3 situations for Machine Learning in your organization
Machine Learning has been the craze for the past few years but many companies are starting to question it’s value as they aren’t getting the ROI they were expecting. In this blog I cover three situations where Machine Learning and Deep Learning might actually be useful in your organization. Please do keep in mind that I’m a bit skeptic of Machine Learning (ML) at the moment and may be somewhat biased against it. Let’s dive in!
- Your organization is very large
If your organization has thousands of employees then it is a good indicator that you have enough internal facing data to build robust ML models with, These could include regression (predictive analytics), Churn models, Customer Segmentation models, if you’re really big then Deep Learning may be a possibility. That being said, I still wouldn’t advise that you blindly hop on the ML train. Before doing any ML or hiring anybody you need to make sure your data infrastructure and pipelines are fully set up, I would also advise that you do some descriptive analytics before diving in. Lastly, I want to stress that you need to go in with a problem first mindset. Remember, your goal is to solve your organization’s problems not implement ML. Think How do we fix the high drop-off rate from ‘add to cart’ to ‘checkout’? Not How do we implement Machine Learning?
Example: Klaviyo is a growing Advertising Platform for eCommerce brands. they’re approaching one thousand employees and have 265,000 customers according to their LinkedIn. They had a data science team, for the next reason I’ll talk about, but their Business Intelligence (BI) team handled their internal facing data and did mostly descriptive analytics until very recently when the Data Scientists started doing their first projects with the internal data. I suspect the reason for this is that they’re internal data is finally large enough to benefit from Predictive Analytics beyond the scope of BI.
2. Machine Learning is part of the product/service
Needless to say if ML is core to the actual product or service your organization provides than ML / ML expertise is a must. Maybe you’re adding a facial recognition feature to your app, or you build highly personalized chatbots, or you’re trying to improve Alexa’s performance. Yeah Jeff Bezos I’m talking to you, although I highly doubt you’re reading! I will again stress that you must solve real business problems and have that problem first mindset I talked about earlier.
Example: Using Klaviyo again, they provide in-depth ad analytics to eCommerce brands. They have identified beneficial use cases for ML, presumably predictive analytics of some kind, to add to the service they provide. Because of this they’ve had Data Scientists work on their external facing data (on their service aka the eCommerce client’s data) long before their internal data.
3. You’re feeling adventurous and feel like giving it a shot
Maybe you’ve heard about the latest in this field and can identify a potential use case in your life or business where it’s possible, not saying it’ll solve your problems but could contribute to what you provide. Though I wouldn’t advise doing this for the sake of profit, you could try it for exploration and experimentation purposes, especially if your a very small organization or one person team allowing you to move fast. Think of as a mini R&D investment, in the same way Google invests in AI research you could invest in AI implementation at a much lower scale. I think this is a good way of looking at it because you shouldn’t expect any ROI in the near future and could quite possibly send money down the drain. This is not very common for business owners, and therefore why I had to make up a fake example :)
Example : Jared is a realtor and he’s heard about really advanced Chatbots made with GPT2. He wants to try adding one of these Chatbots to his website and see if it see if it can help him book more calls, aka potential clients. He finds a Computer Science student that’s passionate about AI to build the chatbot. At the end of the day Jared thinks it’s cool and is impressed by it’s performance and because he didn’t expect anything going in, the actual outcome doesn’t matter to him (unless the chatbot actually hurts his booking rate).
A Final Word
Despite the ML buzz, most organizations don’t actually need to think about using it. Most companies can do fine with a well organized data Infrastructure and some descriptive analytics. I would say that companies should emphasize on being ready to implement ML rather than actually implementing it! That being said If you are an AI company, You’re product/service involves a well thought out case for ML, or a very large organization with a thought out internal ML project then I would highly encourage ML initiatives. Feel free Share your thoughts and connect on LinkedIn!
