An age-old question that keeps coming up in the service industry is, “How should one charge clients?”
Most agencies or companies that provide services will, at some point, consider between charging by unit price or charging by hour. In the machine learning industry, it’s no different. It’s a question that typically comes up at the start of discussions with clients.
The key thing to note here is that different artificial intelligence (AI) companies will be at different states of the machine learning (ML) journey. Some companies may have newly adopted AI, while others may have already implemented ML processes into some of their workflows.
One of Supahands’ USPs is that its services are highly customizable and this even includes pricing. With Supahands, it’s possible to opt for either unit rates or hourly rates.
Depending on the client, there are pros and cons to either form of pricing. This article will shed more light on that, going into detail on how clients can choose the best method of pricing for their specific ML project.
Pricing by unit rate
If pricing is conducted based on unit rate, the entire project will be divided into segments or blocks ie. units, and each unit will have a specific price affixed to it. In the case of data labeling projects, one unit could be one bounding box.
At the end of the project, the total amount paid will be equivalent to the price per unit multiplied by the number of units completed. In most cases, the exact number of units required to complete the project is not known.
This can be good for clients because the pricing structure can be more transparent and straightforward. There’s no need to worry about tracking the hours spent working on the project or worry about whether the time is being recorded correctly.
At the same time, since it’s not bound by time, it reduces the risk of money wasted on underperforming labelers, as well as provides more clarity and predictability for budget allocation.
This also means that the pricing will be more tailored according to the complexity of the task and there’s less need to worry about the throughput of individual labelers. This method of pricing works if there’s a need to focus on accuracy, rather than speed.
On the other hand, what this also means is that even if the labelers get more efficient, there won’t be any change to the charges. It can also be difficult to calculate unit rate if there’s no clarity on the overall project requirements or if not enough is known about the volume of work required.
For Supahands, charging by unit rate is preferable as it makes it easier to manage the cost of delivering the project. This also allows Supahands to measure the performance of each labeler by efficiency and accuracy, rather than how long they spent on the task.
However, since more scoping is required to set the price per unit, it can take a longer period of time to get a quotation for the project.
Pricing by hourly rate
An hourly rate is the price for every hour worked. For example, if your lawyer’s hourly rates are RM500 and you have him for a two-hour meeting, you’ll pay RM1000 for that meeting.
Hourly rates are something that most people are familiar with, which means that it’s typically easier to understand how it works especially when it comes to human resources.
This makes it much easier to kickstart projects as the client wouldn’t have to wait for the project’s unit rates to be calculated and determined. All a client would have to do is give instructions to start work and the labeling team would be able to get right to it.
The other plus point is that as the labelers become more accustomed to the work, they could become more efficient. As they get faster and more accurate, the cost of the project goes down.
As with the unit rate pricing method, there are some cons to hourly rate pricing as well. A client may end up paying more if the labelers work slowly, which may be required to reach good accuracy levels. Or, it could result in a situation where speed becomes the main focus, rather than accuracy. This isn’t ideal for ML deployment.
With this pricing method, it also makes it slightly more difficult to identify and manage each labeler’s quality and thus, doesn’t achieve the ideal level of transparency that Supahands desires.
How to select a pricing method?
There are pros and cons to each method and selecting the best option would depend on what stage in the machine learning journey a company is at.
Going for unit rate pricing would work better if your solution is more mature and you know exactly what needs to be done or if you’re sure that there isn’t much variation or subjectivity for the labeling work.
The other reasons you might prefer unit rate pricing is if you want to prioritize accuracy over speed or you don’t want to see too much fluctuation in how much you end up paying.
However, if you want to kick your project off as quickly as possible and aren’t sure about the expected volume of work yet, you might want to opt for hourly rate pricing. If you also prefer to have fluctuating price ie. pay more at the start and potentially have a reduced rate later on, hourly rate pricing is also the better option.
Start a conversation
When it comes to working with Supahands, it always starts with a conversation. Working with a data labeling company like ours is more than just outsourcing your work. You’re gaining a partner with whom you can discuss options, solutions and pricing structures that work best for each and every one of your projects.