What your electricity bill is already telling you

Welcome to the first edition of Follow the Energy. Each edition follows one of my posts wherever the energy goes, from your site and your bill to the grid, the market and the money, and works it through in full: the method, the numbers behind it, and the questions you'll be asked when you take it into a meeting.

We start where every fix starts: your bill.

In the post that opened my series Cheapest Fix First, I gave you an hour's job:

Pull your last 24 months of bills. For each month, write down the kWh, the maximum demand in kVA (the busiest half-hour's average) and the total in rand. Divide the total by the kWh, and compare each month with the same month a year earlier. A jump well beyond that year's tariff increase, with about the same activity on site, is your first clue to a charge you can control.

It looks too simple to matter. It isn't. Here's why each step is there, a year of bills worked through, and the answers to the questions your finance head is likely to ask.

Why the bill comes first

Every site has its bills. You don't need a meter, a consultant or a budget: only the bills, a spreadsheet and an hour.

The bill is also where the cheapest fixes hide. Step 2 of Cheapest Fix First is “pay less”: the charges you control without buying equipment, such as the capacity you book, the hours you run and your power factor. Some cost nothing to fix, but you can't find them without knowing which months to open.

And it's the baseline for everything after. When you cut energy use or put solar on the roof, you'll prove the saving against these numbers.

What you need

• Your last 24 monthly bills, every page. Ask your supplier for any you're missing.
• A spreadsheet with one row a month.
• A note of anything that changed on site in those two years: a new line, a longer shutdown, a new shift.

Why 24 months and not 12? You'll compare each month with the same month a year earlier, so you need two years.

Three numbers a month

kWh: how much energy you used. It moves with production, and so does most of your bill. On Eskom's Megaflex, about 90% of a day-shift site's bill moves with kWh.

Maximum demand in kVA: the hardest you pulled. Your meter averages your demand over every half-hour. The bill's maximum demand is the highest of those averages in the month (on some tariffs, only in the hours charged for it). A spike of a few minutes counts only as its share of its half-hour.

Two kinds of charge ride on it:

• Capacity charges on your notified maximum demand (NMD), the capacity you booked with your supplier. On Eskom's tariffs they're billed on the higher of your NMD and your highest demand in the last 12 months, so one bad month sets them for a year. Go over the NMD and that month carries an extra charge too.
• A demand charge on the month's peak, on some tariffs. Eskom's Megaflex has one; Miniflex doesn't. Many municipal tariffs charge far more: City Power's industrial medium-voltage time-of-use tariff charges R431.11 a month for every kVA of the month's peak.

Why kVA and not kW? kW is the power that does work; kVA is what the supply has to carry. The gap is your power factor: kVA = kW ÷ power factor. In Demo Foods, my illustrative case study (a model food factory, not a real company), 538 kW at a power factor of 0.86 bills as 626 kVA. At 0.98, the same work bills as 549 kVA.

The total in rand. All of it, excluding VAT, the same way every month.

One division: what a kWh really costs you

Divide the total by the kWh. The answer is your all-in price per kWh: energy, capacity, network and fixed charges together.

It's useful because it strips out volume. Produce 10% more and the total rises, but the rand per kWh barely moves. So when it does move, something other than volume has changed: the price, the hours you run, your peak or the bill itself.

One caution, which we'll come back to: this number is an average. It's a screen for the bill, not the value of a kWh you save.

Compare with the same month a year earlier

Never compare July with June. Eskom's high-demand season runs from June to August, and since April 2025 its winter peak energy rate has been about six times its summer off-peak rate. Production and public holidays move from month to month too. July against last July cancels all of that.

Then ask how much it should have risen. Prices rise every year, so every month will jump; the question is whether it jumped by more than the increase.

• Eskom raises its prices on 1 April. Its average increase for direct customers was 12.74% in April 2025 and 8.76% in April 2026.
• Most municipalities raise theirs on 1 July, by their own amounts. Use your municipality's.
• Use the increase that fell between the two months: November 2025 against November 2024 spans April 2025; July 2026 against July 2025 spans April 2026.

A warning about averages. The headline increase is an average across all customers and charges, and your tariff can move differently. When Eskom restructured its tariffs in April 2025, an illustrative day-shift site on Megaflex saw its average price per kWh rise 6.5%, not 12.74% (then 8.6% in April 2026). The best yardstick is your own: price last year's month on this year's rates from your supplier's tariff schedule. The headline increase is a fair first pass.

