AI and data science in offline businesss

The article will slightly open the curtain and show the most likely near future of many offline businesses. I’ll immediately tell you what I’ll stand on: in the near future, conducting AI-free offline marketing will be romanticized like something pristine — like coffee without sugar, like a face without makeup, like a notebook and a pen. But smart retail executives are always interested in understanding the realistic “possibility space” of a technology’s applications – and that’s what this article was hoped to be about.

I won’t take it complicated.

Neural networks have been used in offline business for years.

Food establishments customize the recognition of visitors to personalize the service and recommendation functionality. Not so long ago there was an interesting conflict between the father of one girl and Target. Stores recommendation system started advising pregnancy-related products to family, which could not but embarrass or pity the father.

“She’s still in high school, and you’re sending her coupons for baby clothes and cribs? Are you trying to encourage her to get pregnant?”

–a few days later–

“I had a talk with my daughter. It turns out there’s been some activities in my house I haven’t been completely aware of. She’s due in August. I owe you an apology.”

Nike recently acquired a startup, with whom they developed a system for determining a specific shoe model size that best suits your needs. Of course, using machine learning (ML). To implement the idea, it required only 6 months of work. Now they have not only an advanced feature but also a streamlined process of its improvement: the application automatically addresses developers data on all cases of return with all the details of the corresponding order.

Amazon also recently made a big step in the use of ML, opening the world’s first local store without consultants and even cashiers. Moreover, you don’t even need to queue up at a self-service cash desk – both customers and their purchases are recognized by dozens of cameras and sensors installed in the store. As the buyer exits from the store, the corresponding amount of funds is got debited from his account.

It is worth mentioning that it was also machine learning that helped determine which products should be put up for sale in the offline store?

Naturally, all the giants of this world, from corporations to government agencies, use machine learning for a variety of purposes.

Instead of having a classification of a thousand general categories of dogs and leopards, you can actually just have five categories of the five levels of diabetic retinopathy and train the model on eye images and an assessment of what the score should be. And if you do that, you can actually get the images labeled by several ophthalmologists, six or seven, so that you reduce the variance that you already see between ophthalmologists assessing the same image.

And if you do that, then you can essentially get a model that is on par or slightly better than the average board-certified ophthalmologist that’s set at doing this task, which is great.

If you were maybe a company that doesn’t have a lot of machine learning researchers or machine learning engineers yourselves, you can actually just take a bunch of images in and categories of things you want to do – maybe you have pictures from your assembly line. You want to predict what part is this image of. You can actually get a high-quality model for that using Cloud AutoML.

Since 2009, we’ve actually been growing the number of ML-related papers posted at a really fast exponential rate, actually faster than the Moore’s Law growth rate of computational power that we got so nice and used to for 40 years but it’s now slowed down. So we’ve replaced the nice growth in computing performance with growth in people generating ideas, which is nice.

There is also a study of student performance in two Belgian schools showed that students who are “in a relationship” have, on average, 15-20% better school performance.

Use cases enumerating can be infinite. The very use of artificial intelligence in retail and eCommerce has already become a topic for scientific research.

Actually, you can go endlessly deep into AI use cases in retail examples. But instead of admiring the consequences let’s better understand the causes.

Online-to-Offline Trends for AI

One of the main reasons for the popularization of the use of neural networks in offline businesses is the reverse regradation to offline businesses. Returning to the examples of romanticization, it was worth including the offline business itself in this list.

It is no longer fashionable to own an offline business without the website. Personalization of content from search engines and stores to social networks and blogs (the boundaries between which are gradually blurred) completely kills the principle of word of mouth. Surely your friend with similar interests has also heard the “same” song, saw the “same” news, is also thinking about buying the “same” phone, which he got recommended by hellova “same” site.

It would seem that a variant with a niche product always arises. But there, as a matter of fact, we usually do not talk about large amounts of information that are needed for feeding to neural networks. So there are carriers of the “old school” of business, which insist on the power of charisma and intelligence. Warren Buffett alone is worth a lot (well, some less after 2018).

Following this trend, more martech providers who play in data will open up the visibility of their databases and technology via client data, audience and artificial intelligence insights to help marketers understand what data is in use, how it’s being leveraged and how it’s working.

A number of principles can be transferred from online business: (It’s funny that 10 years ago we thought which aspects of the offline business could be transferred to online)

  • Market segment prediction / assessment
  • Prediction / evaluation of the target audience
  • Prediction / evaluation of the outcome of a marketing company
  • Personalization service – Hence, based on the varied expectations and service factors, one can predict a customer’s reaction.
  • Definition of real business leverage

Returning to the question of the workflow robotization. Don’t be worried, your employees will not rebel [too much]. The implementation of ML is also worth the time. Yes, if your company has more than 2-3 people, then you will have to reorganize workflow. But the list of necessary qualifications will not change much anyway – the secret to the success of any ML model lies in the correctly collected data and the successfully chosen learning algorithm.

So you only need to teach your analysts to write reports not in notebooks but in Google Sheets.

Other machine learning pros:

  • ML is good for professionals cause it allows you to optimize already debugged workflows, to distribute free resources in the best way;
  • ML is good for beginners because it gives you the opportunity to get into a new niche basis and risks by only collecting and ML-ing some data;
  • for teaching the model, you can use the experience gained earlier;
  • having a trained model, it is possible to obtain predictions of the outcome of certain decisions (preferably, countable);
  • Machine Learning makes it possible to identify the most unexpected (or at least interesting) patterns.

Still not without problems.

As already mentioned, the raw data is the root of success. Frankly, if the date is correctly compiled, but there is not enough of it, then it is too risky to rely on the predictions of the model built on it.

And if the business is young, and it has not yet collected an impressive amount of information about sales and strategies?

Sometimes some geniuses will create some algorithms that somehow can deal well with some small amounts of some data, but even though it wouldn’t be scalable down to solo-baking-cookies-at-home businesses. So we still will need a lot of data aggregation to be able to implement machine learning in most of (very) small offline businesses.

Other machine learning cons:

  • learning/retraining hours;
  • the need for a more detailed analysis of the activities;
  • handling huge volumes of data sucks up a lot of computer power, which you will need to invest in;
  • vendors will often try to make money by blaming themselves a machine learning solutions prophets, when in fact they are only capitalizing on the AI hype;
  • close, long-term sales relationships problem: A smooth salesperson who sends Christmas cards to the purchasing manager every year may not be able to win business with a great “business relationship” if the tangible ROI isn’t there for the retailer;
  • the endless need for maintenance||high cost spent on developing self-maintained systems;
  • taking things globally, machine learning will make it much easier for privacy invasion and hacking, autonomous wars, and prop authoritarian rulers while at the same time being able to create political instability.

Further,

marketers will gain a more nuanced understanding of how these methodologies align with business requirements including compliance, transparency, cost, goals and timing.

Leave a Reply

Your email address will not be published. Required fields are marked *