In short
Every counted interaction records a time, an approximate place, a device, a browser, a language and where the visitor came from. That is a useful picture, but only if you know what the platform deliberately does not count.
Four rules do most of the surprising: your own visits are never counted, bots are excluded, the same browser counts once in any fifteen minutes, and "unique" means a network address it has not seen before. Read those before you read the charts.
On this page
What counts as an interaction
Start here, because almost every "my numbers look wrong" question is answered by this diagram rather than by anything in the charts.
What each interaction records
| Recorded | Where it comes from | How much to trust it |
|---|---|---|
| Time | The server clock | Reliable. Stored in a single timezone, so compare like with like when you export |
| Country and city | Looked up from the network address | Country is usually right. City is an estimate, and mobile networks routinely place people in the wrong one |
| Referrer and referring domain | What the browser volunteers | Recorded as "direct" when nothing is offered — which includes most scans, app opens, and anything typed or pasted |
| Device or operating system | The browser's own description of itself | Good for iOS versus Android versus desktop. Not a device model |
| Browser | Same source | Reasonable, with the usual caveat that in-app browsers often report as something else |
| Language | The first two letters of the browser's language preference | A device setting, not a stated preference. A French speaker on an English-configured phone records as English |
| Network address | The connection | Used for the geographic lookup and for the unique count. Depending on how your account is configured it may be stored in a one-way hashed form rather than in the clear |
Notice what is not in that list: a name, an account, a phone identifier. An ordinary interaction is anonymous. Identity enters the picture through separate mechanisms — a Connection, an App scan by a signed-in person — not through the interaction record itself.
Total versus unique
Two numbers sit next to each other on every item and they mean different things.
- Total is the number of counted interactions, after the gates above.
- Unique increments only when the item has not previously been opened from that network address.
So "unique" is a network address the item has not seen before — not a person, and not a device. That distinction cuts both ways and it is worth knowing which way it cuts for your situation:
| Situation | Effect on the unique count |
|---|---|
| An office, a school, a hotel — many people behind one address | Undercounts. Fifty colleagues may register as one unique |
| A mobile network sharing addresses across many customers | Undercounts, sometimes heavily, which matters at outdoor events |
| One person on the venue Wi-Fi, then on mobile data | Overcounts. Two addresses, two uniques, one person |
| A conference where everybody is on one guest network | Severely undercounts. Use the presence view or event data instead |
The honest summary: total is a decent relative measure over time, and unique is a rough lower bound on people. Neither is a headcount. If you need to know how many humans, you need a mechanism that identifies humans.
Where to look
| View | Shows | Best for |
|---|---|---|
| Item statistics | One item: its totals over time, places, platforms, browsers, languages, referrers | Diagnosing one touchpoint, or reporting on one placement |
| Account statistics | Everything you own, aggregated, plus a map and a recent-activity list | Weekly and monthly review, spotting something that has stopped working |
| Recent activity | The most recent interactions as a list | Confirming something is live right now — the fastest sanity check there is |
| Client and project views | Interactions grouped by who the work is for and which campaign it belongs to | Client reporting, provided you assigned items properly at creation |
| Presence view | Who is inside a geofenced area now | Live operations at an event. Not a cumulative measure |
Reading it honestly
A few habits that make the difference between data that informs a decision and data that merely decorates a slide.
- Separate the placements before you launch, not after.One item per poster, per email, per press release. They cost nothing, and you cannot retrospectively split a number back into its sources.
- Read "direct" as "no referrer offered", not as "typed the address".Scans, taps and app opens all land in that bucket, so for physical touchpoints it will be most of your traffic and tells you nothing.
- Expect a first-week spike and do not extrapolate from it.Curiosity scans, your own team, and testing all land early.
- Compare like periods.Weekday and weekend patterns are strong for consumer touchpoints. Week-on-week beats day-on-day.
- Watch for a flat line rather than a low one.Genuinely zero for a live item usually means something is broken or the allowance ran out — not that nobody is interested.
- Export before you conclude anything important.The charts are for noticing; the export is for arguing.
Getting the data out
Three routes, in increasing order of effort:
- Export — a file of the recorded interactions, for one item or across the account. The right tool for reporting and for anything you want to keep.
- Notifications — an alert on interaction, configurable per item. First-scan alerts are the useful default; per-scan alerts become noise within a day on anything popular.
- Webhooks — each counted interaction posted to your own system as it happens, carrying the place, referrer, device, browser and time. This is how interaction data reaches a CRM or triggers a fulfillment process. Both account-wide and per-item webhooks exist.
The two irreversible actions
The second is quieter. Deleting an item also deletes its interaction history. If you are retiring a campaign and might ever want to report on it, export first and then switch the item off rather than deleting it. Switching off also keeps the address resolving gracefully for copies still in the field, which deleting does not.
Symptoms, causes and fixes
| Symptom | Cause | Fix |
|---|---|---|
| "I scanned it myself and nothing was recorded" | Owner interactions are never counted while you are signed in | Test signed out, in a private window, or on another device |
| Scanned five times, counted once | The same browser counts once per item in any fifteen minutes | Expected. Wait it out, or test from different devices |
| Numbers stopped dead mid-month | Your plan's monthly interaction allowance was reached | Check the allowance. Content kept opening; only the recording stopped |
| Unique count far below the people you saw | Everyone was behind one shared network address | Normal at events and offices. Use presence or event data for headcount |
| Almost all traffic is "direct" | Scans and app opens offer no referrer | Normal for physical touchpoints. Separate placements with separate items instead |
| Cities look wrong | Network-address geography, especially on mobile | Trust country, treat city as indicative |
| Language distribution does not match the audience | It reflects device settings, not fluency | Use it for interface decisions, not audience claims |
| An item shows zero although it is definitely in use | Reset Stats was used, or the item was recreated rather than edited | Check the reset history; recreated items start from nothing |
Things people commonly get wrong
Related reading: Mark settings, Redirection controls, Projects, clients and templates, and Geofencing.