". An International Institute of Refrigeration page quotes 29 to 48 per cent for Europe from sources we did not open. Our reading: refrigeration is one of the larger loads in a food store, but the share varies with climate, store type and what else is on the meter, and a share of the bill is not a saving. It tells an operator where to look, not what a change would recover.

Refrigerant. The EPA's 2011 GreenChill guideline says a centralised direct-expansion system can lose as much as 25 per cent of its charge a year. It reports GreenChill partners averaging 12 per cent in 2008, some targeting under 7 per cent, and individual stores sustaining under 5 per cent, and notes these ratios differ from the regulatory leak-rate definition. A 2005 report cites about 15 per cent for newer and over 30 per cent for older direct-expansion equipment (IEA, 2003, via E6). The guideline's causes of leakage are mostly workmanship and housekeeping: poor brazing, loose or over-tightened fittings, missing valve caps, vibration, thermal movement, corrosion, unsupported pipework. Its remedy is records and routine: log charge, top-ups and repairs, check for leaks every one to two months, and do not top up without finding the leak. Our reading: leakage looks at least as much a matter of installation quality and inspection practice as of birth year, although the IEA figure associates older equipment with higher leakage. We could not separate the two, and a Fiji operator should assume neither.

Faults. A laboratory study of a CO2 refrigeration rack, summarised by the International Institute of Refrigeration, introduced six faults: a freezer-case door left open, ice on a coil, a failed expansion valve, a failed fan motor, and blocked air paths at the condenser and evaporator. They changed compressor behaviour and, for the open door, raised power use. The page gives no effect sizes, sample or climate. Each of these faults can occur in new or old equipment, and each would show in a store's own records.

Component retrofit. A 2000 Southern California Edison laboratory test of one older meat display case found efficient lamps and ballasts cut total system power by 5 per cent, and adding efficient fan motors cut it by 27 per cent in total and cooling load by 13 per cent (E10). It was one case, tested at 75°F and 55 per cent humidity with 2000-era components, with no cost analysis. It shows that component changes can matter. It does not show that age drives cost, and it says nothing about replacing a whole system.

What this adds up to. The overseas material documents mechanisms: refrigeration is a large load, leaks cost refrigerant and energy, faults raise consumption, and component changes can cut it. It gives no measured comparison of older and newer systems in a warm, humid climate, which hypothesis 1 needs. That gap is why the hypothesis stays Inconclusive.

Maintain or replace: what to measure and how to compare

This section is a method, not a finding. It uses no Fiji figures and no worked example with numbers, so that none can be mistaken for results. Any equipment will be replaced eventually, so the practical question is when, not whether. "Keep running it" means "replace later", and later has its own costs.

Measure the existing system

Age is a reason to look at a system, not a verdict on it. These are the records that show what it costs.

1. Refrigeration electricity, on its own meter. A store's bill mixes refrigeration with lighting, air-conditioning, bakery and everything else, and sub-metering the racks or condensing units separates them. Record kWh and not dollars, over at least a full year so that seasons show, with outdoor temperature and humidity, trading hours and stock volumes beside it. A hot month and a faulty compressor look identical on a bill without that context.
2. Refrigerant added. Date, circuit and kilograms of each top-up against the system's charge gives an annual loss rate. Record the refrigerant type and the cost of the gas. The call-out and repair for a leak go in the maintenance record (measure 4), tagged as a leak job, so they are counted once. Given section 3, also record who handled it and what was recovered.
3. Breakdowns and near-misses. Each call-out with date, time reported, time attended, time restored, cause and parts, plus temperature excursions from controllers or data loggers, including those that did not become a loss. Put a cost on each: stock written off or marked down, trading hours lost, staff time.
4. Maintenance spend, planned and reactive, by asset group and year. The trend matters more than the total. The same component failing repeatedly is a different signal from scattered unrelated repairs.
5. Support and spares. Whether the maker or a local agent still supports the controller and key components, how long imported parts take, and how many local technicians can service the system.
6. Condition. Coil corrosion, door gaskets, insulation, fan motors, defrost performance. The overseas guideline and fault studies point to these, and they can be inspected directly.

Price the alternatives

There are usually three options, not two: continue as is, upgrade specific components (fans, lighting, door seals, controls), or replace. The one overseas test we have (E10) found its saving in component changes, so the third option belongs in the comparison. For each alternative to continuing, get installed price from more than one supplier including removal of the old equipment and recovery of its refrigerant; the refrigerant used and the supplier's statement on its availability; the supplier's rated energy use, treated as a claim to be verified by measurement after commissioning; warranty and service terms, and whether the service capacity exists in Fiji; and how trading is protected during the changeover.