Three tripwires

With production about the same, I flag a month when any of these trips:

1. The rand per kWh rose more than about 5 points beyond the tariff increase.
2. The kWh or the kVA moved by more than about 10%.
3. The kVA went over your NMD.

A month's bill wobbles by a few per cent with the days billed, public holidays and the production mix; these thresholds sit above that noise. They're my rule of thumb: tighten them once you know your own site's wobble.

A year of bills, worked through

An illustrative day-shift site on Eskom's Megaflex, with production about the same every month except December, when the shutdown ran longer. Each figure is the change against the same month a year earlier. The site's own tariff rose 6.5% for the November-to-March pairs and 8.6% for the April-to-October pairs.

Four stories sit in that table.

December: a quiet month, explained. A longer shutdown cut output, and kWh fell 20% with it; tripwire 2 doesn't count it, because production moved too. More telling, the rand per kWh rose only 2.5 points beyond the tariff. About 90% of this bill moves with kWh, so only the fixed and capacity charges were spread over fewer units. On Eskom since April 2025, a quiet month barely moves the rand per kWh, which means that when it does jump, volume is rarely the reason.

February: a demand problem the rand per kWh hides. kWh held steady, but the peak rose 14%, from 610 to 695 kVA, past an NMD of 650. The rand per kWh rose only 1.4 points beyond the tariff, so tripwire 1 stays quiet; tripwires 2 and 3 catch it. The cost runs past February: the capacity charges are now billed on 695 kVA for the next 12 months. At 2026/27 rates that's about R750 a kVA a year, so 45 kVA costs about R33,800 a year, plus February's charge for going over. The usual cause is machines or compressors starting together at the start of a shift, and staggering them costs nothing. On City Power's tariff, the same 85 kVA would have added about R36,600 to February's bill alone.

March and April: the readings, not the site. kWh fell 12% in March and rose 13% in April while production held. That's the mark of an estimated reading and its catch-up: look for an “E” or “Est” beside March's reading. A run of estimates hides errors and makes every other check unreliable. If kWh and kVA jump together and stay there with nothing new on site, ask instead about the meter's multiplier: a wrong one puts every kWh and kVA out by the same factor.

June to August: the hours, not the amount. kWh and kVA barely moved, but the rand per kWh ran 8 to 9 points beyond the tariff for three months, and only in winter. That points at when the site uses power. Here, part of the day shift had moved into the evening, into the 17:00 to 20:00 winter peak, when energy costs most. Divide each bill's peak kWh by its total: if that share rose, you've found it.

In rand: last July the site paid R768,000 for 240,000 kWh, R3.20 a kWh. This July it paid R914,500 for 242,400 kWh, R3.77 a kWh. At the 8.6% tariff rise it should have been about R3.48. The gap, R0.30 on every kWh, is about R72,000 in July alone.

Reading any flag

What a kVA and a kWh are worth

Once you've found a flag, someone will ask what fixing it is worth. Four numbers, on Eskom's 2026/27 rates for a supply between 500 V and 66 kV within 300 km of Johannesburg, excluding VAT:

• A kVA off an NMD that sits above your real peak: about R750 a year, on Megaflex or Miniflex. Lowering the NMD takes a letter and, normally, 12 months' notice.
• A kVA off the month's peak: about R315 a year on Megaflex, and nothing on Miniflex.
• A 10% cut in kWh at a day-shift site on Megaflex takes 5.9% off the bill if it all comes off-peak, 9.2% if it's spread evenly, and up to 17.7% if it all comes in peak hours.
• A rooftop solar kWh used on site saves about 205c on Megaflex in Johannesburg, against a day-shift site's average of 262c.

That last one leads to the most common mistake.

The questions you'll be asked

“Of course it went up. Eskom raised its prices.” Yes, and you've allowed for that: the flag is only the part above the increase. Because the headline increase is an average, the best yardstick is last year's month priced on this year's rates.

“So every kWh we save is worth R3.77?” No. The rand per kWh is an average that includes capacity and fixed charges, and a saved kWh doesn't touch most of those. What a kWh saves depends on its hour: much less than the average off-peak, much more in a winter peak hour. A solar proposal that values every kWh at your average overstates the saving: on Megaflex, a rooftop solar kWh saves about 205c against a 262c average, so the average overstates it by about 27%.

“We produced less this year. Doesn't that explain it?” Less than you'd think. On Eskom since April 2025, only about a tenth of a day-shift site's bill stays put when kWh falls. A 20% fall in kWh lifts the rand per kWh by about 2.5%; a 10-point jump needs another reason.

