When it comes to machine learning, there is always a tension between collecting as much high-quality data as possible while staying within a limited budget. For machine learning algorithms that require human-annotated data, that tension is even greater due to rising labor costs.

As you can see in this graph PayScale.com put together, wages rose considerably during the fourth quarter of 2017. This continues a trend we’ve seen over the past several quarters, and it shows no signs of slowing down.

Suppliers of human-generated data can entertain only a few options to minimize the impact of increasing labor costs on our bottom line:

Labor Costs

Option 1: Maintain status quo

We can ignore the graph above and stick to our existing pay rates. The risk here is that we would lose good workers to better-paying suppliers, eventually draining our bandwidth and negatively impacting quality.  

Option 2: Increase rates

We could transfer the cost of higher wages onto our clients. It’s possible they are aware of rising labor costs, and they might even understand our need to transfer those costs to them. But we figure that’s extremely unlikely. Raising rates on our clients would be a last resort, and only after we have exhausted every other option.

Option 3: Outsource to cheaper areas

We can hire workers only in areas where labor is cheaper. For example, we shift to only hiring in states with a low minimum wage. This strategy has a several unfortunate consequences, but the most dangerous is one of biased data. By avoiding high minimum wage states, we would end up with workers concentrated in the Midwest and the South, which means that your machine learning will work great for those areas, but will underperform elsewhere.

Option 4: Do more with less

We could decrease internal overhead through layoffs, process changes and other moves that will make us try to do more with less. In striving for continuous improvement, the better suppliers are doing this already. But the process can be slow and painful.

Option 5: Invest in innovation

We could dedicate a large research and development to innovation. We could embrace technology that eliminates manual steps through the automation of repeated processes, gives our workers and internal teams more tools to do their jobs more efficiently and identify and implement alternative approaches that are more efficient.

In case it’s not obvious yet, we at AI Data Innovations have devoted ourselves to Option 5. Our dedication to innovation means we are constantly experimenting with new tech to save time and resources and improve efficiency. Our R&D team never stops exploring new ideas to improve processes and make the most of the resources we have.

This means rising labor costs don’t affect us as much as they do others. Our bandwidth and quality are minimally impacted and the service we provide to our clients is top-notch, no matter what region our workers come from, how much data the client needs, or how limited their budget is.