How to tell whether your advertising is working

Your advertising is working if it brings in more money than it costs, and the only way to know is to decide what you are counting before it starts, track each channel separately, and give it eight weeks before you judge. That takes a notebook and ten minutes a week, not an analyst. Most small businesses never do it, which is why most small businesses cannot say which of their advertising is worth keeping.

We run a marketplace for local advertising space, and the question we hear most from advertisers is not "what does it cost" but "how will I know". This is the honest answer.

Decide what "working" means before you start

"Awareness" is not a result. Nobody can count it, and a channel that cannot be counted can never be dropped, which suits whoever is selling it. Working means a number: enquiries, quotes sent, jobs booked, new patients, tables filled on a Thursday, scans of a code. Write the number down before the advertisement goes up, along with what it is now. "We get about four new enquiries a week. Working means seven."

Pick the number closest to money that you can actually count. Enquiries are easier to count than jobs but further from revenue; jobs are closer but slower to arrive. For most trades and services, quotes or bookings is the right place to look.

Five ways to track a channel

Every channel needs its own way of being counted, or the numbers blur together and the radio gets credit for what the poster did. Use at least one of these per channel.

A different phone number

The cheapest trick in advertising and still the best. A second mobile number, or a diversion number from your phone provider, printed only on that one advertisement. Every call to it came from that advertisement and nowhere else. Costs a few dollars a month.

A QR code that counts scans

A code on a poster, a counter card, a flyer or a fence sign, pointing at your website or booking page. A code that counts tells you how many people scanned it and when. A campaign booked through Pritty Ghood comes with its own QR code, and the advertiser can see the scans on their account by week and by time of day, which answers not just "did anybody look" but "when do they look", and that decides whether the poster is in the right place.

A code word, or "mention this ad"

"Mention this poster for ten percent off." "Ask for the Saturday special." Old, and it still works, because it gives the customer a reason to tell you where they saw you. Works best for retail, hospitality and anything with a counter.

A landing page nobody else is sent to

A page on your website that only one advertisement points to, so its visitor count is that advertisement's count. Cheap if you already have a site. Combine it with the QR code and you get both numbers.

Ask every new customer, and write it down

"How did you hear about us?" asked by whoever answers the phone or works the counter, and the answer written in a notebook or a spreadsheet column. Every new customer, every time, for the whole eight weeks. It is the least precise of the five and the most useful, because it catches the channels you were not tracking, including the ones you are not paying for.

The arithmetic

Three numbers.

What a customer is worth. Not the first job. What the average customer spends with you over a year or two, less what it costs you to deliver it. A cafe customer who comes twice a week is worth a few hundred dollars a year in margin. A new physio patient might be worth several hundred over a course of treatment. A roofing customer is worth thousands, once.

What the channel cost. Everything, for the eight weeks: the space, the printing, the design, the time.

Break-even. Cost divided by the value of a customer is the number of customers the channel has to bring in to pay for itself. A poster costing $500 for eight weeks, in a business where a customer is worth $250, needs two customers. Anything above two is profit, and the channel stays.

Write the break-even number next to the channel on your plan. It turns the review from a feeling into a comparison.

The eight-week rule

Two weeks proves nothing. In the first fortnight most people who see an advertisement have not needed you yet, and the ones who have are still deciding. Most channels show what they are going to do by week six, and eight weeks gives a clean number to compare against the eight weeks before.

Decide the review date before the advertisement goes up and do not judge the number before then. You will look, of course, and it will look bad, because small numbers early always do. Looking is fine. Deciding is not.

Reading small numbers honestly

A small business runs on small numbers, and small numbers lie. Three scans in one week is not a trend. Two calls on a Tuesday is not "Tuesdays are best". Our own analytics screen refuses to name a busiest hour until there are at least three scans in it, for exactly this reason: two scans in one hour is a coincidence, and calling it a pattern is how somebody reprints a poster on the strength of nothing.

Some rules for reading them.

• Compare eight weeks with eight weeks, not one week with the week before.
• Count the customers, not the impressions. Two thousand people "reached" is not two thousand customers, or two hundred, or usually even two.
• If the number is under ten, treat it as a direction, not a measurement. It says "probably worth continuing" or "probably not", nothing finer.
• One big job can make a bad channel look good. If a $30,000 roof came from a flyer, that is wonderful, and it is one data point. Run the flyer again before you decide it works.

When a channel is not working

If the eight weeks are up and the channel is below break-even, do one of three things, in this order.

Change the advertisement, not the channel. The place may be right and the message wrong. A poster that lists everything you do gets read by nobody; one that names one job, one offer and one action gets read. Change the message, run another eight weeks.

Change the place. Same message, different spot. A physio's poster in a gym and the same poster in a real estate office are two different experiments, and only one of them was ever going to work.

Drop it. If the message and the place have both been tried and the number is still short, stop and move the money. Nothing about it was wasted: you now know something about your customers that you did not, and the next channel starts from that.

Where the money goes next depends on where your customers are when they decide. We wrote a guide to advertising a small business in New Zealand that goes through the channels and who each suits, and there is a page for most trades and professions at /advertise showing which kinds of local business their customers already walk into.

A worked example

A painter in Tauranga puts an A2 poster in a paint shop for eight weeks, with a QR code to a page on her website and a second phone number printed under it. The space costs her about $55 a week, roughly $450 for the eight weeks, plus whatever the signwriter charges for the poster. Her average job is $4,000 with roughly $1,200 of margin, so break-even is one job.

Over the eight weeks: 31 scans of the code, mostly on Saturday mornings; 6 calls to the second number; 4 quotes; 2 jobs. Two jobs is $2,400 of margin against about $500 of cost. The poster stays. The Saturday morning pattern tells her the paint shop's weekend customers are the ones deciding, so she books a second poster in the paint shop across town rather than a screen in a gym, and changes nothing else.

Had the scans been 31 and the jobs zero, the message would be the suspect, not the place: people were interested enough to scan and not interested enough to call, which usually means the page they landed on did not give them a reason to. New page, another eight weeks.