The Coupon That Ate Bed Bath & Beyond's Best Customers

Bed Bath & Beyond’s blue 20%-off coupon became one of the most recognisable marketing assets in American retail. It also became the thing that made their customer data useless.

The coupons were mailed indiscriminately, to everyone, at roughly the same frequency. There was no segmentation by purchase history, basket size, or customer lifetime value. A customer who had spent $3,000 in the past year got the same coupon as someone who had bought one item on clearance. The result was a system that actively rewarded deal-seeking behaviour while training high-frequency buyers to postpone purchases until the discount arrived.

What Indiscriminate Discounting Does to Your Data

When every customer eventually buys on promotion, revenue data stops telling you what it is supposed to tell you. You cannot distinguish a customer who responds to a 20% discount because they are margin-sensitive from a customer who would have paid full price but now habitually waits.

This matters because the two segments have completely different CLV (customer lifetime value) profiles. The margin-sensitive shopper might be worth a coupon if the discount converts an otherwise lost sale. The habitual waiter is worth stopping, they would have bought anyway, and every coupon you send them is pure margin destruction.

Without proper segmentation, Bed Bath & Beyond could not see this distinction. Their marketing database showed millions of coupon users. It could not tell them how many of those users were trained into that behaviour by the coupons themselves.

The Attempt to Stop, Without the Data to Do It Surgically

In 2019, Bed Bath & Beyond brought in Mark Tritton as CEO, partly to modernise the business. One of his stated goals was reducing coupon dependency. It was the right call. The execution was not.

The approach was to cut coupons broadly, reducing volume, reducing frequency, pulling back across the board. There was no evidence of a targeted strategy based on customer segments. Who kept seeing coupons? Who was cut off? Without LTV segmentation, the blunt cut was the only available tool.

Same-store sales fell sharply. The conclusion drawn by the market, and internally, was that coupon-cutting hurt revenue. That is probably true in the short term. But the more important question is whether the revenue that vanished was worth keeping.

A customer who only buys when discounted and who costs you margin every visit may be generating revenue on the income statement while destroying value on the balance sheet. Without customer-level profitability data, you cannot know. Bed Bath & Beyond cut the coupons without knowing who they were cutting.

What the Data Would Have Shown

A proper LTV segmentation model on Bed Bath & Beyond’s customer base would have divided buyers into at least three buckets:

Full-price loyalists: customers who historically bought at or near full price with high frequency. These are the people a coupon program actively harms: they learn to wait, you lose margin, and you risk training them into a new behaviour that then requires ongoing incentives to maintain.

Promotion-activated buyers: customers who genuinely do not convert without a discount. For this group, the question is whether the discounted conversion is profitable enough to justify the margin sacrifice. Some will be; many will not.

Deal-seekers: customers who buy only on deep discount and rarely return between promotions. These customers may never be profitable to acquire, and a coupon program disproportionately attracts them.

The fix was never “stop all coupons.” It was segment the list, stop sending to the first group, test the discount thresholds needed to activate the second group, and phase out the third. That requires customer-level data, purchase frequency, average transaction value, channel, time between purchases, and a CLV model built on it.

The Lesson for Businesses Running Blanket Promotions

The Bed Bath & Beyond story is often told as a tale of bad execution by a new management team. The real failure happened years earlier, when the company built a revenue base on discounting without building the measurement infrastructure to understand what that discounting was actually doing.

By the time the problem was obvious enough to act on, the data infrastructure to act surgically did not exist. The only lever available was a blunt one.

Blanket promotions tell you what your promotion-influenced revenue looks like. They do not tell you what your underlying demand looks like, which customers are worth keeping, or what share of your revenue would survive without the discounts. That distinction is exactly what a customer analytics function exists to answer, and it needs to be built while the business is still growing, not when it is already declining.

Bed Bath & Beyond filed for bankruptcy in April 2023. A useful data infrastructure would not have saved the business on its own. But a business that knew the LTV of its customer segments could have made different decisions at every point in the decade before the end, about which customers to invest in, which promotions to run, and what the real cost of coupon dependency was.

The data was there. The segmentation was not.