Denouement
blog.staffjoy.com
blog.staffjoy.com
> Control money to make money. In workforce management, I now realize that controlling payroll data is key. It’s tough to quantify the creation of value otherwise.
My (fuzzy) interpretation: since we weren't integrated with the payroll/hr/accounting systems, we could not prove the efficiency gains our product was meant to bring with hard numbers.
On the one hand I agree that integration with payroll systems is probably critical in this space.
On the other hand, if your product brings a 10% efficiency improvement (which is still huge, not discounting that), it's still going to be hard to prove that, even with access to raw payroll data. Raw payroll data is probably too noisy.
Our V1 timeclock product garnered a lot of interest because it directly led to people getting paid, and it integrated with scheduling to make it easy to tell whether records were correct. I now understand that many more established players, like Kronos, started with payroll as a core feature, then diversified to scheduling.
How this lesson applies to my newco Moonlight - we're charging a transaction fee and managing payments, rather than doing something like a SaaS subscription model.
Anyway, first rate blog post and handling of a difficult situation!
There are two categories:
Both are multiple good/service providers
Both are multiple goods/service recipients
I think the difference for something like Uber and Airbnb is that they embrace chaos. It is better for them to have someone take a different Uber or stay at a different Airbnb everytime.
The problem I think it's that you have to be sleazy enough to pull this off. You have to be like Airbnb with its promotion of instant booking going side by side its promise of community and kumbaya blah blah. I trust I don't have to say anything bad about Uber. Plenty of ink had been spilled already.
You want a constant stream (or late enough pool) of work and of workers.
Here's an idea to test: do employees react favorably to "quantization" of work?
Example: we think our conversion sucks because blah blah. We want to create a new set of landing pages and develop a way to show it 10% of the time. We want a whole new team to do it so it doesn't distract of full time staff from the fires they're putting out with the real application.
Just spit balling... Would love to hear what everyone thinks about it.
Thing is, it was part of a comprehensive system that handled accounting, inventory, and payroll as well as scheduling.
If we didn't have the integration with payroll, to know who the employees were and their status, it would hardly have been worth using a stand-alone scheduler if that meant maintaining a shadow copy of the employee file.
Good, cheap (and let's be honest, gullible) technical talent is hard to find.
I'm more proud of the Autoscheduler algorithm in Julia, which uses a bunch of different models in a kind of Dynamic Programming approach. Check out its README for an explanation: