Original data · free to cite
When Melbourne's west books carpet cleaning
Thirty-one months of one cleaner's actual job book, published in full. Including the part that made us wrong.
Everyone in this trade will tell you summer is busy and winter is quiet. Nobody shows their working. We run a one-van carpet cleaning business out of Point Cook, so our diary is the local demand curve, and we have 31 unbroken months of it, January 2024 to July 2026, 2,100+ jobs. This page publishes it, along with the two things that made our own first reading of it wrong.
Take anything here, the charts, the tables, the numbers, and use it. Attribution with a link is appreciated and not required, and you do not need to ask us first. If you want the underlying series in a spreadsheet, emailinfo@greenlioncarpetclean.com.auand we will send it.
1. The year has two peaks, not one
Every month is expressed as an index against the 31-month average, so 100 is an ordinary month. We publish the index rather than raw job counts, the shape is the finding, and the shape is what you can compare your own numbers to.
| Month | All months | Advertising held constant | Years pooled |
|---|---|---|---|
| January | 114 | 116 | 3 / 2 |
| February | 120 | 113 | 3 / 3 |
| March | 108 | 102 | 3 / 3 |
| April | 88 | 83 | 3 / 3 |
| May | 87 | 105 | 3 / 2 |
| June | 78 | 98 | 3 / 2 |
| July | 80 | 80 | 3 / 2 |
| August | 91 | 85 | 2 / 2 |
| September | 114 | 107 | 2 / 2 |
| October | 115 | 108 | 2 / 2 |
| November | 115 | 108 | 2 / 2 |
| December | 102 | 96 | 2 / 2 |
Read the solid line. Demand peaks in January at 116, holds a second, broader peak through spring, September, October and November all sit above 105, and bottoms out in July at 80. April is the other genuinely quiet month at 83. The spring peak is the part that surprised us: it is nearly as strong as summer and it lasts three months instead of two.
2. The part where we were wrong
Pool all 31 months naively and June looks like the worst month of the year, index 78, a deep midwinter hole. We believed that for a while. It is an artefact. June is the month this business twice cut its own advertising to near zero; in one of those Junes the spend was nil and it was the lowest month in the entire series. Hold advertising roughly constant and June climbs to 98, an ordinary month.
The dashed line on the chart is the same data with our own decisions still baked into it, so it is no second opinion. We have left it there deliberately, because it is exactly the shape you would get if you asked any cleaner "when is your quiet season" and they answered from their gut.
The lesson generalises past our trade: a small business's own records measure its own behaviour at least as much as its market's. April and July survive the control and are real. June did not.
3. Demand is far more local than a service area suggests
414 customers, 58 suburbs, last job between September 2024 and July 2026. Share of customers, not of jobs.
One suburb, the one we are based in, is 33.1% of the book. The top three are 52.2%. The top nine: every one of them inside the Wyndham or Hobsons Bay council areas, are 79.7%. The remaining 41 suburbs, scattered from Brunswick to Doncaster, are 11.8% between them, and most contributed a single household.
We advertise across a far wider radius than that, and the map keeps telling us the same thing: distance decides this trade. A van that can technically reach a suburb in fifty minutes still almost never gets booked there.
Method, and what this data cannot tell you
- Seasonality
- ServiceM8 · report_jobs_by_month_yoy, read 2026-08-18. 31 complete months, January 2024–July 2026. The in-progress month at time of reading is excluded: a part-month in a seasonal series invents a trough.
- Advertising control
- Google Ads API · customer spend by segments.month. The controlled series keeps only the 27 months where spend was at least A$1,500, which removes the months the account was throttled or switched off.
- Suburbs
- ServiceM8 client extract, read 2026-08-09. 414 customers with a most-recent job between September 2024 and July 2026. Suburbs with fewer than three customers are pooled, so no row can be traced toward a household.
- Not published
- Revenue, absolute monthly job counts, and anything identifying a customer. We will not send these on request either.
Four limits worth stating plainly
- n = 1 business. This is one van in one corner of one city. It is a real sample, not a representative one.
- Suburb shares are customers, not jobs. A household that books three times counts once.
- The geography reflects where we advertise and how far we drive, not only where demand exists. Read it as demand we can serve.
- Two to three years per calendar month. Every monthly figure pools two or three observations; the table shows which.
Using this
Published under CC BY 4.0: quote it, chart it, put it in a newsletter, argue with it. We would like a link back to this page where it fits naturally, and we would rather you take the numbers and disagree with our reading than not use them at all. Questions, corrections or a request for the raw series: info@greenlioncarpetclean.com.au.
Green Lion Carpet Clean · Point Cook, VIC · data read 2026-08-18. A more general fact sheet is on our press page.