Businesses keep making the same mistakes. Here's one I see periodically, illustrated by a startup I consulted with briefly a few years ago. It was an app/discount card that gave people discounts at local businesses. Commonly 10% off or so. They had an online map showing where you could use their card. I knew they were doomed when I saw two florists on the map a mile apart on the same road. All their card did was spark a price war. 

Loyalty cards can be great, but there's two sides to success. 1) The business should use programs like this to gain marketing insights into their customers so they can do personalized, one-to-one marketing. Kroger does a great job of this. The more you use your Kroger loyalty card, the more you get a personalized batch of coupons for stuff you actually use. The business gives up a share of the profit in exchange for greater insight into their customers' habits, not just a hoped-for increase in sales volume. They can then use these insights to laser-focus their marketing efforts on their most profitable customers and drive better engagement, which then increases profits. 2) From the consumer side, a loyalty card is worthwhile if the more you use it, the more valuable it becomes. Chick-fil-A is a good example of this. When you start using their app, you get 10 points for every dollar you spend. Then you move up a level and start getting 11 points. Finally you can hit the top level and get 12 points per dollar. Thus the app ends up being 20% more valuable to the user. They also periodically give you a free sandwich or such. This builds on a positive reward feedback loop in the consumer's brain and thus builds brand loyalty (by playing on the consumers' sunk cost or loss aversion cognitive biases). Otherwise, the app or card just becomes a package of coupons that can be thown out without a second thought, much like bundles of physical coupons one gets in the post. 

Discount apps or cards that don't gather valuable marketing insights for the business or become more valuable to the consumer over time are, in my opinion, doomed to failure. I've seen several of these come and go and have yet to be proven wrong.