Compare them over the same years

Compare options by the present value of every cost over one common horizon of N years, using the business's own cost of capital r, the same rate it applies to other uses of funds (Part 2). Choose N long enough that the existing system would certainly be replaced within it, and let k run from 0 to k(max), the latest year the existing system can credibly be kept, judged from measures 3 to 6. Every option then ends the horizon with the same kind of asset, a replacement system of age N − k, so the old system needs no terminal value. Cash flows fall at the end of each year, and the replacement is bought at the end of year k and runs from year k + 1. For the option "continue for k years, then replace", where k = 0 means replace now:

The cheapest k is the best timing on cost alone. The subscripts matter, because two clocks run at once. Calendar year t sets prices: the tariff, gas, labour and parts. Asset age sets performance: how much electricity, refrigerant and repair the equipment needs. The replacement's running cost in year t therefore depends on both its age, a = t − k, and the calendar year t, and its residual value depends on its age at the horizon, N − k. The existing system's cost in year t likewise depends on its age in that year and on the calendar year. Treating either cost as a function of one clock alone misstates the comparison. Each term comes from the measures above or from a quote:

This comparison is consistent only if some rules hold:

• One horizon, one duty. Every option keeps the cold chain running for the same N years and ends the horizon with a working system. Setting N beyond the old system's credible life and limiting k to k(max) means the old system never needs a terminal value, and no option keeps it past the point measures 3 to 6 say it can be kept.
• Two clocks, calendar year and asset age. Its quantities of energy, refrigerant and repair follow its age, its prices follow the calendar year, and its residual value follows its age at the horizon. A later replacement is younger when the horizon ends and so is worth more at that point. Holding the residual value constant across k, or ignoring it, would therefore favour replacing early.
• One price basis. Real or nominal cash flows throughout, with a matching r. The tariff path is in the same basis.
• Each cost once. Changeover disruption, removal, refrigerant recovery and staff training go in I. The failure costs of the system that is running go in C(old) or C(new). Anything recovered from the old equipment is netted in I and not entered again as a separate residual value. Disruption counted in I is not counted again as a failure cost. A call-out or repair invoice sits in one place only: upkeep, tagged by cause, with the refrigerant line holding the gas alone and the stock lost in a breakdown sitting under breakdowns.
• Same scope on both sides. Every option counts the same categories of cost. A component upgrade is another schedule of spending and running costs, compared in the same way.
• Failure is a cost, not an afterthought. The failure cost in C(old) should come from the operator's own call-out and excursion history. We do not supply a figure.
• Tax and financing. This comparison leaves out tax, depreciation allowances and financing terms. If they are included, they are included for every option on the same basis.

Investment price and replacement life. Two assumptions sit behind the formula and should be stated whenever it is used. First, the installed price I is in constant money and is taken to be the same whichever year k the replacement is bought, because we do not forecast real price changes. If an operator expects the real price to change, I becomes a price quoted for each k. Second, the replacement's useful life L is taken to be at least N − k for every k considered, so it is not itself replaced inside the horizon. If L is shorter, a second replacement and its own residual value must be added. The residual value S(a) is the value of the remaining life of a replacement a years old, estimated by one declared method and applied to every k. All three are assumptions and not measurements.

Draft 01 offered a one-year version of this test, the annual running-cost gap set against one year's cost of the capital. That is a special case, valid only if costs are flat and lives are equal, and the multi-year version replaces it.

Reading the result

Our reading: if PVC is lowest at a large k, and top-ups are steady, energy use per unit of cooling duty is flat over the years, call-outs are few and quick, and parts and technicians are available, the records support continuing, and a well-kept older system is a sound choice. If PVC falls toward k = 0 because top-ups are rising, the same components keep failing, excursions are frequent, or spares and service are disappearing, they point the other way. If PVC is nearly flat across k, waiting is cheap and a year of better measurement may be worth more than a decision now. Age alone predicts none of these.

Finally, test the ranking. Change the tariff path, failure cost, horizon and r over defensible ranges, and note which changes flip the best k. If none do, the choice is robust. If one does, that input is the one worth measuring better before committing capital (claims register C11). The tariff position in section 3 is one reason to do this: price the measured kWh at the tariff in force and test a spread of other tariffs, and do not forecast one. Whether a given operator already keeps these records is not something public sources show, and we draw no conclusion about any operator's measurement. Where the records do not exist, starting them is the cheap first step.

Hypotheses, limits, changes and sources

Where the four hypotheses stand

No outcome changes. This part tested none of them. It asked only what its public evidence adds to each, using Study Brief v0.4 and Claims Register v0.6.

1. Ageing infrastructure adds operating cost (C2): Inconclusive on public evidence, as recorded on 27 October. Nothing here measures how age changes cost in a store, and the overseas leakage evidence is mixed on whether age or workmanship matters more.
2. Apparent competitive disadvantage also stems from layout, presentation, assortment, convenience and customer expectations (C3): untouched by this part.
3. Location, trade area and timing of new-format entry may matter more than store age (C4): untouched. Tariffs and refrigerant rules apply to every operator, so they cannot explain differences between stores.
4. Keeping an older store can be the commercially rational choice (C5): section 5 gives a symmetrical cost-side method, and no evidence that retention is the better investment in any store. Evidence that an older system performs adequately would not by itself show that.

