The ten data problems that size a battery for you

1. kW read as kWh (or the other way round)

A 15-minute value can be energy (kWh in the interval) or average power (kW). At 15 minutes, kW = 4 × kWh: 100 kWh in a quarter-hour is an average of 400 kW. Read the wrong way, the site's consumption and peak are four times too high or too low. Check the column header and the unit row; if there is none, compare the year's sum with the invoices. At 30-minute data the factor is 2.

2. A meter register instead of interval values

Some exports give the meter's running total, which only ever goes up. Consumption per interval is the difference between consecutive readings. Watch for drops in the register: a meter exchange or a rollover, which must not become a huge negative or positive interval.

3. The wrong or a mixed interval length

The interval is the most common step between timestamps: 15 minutes for most C&I meters, 30 in the UK, 60 in many portals. A file that changes length in the middle of the year usually means a meter exchange or a new resolution, and needs two imports rather than one.

4. Timestamps that label the end of the interval

A file whose day starts at 00:15, or that contains 24:00, labels each interval by its end. Read as the start, every value moves one step late — which shifts peaks across month ends and across tariff time windows. It is common: in a review of 24 half-hourly formats from UK portals, some files labelled the period end and others the start (Leigh Dodds, 2021).

5. Daylight saving days

In local time, the spring day has 23 hours (92 quarter-hours) and the autumn day 25 (100). A file with exactly 96 values on every day ignores daylight saving: it is either in UTC or in winter time all year, and an hour of the autumn day is missing or doubled. That matters most in winter, when many peaks fall.

6. Duplicates

The same interval twice with the same value is harmless: count it once. The same interval twice with different values usually means two meters or two channels mixed in one file, and needs a decision before anything else.

7. Gaps

Missing intervals understate consumption and can hide the peak a battery has to cut. Short gaps of up to an hour can be interpolated; longer ones filled from the same weekday's profile, scaled to the bill, or left out. Zorowatt's checker marks the data as an error when more than 10% of the intervals are missing, and every fill is a proposal you accept, never automatic.

8. Stuck zeros and flatlines

More than four hours at exactly 0 kW on a site that otherwise never drops to zero suggests a stuck meter or a shutdown; the same non-zero value for more than six hours often marks an estimated or substituted reading. Both thresholds are Zorowatt's estimates. Ask the client before you keep or replace them.

9. Spikes

A single value far above the site's usual top load — Zorowatt flags anything above three times the 99th percentile, an estimated threshold — or above the connection's physical capacity is usually an error. If it is real, it was billed: check it against the invoice's kW max. If it is not, a battery sized on it would be oversized, and the savings promised on it would never appear.

10. Coverage

Less than 12 months means the annual peak and the seasons are extrapolated. Zorowatt treats a dataset as good enough for a study only with 12 months and at least 90% of the intervals measured; below that the report says the result is provisional.

Then check against the invoices

The last test catches most of the others: compare each month's kWh and maximum kW from the data with the invoice for that month. A factor of 4 points at kW vs kWh, a one-month shift at the date order or the labelling, a small steady gap at missing intervals or estimated readings.

The free meter data checker runs all ten checks in your browser and tells you what it assumed — layout, interval, unit, labelling, date order — so you can see and correct it. Nothing is uploaded.

Related guides

• What to ask your client for before a battery study
• How to size a battery for peak shaving
• What a battery feasibility study contains

Sources

• Leigh Dodds, 24 different tabular formats for half-hourly energy data, 12 Jul 2021

Last checked: 2026-10-30. Every figure in this guide comes from the source linked next to it, with the retrieval date Zorowatt's research recorded (28–30 October 2026), or from arithmetic shown step by step. The thresholds for gaps, stuck zeros, flatlines, spikes and coverage are Zorowatt's own estimates, as used by its data checker.