November 11, 2012

Psychic water bills

I’ve just received a water bill which, among other information, estimates my average daily water use for November.  That’s a pretty good trick for something that must have been mailed in the first few days of the month. They mean October, I assume.

Apart from their off-by-one labelling of months, Watercare are interesting because of their ‘estimated’ water usage.  They recently changed to sending monthly bills, but they still only try to read the meter ever second month, and in my case have failed to find it twice.  My first bill on moving in was for three months, and it was relatively high. I fixed the leaky seal in the toilet and expected the bills to go down.  The following month, the meter wasn’t read but my estimated daily water use went up about 7%.  The next month, again there was no meter reading, and the estimated daily use was another 10% higher. The next month the estimated daily use was down about 8%, again with no reading.

I can see why the estimated total usage would fluctuate based on the varying time between estimates, but it’s hard to see what basis Watercare had for estimating I was using more water in September (actually August) than in August (actually July) without any actual data.  I wouldn’t have expected the average Aucklander to use more water in winter, and a research report from Branz confirms my expectation.

This month, now they have found the meter, the estimated use has fallen about 85%, catching up on three months of overbilling.

avatar

Thomas Lumley (@tslumley) is Professor of Biostatistics at the University of Auckland. His research interests include semiparametric models, survey sampling, statistical computing, foundations of statistics, and whatever methodological problems his medical collaborators come up with. He also blogs at Biased and Inefficient See all posts by Thomas Lumley »

Comments

  • avatar

    I presume they’ve changed to monthly billing for better cashflow but it is a pain, esp when combined with optimistic estimates.

    11 years ago

    • avatar
      Thomas Lumley

      To give them credit, I think it was to improve their customers’ cashflow — people were getting unpleasantly surprised by three-month bills.

      11 years ago

  • avatar

    I’m always intrigued by their estimate of the number of people in your household as shown in their graph. They always have the wrong number of people in the graph and I’d love to be able to tell them the correct number to see how we compare versus other families of the same size.

    It normally shows us as a little above average for a 2 person household (but we’re a 4 person household). They did some estimates which were way out because we were out of the country (including a <$10 bill!), and then we got a big bill correcting for their mistakes and showing our usage as above average for a six person household!

    11 years ago

    • avatar
      Thomas Lumley

      You can tell those numbers of people are fairly bogus, because the ranges don’t overlap at all. It can’t really make sense that the highest-use 4-person households use less than the lowest-use 5-person households.

      It looks like they obtained averages for each number of households and then just split up the spaces between them. In fact, the upper endpoints of the intervals form a perfect quadratic, so they probably were from a regression model rather than simple averages.

      11 years ago

  • avatar
    Martin Kealey

    Like the power companies, Watercare are not totally devoid of data on which hto make estimates on a new account: they have the usage history for the previous occupants.

    They have aggregate outflows from reservoirs and treatment plants.

    And in theory they could also use weather data to estimate the likelihood of activities like “washing the car” — though somehow I doubt they bother.

    Do you have an opinion on whether these are (1) fair estimators and (2) no more likely to penalize the consumer than “simpler” methods of estimating?

    11 years ago

    • avatar
      Thomas Lumley

      Basing it on past usage and seasonal aggregate demand would make sense, and anything that increases accuracy of estimation should be good for their customers because it would reduce random month-to-month fluctuations in billing. I don’t think these factors can explain what they are doing, though.

      Seasonal aggregate demand should not be that much higher in September than June, and the only evidence I’ve been able to find supports this view. Past usage can’t justify changes between months for a completely new account — I was a renter before, so I’ve never been in Watercare’s system.

      11 years ago

  • avatar
    Watercare Insider

    The way Watercare calculates estimated bills is very simple:

    (average daily consumption based on last two consecutive actual readings) x (days in the estimation period) x 90 %

    The result is rounded up, or down, to the nearest kilo-liter (aka 1,000 liters or kL).

    Example:

    • if actual reading 1 is: 100 kL for 29 days

    • if actual reading 2 is: 140 kL for 31 days

    • if estimation period is: 30 days

    –> average daily consumption for reading 1 and 2 combined = (100 + 140) / (29 +31) = 4 kL

    –> estimated reading for 30 days = 4 x 30 x 0.9 = 108 kL

    So basically Watercare is looking at the last two monthly ‘true’ consumptions (i.e. when a Meter Reader came to read the meter), and charge 90% of the same average for the estimated period.

    Of course, there are exceptional situations that might explain why your estimated bills do not match this formula.

    For example:

    – prior to 1 July 2012, meters were read every three months, or even every six months. So the estimated bills sent from July this year have been using very old data set. As Watercare reads water meters every two months now, the estimated bills will be more and more accurate as time goes

    – when a Meter Reader cannot read the meter due to a locked gate, dog on property, meter covered by car, boat, etc., then no actual read is recorded for this month, obviously. This has two consequences:
    1. the current bill has to be manually estimated; and
    2. the next bill will be an estimation too, because that is how it was supposed to be in the first place. So you end up with two estimated bills in a row.
    The lack of a recent actual read means that those two estimated bills have to be based on quite old readings, hence possible discrepancies with annual consumptions.

    – a previous actual reading might have been wrongly recorded (e.g. Meter Reader inverting digits, or reading wrong meter), or a leak at the property generated high readings that were used for estimation later on because not recorded as ‘anomalies’

    Once again, the above examples are exceptions, and Watercare is getting better and better at calculating estimated bills, thanks to actual readings coming in every two months.

    If you receive an estimated bill that is way too high or too low, simply read your meter, call Watercare on (09) 442 2222 and give the reading: if the difference is significant, you might get a rebill / credit.

    11 years ago

    • avatar
      Thomas Lumley

      Yes, that’s what the Watercare documentation says, and it’s a sensible approach, but as far as I can see this would imply the daily use estimates would not change until there was an actual reading. This conflicts with my bill.

      It doesn’t actually bother me, since I knew it was all going to work out eventually, I’m just interested in why the pattern of estimates was so strange.

      11 years ago

      • avatar
        Watercare Insider

        Yes, the pattern seems strange, as one would assume that – if there is no actual reading taken by a Meter Reader on a particular month – then all estimates that follow should be identical to the previous one.

        For example, if a consumption is estimated at 12 kL in August, and if no actual reading took place the following month, then one would expect both the Sept manual estimate AND the Oct auto-estimate to be 12 kL as well. Not 13 kL or 11 kL.

        However there are two factors that could make estimates slightly different from one month to another:

        – the estimated PERIOD: if you check your invoices, you will see that the estimated periods are never exactly the same, e.g. 32 vs. 30 days. That is enough to give different estimated consumptions, even if they are based on the same daily average usage.

        – manual estimations are calculated by HUMANS: as noted above, if a scheduled actual reading does not take place, then a Watercare agent has to manually calculate an estimated consumption for that month. Estimation software are programmed to follow a set of rules, e.g. “meter readings used for an estimation calculation have to be at least 30 days apart” (up to 10 rules sometimes apply), but human operators might use a simpler set of rules to make calculation quicker for them, resulting in an estimation that could be different from one calculated by the computer previously.

        As you rightly pointed out, as soon as an actual read is taken by a Meter Reader all those estimations are ‘washed out’, i.e. the customer is billed for the correct amount over two months period.

        11 years ago