“Couldn't the utility simply have got it wrong?” It happens: estimates never corrected, a wrong meter multiplier, the wrong tariff. Errors pay back once, as a refund or a credit, and only if you claim in writing. Refunds are usually limited to about three years, so don't wait.

“Isn't this too simple to trust?” It's meant to be simple. It doesn't tell you what's wrong; it tells you which months to open and which lines to read. It costs an hour and needs no meter data.

What this check can't tell you

• When in the day you use power. A bill shows totals by time-of-use period at best; half-hourly meter data shows the shape of every day.
• Whether each charge is right. For that, check the tariff, the NMD, the power factor, the readings, the multiplier, the peak share and the load factor: the seven checks in my free Bill Check.
• What the site could save by using less. That takes the site itself: step 1 of Cheapest Fix First, where the series goes next.

Your next hour

1. Build the sheet: 24 rows, three numbers, the division, and each month's change on a year earlier.
2. Mark every month that trips a tripwire, and note what changed on site that month.
3. Open those months' bills and read the lines.
4. Work through the seven checks in the free Bill Check.
5. Ask your supplier or meter provider for half-hourly data. You'll need it for what comes next.

When you've done it, tell me what your first flag was. Email me at thabiso.lungakali@gmail.com, or answer under the post on LinkedIn. I read every reply.

Thabiso Kali
Energy engineer who speaks finance

The site in the worked example is illustrative. All rand exclude VAT. Tariffs change every year, so check your supplier's current schedule. A general guide, not advice on a particular bill or supply agreement.

Sources and working

Every figure in the edition, where it comes from and how it was worked out. All rand exclude VAT; Eskom figures are for a supply between 500 V and 66 kV within 300 km of Johannesburg unless the line says otherwise.

12.74% (April 2025)

Source: Eskom, Historical average prices and increase

Working: NERSA's approved average for Eskom's direct customers, 2025/26

8.76% (April 2026)

Source: Eskom, Tariffs and Charges Booklet 2026/27

Working: The average for direct customers

Winter peak about six times summer off-peak

Source: Eskom, 2025/2026 price increase

Working: The ratio fell from 8:1 to 6:1 on 1 April 2025

June to August; capacity charges on the higher of the NMD and the 12-month high; an extra charge over the NMD

Source: Eskom's booklet, as summarised in my Bill Check

Working: Municipal rules vary

About 90% of the bill moves with kWh; 6.5% and 8.6%; R750 a kVA off the NMD; R315 off the monthly peak; 5.9%, 9.2% and 17.7%

Source: My analysis of an illustrative 1 MW day-shift site on Megaflex and Miniflex, 2024/25 to 2026/27, from Eskom's published rates, checked by a second, separate calculation

Working: 240.98c ÷ 226.36c = 1.065; 261.62c ÷ 240.98c = 1.086; R750.60 and R315.48 a kVA a year at 2026/27 rates

Lowering the NMD takes a letter and, normally, 12 months' notice

Source: Eskom's NMD rules

Working: "Normally": your supply agreement decides

205c against 262c; overstated by about 27%

Source: My analysis of Megaflex 2026/27 rates against PVGIS hourly solar output for Johannesburg

Working: 261.62c ÷ 205.31c = 1.274

R431.11 a kVA a month

Source: City Power, 2026/27 tariffs, industrial medium-voltage time-of-use

Working: 85 kVA × R431.11 = R36,644

538 kW, 0.86, 626 kVA; 549 kVA at 0.98

Source: Demo Foods, the illustrative case study in the Energy Blueprint

Working: 538 ÷ 0.86 = 626; 538 ÷ 0.98 = 549

The worked example

Source: Constructed to be realistic for that day-shift site; not a real site

Working: 695 − 650 = 45 kVA; 45 × R750.60 = R33,777. July: R768,000 ÷ 240,000 = R3.20; R914,500 ÷ 242,400 = R3.77; R3.20 × 1.086 = R3.48; (R3.773 − R3.475) × 242,400 = R72,138. A 20% kWh fall with 10% fixed: (0.9 × 0.8 + 0.1) ÷ 0.8 = 1.025

Refunds usually limited to about three years

Source: My Bill Check

Working: "About": it depends on the supplier and the claim

The tripwires (5 points, 10%)

Source: My rule of thumb

Working: Said so in the edition