Saturday, February 9, 2019

What the Lumen5 text to video service taught me about...text

Last month, Sheila Scarborough wrote about repurposing content. One of her suggestions was to use Lumen5 to convert a blog post to video.

Lumen5 crawls the blog posts and makes a short draft video with key phrases and matching stock imagery. All we have to do is tweak the draft, maybe change some of the phrases or images (machines are not THAT smart!) and boom – we have a new video.

Good old artificial intelligence, or A.I., is supposed to make this process easy.

So I thought I'd try it.

I selected one of my more popular Empoprise-BI posts from 2018: Before Blogger is sunsetted, I'd better write this post on Google's customers.

When I loaded it into Lumen5, the A.I. went haywire.

And I can understand why.

My blog post writing style is notoriously dense. So Lumen5 had trouble automatically selecting portions of text to highlight in a video. When I tried to select text manually, even I had trouble.

Just for fun, take a couple of minutes to read my original post.

Now watch the video.

If I am going to convert text posts to video...

...they need more streamlined writing.

(Let's see if I did it.)

Friday, February 1, 2019

Why you shouldn't be pro-choice when it comes to product feature prioritization

I spent a decade as a product manager back in the day (way back - it was waterfall time), and I've interacted with product managers ever since. I think I can safely say that every single product manager has to deal with feature prioritization - namely, that there are more features that need to be added to the product than developer hours available to implement them all.

Once you get past the bad methods of prioritizing features (for example, prioritizing feature X because the development engineer thinks it's really cool and reminds her of The Matrix), you realize that you need to engage in some type of market analysis. Now I've worked with very small markets, where the total number of customers for my products can be counted in the hundreds. The snack chip makers who will be advertising in this Sunday's Super Bowl (there, I said it) deal with a somewhat larger number of customers.

But how do you get meaningful data from customers? After all, customers are focused upon their own needs, which in some respects resemble a 1960s protest rally.

By Bundesarchiv, Bild 183-1989-1106-405 / CC-BY-SA 3.0, CC BY-SA 3.0 de, Link

What features do we want?
When do we want them?

In a universe with unlimited choices, people will choose all the things.

There are other issues with such surveys, as Jared Boyer of Price Intelligently notes:

Asking respondents to “check all the features that are important to you” can generate bias since respondents can easily be swayed by the concept that “more is better.”

But what if people are REALLY forced to choose? Here's how Boyer recommends that this be done:

[F]orcing respondents to simply choose a favorite feature and least favorite feature gives you a better opportunity to hone in on the aspects of your offering that are most attractive to customers.

Yes, Boyer advocates that you choose a SINGLE favorite feature and a SINGLE least favorite feature. In the first case, choose the ONE feature that you're not willing to give up, bearing in mind that you'll give up everything else. In the second case, choose the feature that you may NEVER EVER get.

Now THAT will make you think.

Even if you're not a product manager, or don't associate with those types of people, you can play this game for yourself. Choose your favorite application - Microsoft Word, Facebook, Angry Birds, whatever. Think about your favorite and least favorite features. Really think. Now imagine that 41 million 18-24 year olds in the United States are doing the same thing.

You can get a lot of data from that.