£80 of rewards, £7,000 of new business

Years back, one of the lunch businesses I worked with turned on a recommend-a-friend reward. Over the following months they gave away something in the region of £80 in credit. The new business that came through it was over £7,000. I've seen a lot of marketing spend in this trade and I've never seen another number like it.

It works because it isn't really advertising. When somebody tells a colleague "order from these, they're good and you get a quid off", that's a recommendation from a person the colleague already trusts, delivered at the exact moment they're deciding what to have for lunch. No leaflet can do that.

Your customers sit in rooms with your next customers. That's not true of most businesses. A hairdresser's customer might mention them once a month over a drink; your customer sits four feet from six other people who all have to eat at half twelve.

It also solves the awkward part of growing a round, which is that the value of a stop goes up with every extra person on it. Two orders at an office is a detour. Eight is a proper stop. Your existing customers are the only people in a position to turn the first into the second, and they'll do it if you make it easy and give them a reason.

Keep it small. A pound or two for the person doing the recommending is plenty, and I'd give something to the new customer as well - it gives your existing customer something to say. "Order from these and we both get a quid off" is a much easier sentence than "do me a favour".

Resist the temptation to be generous. A fiver each way doesn't get you five times the referrals, it just costs you five times as much and starts attracting people who are in it for the fiver. The reward isn't really paying for the introduction; it's giving a helpful person permission to be helpful.

Pay it as credit against a future order rather than cash. It costs you the margin rather than the full amount, and it brings both of them back in.

Not on day one. A referral scheme needs somebody to do the referring, so it does nothing until you've got a base - I'd say wait until you've got fifty or so regulars, which for most rounds is a few weeks in. Before that, your effort is better spent on flyers and getting your name in front of people who've never heard of you.

Once it's on, leave it on. This is a slow, steady drip rather than a campaign, and that's exactly what you want: a handful of new customers a month, arriving pre-warmed, at a cost of about two pounds each.

The commonest way to waste a referral scheme is to set it up and never mention it. Put it in your newsletter. Put it on the bag. Mention it to anyone who compliments the food, because someone who's just told you they like your sandwiches is about as warm as a customer gets.

Pekkish has a recommend-a-friend feature built in, so the reward is applied automatically to both accounts and you can see which customers are actually bringing people in - which, by the by, is worth knowing for its own sake. The handful of people doing most of your recommending are the ones worth looking after.