Most writing about customer feedback assumes you have a lot of it. Statistical significance, confidence intervals, sample sizes, trend lines over quarters. Then you run a form in a shop that serves ninety people a day, get twelve responses in a month, and none of that applies. You're left holding a dozen paragraphs and a vague sense that twelve isn't enough to conclude anything.
Twelve is enough to conclude some things and not others. Telling one from the other is most of the skill.
What twelve can't tell you
It can't tell you a rate. If four of your twelve responses mention the wait at the counter, you do not know that a third of your customers find the wait a problem. You know four people did, out of a group that selected itself by being willing to fill in a form. The next twelve might produce one mention or seven. Small numbers move around a lot, and the arithmetic that turns four into 33% is doing something the data can't support. For the same reason it can't rank anything: if three people mention parking and two mention the music, parking is not your bigger problem, because those two counts are indistinguishable at this size.
It also can't tell you whether things are improving. A month at an average of 4.1 followed by a month at 4.4 is not a trend, it's two small samples with a gap between them. I've watched people make real decisions on the back of that gap, and it's basically reading tea leaves with a spreadsheet. False precision in feedback metrics goes further into what those decimal points are hiding.
What twelve can tell you
It can tell you that a thing exists. This is the part people undervalue. If one customer writes three sentences explaining that your booking confirmation email doesn't say which entrance to use, that's not a statistical finding, it's a fact about your booking email. You don't need a second respondent to confirm the entrance isn't mentioned. Go and look, and if they're right, fix it.
A lot of feedback works this way. It reports something checkable rather than expressing a preference. One person saying the toilet lock is broken is sufficient evidence about the lock. One person saying the toilets are unpleasant is a matter of opinion and needs corroboration. Sorting your dozen responses into those two piles takes five minutes and tells you what to do with most of them.
It can also tell you what people notice. Twelve responses cover a wide field of possible things to mention, and what people choose to talk about is information in itself. If nobody mentions your food and everyone mentions your staff, that's a hint about where your experience is distinctive, even at this sample size. Silence about something you're proud of is worth sitting with.
And it can give you language. Twelve open answers will contain the phrases your customers use for what you sell, which is more useful than it sounds for anyone writing a menu, a website, or a sign.
How to actually read them
Read all twelve in one sitting, in full, rather than skimming as they arrive. Feedback read one at a time over a month never becomes a picture, and this is the whole advantage of a small volume: you can do the thing that's impossible at scale, which is read every word.
Don't score them. Resist the urge to tally into categories at this size, because categorising twelve items produces counts of two and three that will then get treated as measurements. Instead, write down the specific claims. "Confirmation email is missing the entrance." "Coffee was cold twice." Concrete lines you can go and check.
Then check them. That's the step that turns small feedback into useful feedback. Go and read your own confirmation email. Stand at the counter at 12:30 and time the queue. Order the coffee yourself. A dozen responses can't establish how common something is, but they can point you at exactly where to look, and your own eyes settle the frequency question better than a bigger sample would.
Keep the ones you can't verify. A single unrepeated comment isn't nothing, it's a hypothesis waiting for a second data point. The value of writing them down is that when the second one arrives four months later you recognise it instead of treating it as new. When the same complaint shows up twice is about exactly that moment.
The response count is itself a finding
Before you accept twelve as your ceiling, ask why it's twelve. Ninety customers a day and twelve responses a month is a very low rate, and the reason is usually mechanical: the QR code is somewhere nobody looks, or the request arrives three days late, or the form opens with eight questions before anything a customer would want to answer.
Fixing that is often easier than analysing what you've got. Response rates respond to placement and timing far more than to survey design, and the difference between twelve and sixty is usually a decision about when you ask rather than a marketing effort. Low response rates are feedback too covers what the number itself is telling you, and why timing matters as much as the question covers the moment.
Remember who your twelve are, too. They're the people willing to fill in a form, which skews toward the strongly pleased, the strongly annoyed, and the unusually conscientious. The quiet middle is missing, and that group is where most of your revenue sits. The silent majority in your feedback has more on what that does to the picture.
When the numbers do start to add up
Small volumes accumulate. Twelve a month is a hundred and forty a year, and a hundred and forty is a genuine picture of your business, particularly if you can see it against the reviews you've collected publicly over the same period.
That's the case for keeping it all in one place from the start rather than in a folder of monthly exports. Qria holds the structured responses you collect and the reviews landing on Google or elsewhere in one view, and summarises what keeps recurring across both, which matters more at small volumes than large ones. When you only get a dozen responses a month, the pattern is only visible across a year, and nobody remembers a year of individual responses without help.
The mistake with a dozen responses isn't reading too much into them. It's dismissing them because a dozen sounds unscientific, and then not fixing the confirmation email that one person took the trouble to tell you about.