What the evidence cannot establish

• That older equipment costs more to run in Fiji, or that it does not. We have no measured comparison by equipment age from a warm, humid climate and no store data.
• What a Fiji supermarket spends on refrigeration electricity, refrigerant, breakdowns or upkeep. The overseas shares are ranges from cooler markets.
• What any saving from replacement or retrofit would be here. The one laboratory retrofit result (E10) is one case in controlled conditions.
• Which electricity rate applies to a supermarket today or next year. We have not verified commercial band rates from an FCCC schedule, a further surcharge increase is under assessment, and the tariff decision due by 31 March 2027 is not made.
• What Fiji's refrigerant law means in practice for existing systems. We read the principal Act and the 2020 amending Act through summaries, not regulations, the Gazette, or the instrument behind the 1 August 2025 equipment ban. We cannot say what the 2030 date does to supply and price of refrigerant for existing systems, and we are not giving legal advice.
• Which refrigerants Fiji supermarket systems use.
• Whether any named operator's equipment is old, failing or sound. This draft names none.
• That the overseas leak-rate figures are representative. They come from US programme participants, a 2003 IEA figure reached through a 2005 draft, and a vendor-published summary.
• How equipment age affects the risk of failure. We found no source that measures it.

Changes from Draft 01 and related corrections

• Tariff. The FCCC's 3 November release itself (read in full) confirms the deferral to 31 March 2027 and the surcharge to December 2026. FBC News and Fijivillage of 3 November corroborate it and are treated as secondary. Part 2 v0.2, v0.3 and Evidence Record E14 and D6 now carry dated corrections.
• Refrigerant law. Draft 01 said HCFC refrigerant for topping up existing systems "has a stated end date" and that the amending Act said nothing on servicing. Both statements are withdrawn or narrowed. The 2030 date applies to bulk import, export and manufacture. The principal Act has handling, recovery, licensing and release provisions that Draft 01 had not read.
• Calculation. The one-year comparison is replaced by the multi-year, timing-based comparison in section 5. A final check on 4 November tightened it: the replacement's running cost and residual value now follow its age, the horizon is set so the old system needs no terminal value, and each cost is counted once. A further correction on 4 November (evening) separated calendar-year prices from asset-age performance, corrected the direction of the residual-value bias, and stated the investment-price and replacement-life assumptions. A final production note assigns each call-out and repair cost to upkeep only, so it is not counted again under refrigerant.
• Structure. The cost lines come first, then the rules, then overseas mechanisms, then the method.
• Unchanged. The four hypothesis outcomes, the limits, and the absence of any Fiji cost estimate.

Sources, and how each was read

"Full" means we opened the page or document through our fetch tool, which returns a summary and quotations and not a page-by-page reading. "Headline" means only a search-result title. "Record" means we rely on our own Evidence Log or Record entry, not re-read here.

Fiji electricity

• FCCC media release, "Electricity tariff deferred to March 2027 and fuel surcharge continued to December", 3 November 2026. Regulator's statement. Full (PDF).
• FCCC media release, "Electricity tariff review update", 1 May 2026. Full (PDF).
• FBC News, "Fuel surcharge remains until December", 3 November 2026. Secondary report. Full.
• Fijivillage, "Electricity tariff deferred until March 2027, fuel surcharge to continue until December 2026", 3 November 2026. Secondary report. Full.
• FBC News, FCCC completes review of fuel surcharge and bus fare applications, 21 May 2026 and FBC News, EFL acknowledges interim fuel surcharge authorisation, 26 May 2026. Secondary reports. Full.
• Fiji Sun, "FCCC told to scrap electricity tariff decision, restart review", 14 January 2026. Full.
• The Fiji Times, "FCCC approves tiered electricity tariff increase, effective January 2026" and "EXCLUSIVE: FCCC suspends electricity tariff for public consultation". Headline only.
• EFL public notice, tariff review December 2025. Part 2 Evidence Record E14. Record.

Fiji refrigerant law and equipment

• Ozone Depleting Substances Act 1998, PacLII. Legislation. Full.
• Ozone Depleting Substances (Amendment) Act 2020, PacLII. Legislation. Full.
• The Fiji Times, "Fiji bans HCFC-loaded equipment from August", 23 July 2025. Report of a ministry announcement. Full.
• Reuters, 2 November 2026, and Part 2 Evidence Record E21 to E23, for the Danfoss end-of-life statement. Record.

Overseas technical evidence

• US EPA GreenChill, Commercial Refrigeration Leak Prevention and Repairs, EPA 430-B-11-001, May 2011. Full.
• US EPA / ICF supermarket refrigeration report, revised draft, 30 November 2005. Evidence Log E6. Record.
• ACHR News, "The effects of retrofitting on an older display case", September 2000. Evidence Log E10. Record.
• International Institute of Refrigeration, "Faults and energy consumption of CO2 refrigeration systems in supermarkets". Summary page, authors and sample not named. Full.

Our own documents: Study Brief v0.4, Claims Register v0.6 (C2 to C6, C11), Evidence Log v0.12, Part 2 Narrative v0.2 and v0.3 and Evidence Record v0.2 (as corrected), Part 3 Narrative Draft 01 (preserved).