Amazon has been running a new free-sample program since at least August 2018, as Axios noted today. And it may be a very smart move--or an incredibly dumb one that could surpass the now infamous intrusiveness of Facebook.

The company has run hot and cold, as any might. But Amazon's highs and lows are extreme. For example, Prime Day has been maybe the single most clever retail move since Macy's launched its Thanksgiving Day Parade. Yet recruiting an in-house Twitter army to combat negative stories about working there was a huge PR loss.

This move seems on the edge and liable to go either way.

Sampling is an old and effective marketing technique. If people like the way your product works and it's not likely to be a one-time purchase, then letting them try it increases the chance that you'll reach people who will ultimately like it. That can mean a quicker path to sales and brand loyalty.

However, there's a potential problem because of the way Amazon premiered this program--or didn't, publicly at least, to the millions and millions of consumers who were apparently automatically opted in without notification. (I have reached out to Amazon about this.) You can opt out but need to know the program exists and then go to the appropriate page on the site.

The samples program was once a paid one for Amazon Prime members. You could buy samples and then receive future purchase credits for the amount. That changed. Axios traced it back as far as a Twitter message from someone who had received a sample without expecting one. I was able to find differences on the company's sample webpage dating to October 19, 2018, using cached versions.

The page now says that samples are like "Amazon's product recommendations," which means they're based on analysis of historical data. The page also mentions that Amazon "surprises select customers with samples that we think will be delightful and helpful."

There are some big differences between Amazon's approach and traditional sampling. In the latter, consumers look at what is available and then try it. The arrangement might get them to consider something they never have before and that no one might predict on the basis of their previous-purchase histories. And people aren't forced to take things that might cause embarrassment or resentment.

By depending on data mining and then sending physical products, Amazon steps beyond recommendations--which, in my experience and other people have told me, can be wildly off. Having product show up unbidden at your doorstep based on your activity on Amazon (and any other data it might acquire for a fuller picture) could be more unnerving than seeing ads pop up on Facebook that key off messages or emails you might have sent.

And what happens if a misreading of data creates a terrible response among consumers? Here's an awful hypothetical. Parents have been purchasing products for their baby. A sudden illness causes the infant's death. And then, the following week, samples of baby formula or diapers appear in the mailbox. Or a dire financial setback means someone cannot keep up with car payments and the vehicle is repossessed shortly before an auto air freshener arrives.

Or someone just finds it incredibly disturbing to find a retailer tracking their every move and pushing products at them.

Or someone doesn't get samples and wants to know why they're penalized.

We all know that the tech industry has gone down a road of monetizing users in ways that are often creepy. Look at the drubbing Facebook has rightly taken. Or the negative attention leveled at other tech giants, like Google.

If people don't complain overly about this new program and it can stay out of the public eye, it might work well for Amazon. At least in terms of getting promotional money from advertisers. But data can be deceiving, whether in the relevancy of an ad that a website shows or in the assumption that a stratagem is wise. Facebook made both mistakes and now is in a constant battle. Amazon is hardly immune.