Products You May Like
By Huy Hoang, Founder, Simply Maid
Most founders who talk about “distributed teams” mean people on Slack in different time zones. My version is stranger. Simply Maid delivered its first clean in Crows Nest, Sydney, in August 2015, and for years we did one thing:house cleaning in Sydney. Ten years on we operate across seven Australian cities, and almost nobody who works for us has ever been in the same room as a colleague. A cleaner arrives at a stranger’s home alone, does two to four hours of skilled physical work with no supervisor within fifty kilometres, and leaves. Then does it again the next day at a different address.
That structure forces you to answer questions most businesses can dodge. How do you hire well when you cannot watch someone work? How do you know the job was done properly when the only witnesses are the cleaner and the customer? Here is what a decade of getting it wrong, and occasionally right, has taught me.
Recruit for the job that actually exists
Early on I hired for cleaning skill. That was a mistake, because cleaning is teachable and the job is only partly cleaning. The real job is turning up on time, alone, to a home you have never seen, reading a customer’s unspoken expectations, and handling whatever is behind the door with grace. So the filters that matter are reliability history, communication and judgement. Police checks are the floor, not the bar.
The strongest signal we have found is behaviour in the first fortnight. Someone who confirms promptly, arrives early to the first three bookings and messages when something is off will almost always still be with us a year later. Someone brilliant with a mop who goes quiet for a day will not. We now treat those two weeks as the real interview.
Reliability is the product
Customers rarely churn because a skirting board was missed. They churn because nobody showed up. When every employee works alone, a single no-show is not a staffing hiccup; it is the entire brand failing in one household. So we measure reliability before quality: on-time arrival, completion rate, responsiveness to schedule changes, and whether a customer asks for that person again. Rebooking a named cleaner is the most honest metric we have, because it is the customer voting with their calendar.
One structural decision helped more than any policy, and we learned it the hard way when we expanded intohouse cleaning in Melbourneand tried to run a new city like a dispatch queue. We now match cleaners to regular customers and protect that pairing. The cleaner gets predictable income and a home they know; the customer gets someone who remembers where the spare vacuum bags live. When a pairing is stable, the reliability numbers look after themselves. When we break pairings to chase scheduling efficiency, the numbers fall within weeks.
Quality control without a supervisor
You cannot inspect your way to quality when nobody is there to inspect. What you can do is design the feedback loop so it closes fast. Every clean is rated by the customer, and a low rating triggers a re-clean, at our cost, within days. The stakes are highest inend-of-lease cleaning, where a tenant’s bond depends on a property manager’s inspection, so that is where the guarantee earns its keep. It is not marketing, it is our quality control department. It gives customers a reason to tell us the truth rather than quietly leaving, and it gives cleaners an unambiguous standard to work to.
The other half is payment design. Bookings place a hold on the customer’s card rather than a charge, which only settles once the clean is complete. It sounds like a billing detail, but it changes the psychology on both sides: the customer is not paying for a promise, and the cleaner knows the money follows the work.
Retention is mostly about pay and respect. We run a cleaner-first pay philosophy: the person doing the physical work takes the largest share of every job, and we build the business model around what is left, not the other way round. People who work alone notice very quickly whether the company treats them as the point of the business or as an input cost.
The 2026 rebuild, and what I got wrong
This year I rebuilt our entire operating platform from scratch. I am not a developer and we had no engineering team. I directed AI coding tools and shipped a working platform in roughly six months, including migrating around 800,000 historical records from the system we had outgrown. I am proud of that. I also made mistakes that cost weeks.
The biggest was carrying the old data model across almost wholesale. A decade of organic growth had left us with about 120 database tables, many of them duplicates or leftovers from forgotten experiments. Instead of designing the model the business actually needed, I migrated the mess and spent months untangling it. Roughly a third of those tables turned out to be removable.
The second was inconsistency in small things: money fields named three different ways, two spellings of longitude, duplicate created-date columns. Trivial individually, until a rename in one table silently broke the page our operations team used to see which cleaners were available, in production, on a weekday. When every employee works alone, the platform is the only colleague they have. When it goes down, so does everyone.
The lesson generalises beyond software. Distributed workforces do not tolerate ambiguity. Whatever is unclear in your systems, your pay structure or your standards will be resolved by a thousand individuals making a thousand private decisions, and you will only find out through the ratings. The job of the founder is to remove that ambiguity before it reaches the front door.
About the author
Huy Hoang is the founder ofSimply Maid, an Australian home-cleaning company founded in Sydney in 2015 that now operates across seven cities. A non-technical founder, he rebuilt the company’s booking and operations platform in 2026 using AI development tools. His work has been featured in the Sydney Morning Herald, Domain, Mental Floss and House Digest.