How we stopped the ad spend leak at a six-site coffee chain

About the project

A chain of six coffee shops in one city. Advertising ran as a single campaign covering the whole city, on $2,400 a month, and the reporting came down to one sentence: orders are up.

The challenge

Nobody in the company could say which of the six sites produced orders and which only spent. There was one campaign, geotargeting was set to the whole city, and in the report all six sites looked equally successful.

What we did

Split the advertising into six campaigns, one per site, each with a 1.2 km radius around it. That immediately showed a picture nobody had seen: two sites in residential districts produced 41% of orders on 17% of budget, while two in the centre ate 38% of budget and returned 9% of orders. The central sites were competing with each other for the same audience, and for people who would have walked in anyway.

What worked

Switching things off. We stopped advertising at the two central sites entirely, and revenue there fell 4%, not the 38% the owner feared. That money went to the residential districts, where every dollar worked three times harder.

The second thing was Google Maps listings. Local SEO for a coffee shop is an order of magnitude cheaper than ads: somebody searches for coffee nearby and sees you without clicking an ad at all. We filled in all six listings with hours, photos and menus.

What went wrong

In the first month we measured success by total orders and nearly concluded the split had failed, because the total simply did not move. The trend was only visible site by site: three growing, two falling, one flat. The chain average hid all of it.

What we should have done

Split the campaigns before the first dollar, not after a year. The rule is simple: if a business has several physical locations, each one gets its own campaign. Otherwise you are not measuring, you are averaging.

What a proper plan does to the economics

Before: $2,400 spent, 890 orders, $2.70 per order. After: $1,650 spent, 1,640 orders, $1.00 per order.

Spend fell by $750 a month and orders rose by 750. Over a year that is $9,000 of budget saved and 9,000 extra orders at an average ticket of $4.20, which is another $37,800 of revenue that simply was not being collected.