Do discounts work? When to run an offer and when not to

A discount works when it brings in people who wouldn't have come and the margin you keep covers the cut. If only your usual customers take it, you lose money. Before you launch an offer, work out how many extra sales you need to break even, give it an end date, and limit it to a slow time slot or one specific product.

In short

• A 20% discount doesn't need 20% more sales. It needs a lot more: with a 60% margin, 50% more. The table below gives you your own number.
• The little data there is (one big chain, one study on purchase intentions) suggests an offer tied to a time slot moves people in that slot, and that getting customers used to discounts can lower what they are willing to pay later.
• Before cutting prices, try a plan, a combo or a reward for coming back. If you do discount, give it an end date, a limit and a three-week test.

What does the evidence say about discounts?

There are no independent studies on discounts in cafés or small neighbourhood businesses. What exists is two hints of a different size.

An offer tied to a time slot moves people in that slot. In October 2019 Starbucks tested a Thursday "happy hour" in the US from 2pm to 7pm: buy one grande or larger handcrafted drink, get another free. An inMarket study measured 11% more foot traffic in that window (15.6% more at 6pm), comparing the four weeks before with the four weeks of the promotion; outside the window traffic barely changed (+0.24%) (Restaurant Dive). The article gives no sales or margin figures: it counts people who walked in, not money that was left over. And it is a huge chain running a buy-one-get-one-free offer, nothing like your street.

Discounts can change what customers think your product is worth. A 2018 study in Die Unternehmung found that people who had previously seen a price promotion showed lower purchase intentions in a period without a promotion than people who hadn't. The authors explain it with two mechanisms: lower willingness to pay and a lower reference price (Bambauer-Sachse and Massera, 2018). It is a study of purchase intentions, not of real café sales, so take it as a warning, not a rule.

What we take from both: a discount is a tool for a specific problem (an empty time slot, a product left over), not a permanent state.

How many extra sales do I need for a discount to pay off?

More than it seems. When you cut the price you lose margin on every sale, including the ones from people who would have come anyway. The sum to break even is:

Extra sales needed = discount ÷ (margin − discount)

Here "margin" is the part of the price you keep after paying only what it costs to make what you sell (coffee, milk, cup, the raw materials of the service). Leave out rent and fixed salaries: they don't change if you sell one more. If you don't know yours, working out what each product costs to make is the first step; meanwhile, use the table with three different margins.

Extra sales needed in that slot to earn the same as without the discount (our own calculation, it is arithmetic, not market data):

"Doesn't work" means that with that margin a 50% cut leaves you with no profit on each sale: it doesn't matter how many people come.

What does it look like with a real coffee?

An example with made-up numbers, only to illustrate: a latte at €2.20 that costs you €0.65 to make (coffee, milk, cup). You keep €1.55 per coffee, a 70% margin.

With a 20% discount and 10 normal coffees in that slot, you need 4 new customers just to end up where you started. With buy-one-get-one-free you need more than three times the coffees (and you don't have that many tables). Those are the sums almost nobody does before putting the sign up.

And if the tables are empty anyway?

That is the best argument for an offer in a slow hour: if the place is empty and staff are already paid, one extra customer adds their margin almost clean. But the sum above still holds: the risk is that people who already came at that hour, or at another one, "move" into the cheap slot, and then all you did was cut the price.

When does an offer make sense?

• A specific, empty time slot. You know which one because you measured it (if not, start with 12 ideas to fill your café's afternoons, which explains how to track it).
• A product that would go to waste. The day's pastries, the last cake, fruit that won't last. Discounting what you were going to throw away costs less than it looks. Spain's retail law even leaves prices applied to cut food waste out of the "previous price" calculation (see below).
• To launch something. A new product or service you want people to try.
• With an end date and a limit. "Until 30 November" or "the first 20". An offer with no end becomes your price.

When is it better not to?

• When you don't know your margin. First find out what each thing costs to make.
• When your problem is few new customers, not few visits from the usual ones. A discount won't give you visibility: that is fixed on your Google profile and with word of mouth.
• For everyone, all day. It is the most expensive way: you cut the price for everyone who would have paid in full.
• When your margin is low. With a 50% margin, a 30% discount asks you to sell more than double.
• When it repeats every week. Customers learn to wait for the offer.

What can I do instead of cutting the price?

A discount is a gift to everyone. The alternatives give something specific in exchange for something specific:

A reward for coming back is the most honest of the five: it doesn't lower the price of the first coffee, it rewards people who already chose you and, if you use a digital loyalty card (or a paper one), it shows you how often each one returns. One honest note: stamping with a phone takes a few seconds longer than a paper stamp.

What does the law say if I advertise a reduced price?

If you advertise a reduced price and compare it with the previous one, article 20 of Spain's Law 7/1996 on retail trade says the reference price is the lowest you applied to identical products in the previous thirty days. Prices applied to cut food waste on products close to their expiry or best-before date don't count, and the rule doesn't apply to a product put on sale for the first time (consolidated text in the BOE, in Spanish). So raising the price the week before in order to "cut" it afterwards is not an option. This is general information, not legal advice: if in doubt, ask your accountant or your town hall's consumer office.

How do I test an offer without betting the shop?

Copy this into a notebook or a note:

Offer test (3 weeks)

1. Time slot or product: _______ (the emptiest slot, or what's left over)
2. My margin per unit: _______ € (price − cost to make)
3. Offer: _______ (discount, combo or reward)
4. Sales needed to break even: _______ (use the table above)
5. End date and limit: _______
6. Baseline: tickets in that slot over the previous 2 weeks = _______
7. Result at 3 weeks: tickets = _______ · total margin = _______ € · new customers or the usual ones?
8. Decision: repeat · change · drop

Look at line 7: compare the slot's total margin, not just how many people came. And note the weather (rain, heat, holidays), which moves an afternoon more than any sign.

Frequently asked questions

What discount is reasonable?

It depends on your margin, not on a magic number. Look at the table: the higher the discount and the lower the margin, the more extra sales you need. If you can't say how many extra sales you expect, start with a small offer or a combo.

Is a buy-one-get-one-free a good idea?

Rarely for a small business: it equals a 50% discount. With a 70% margin you need to multiply that slot's sales by more than three. A second coffee at half price only for people who come with someone costs less.

Do discounts make customers get used to paying less?

There are signs it can happen: a 2018 study on purchase intentions saw it after people had seen a promotion. We found no studies on cafés, so the prudent thing is to keep the offer one-off, with an end date.

How do I know if it worked?

Compare the slot's total margin (not just tickets) with the previous two weeks, over three weeks, and check whether the people who came were new customers. It is a practical rule of thumb, not a scientific one.