Skip to content
Browse the docs

How detection works

Spurn scores every paid click from 0 to 100 by combining independent pieces of evidence, then turns the score into a verdict using your site's sensitivity.

Which visits Spurn looks at#

Spurn only looks at paid landings. A landing is paid when its URL carries an ad platform's click id, or UTM tags with a paid medium. Everything else is organic and is ignored: never sent, stored or billed.

PlatformClick id parametersWhat Spurn does
Google Adsgclid, gbraid, wbraid, gad_sourceDetects, and blocks with IP exclusions
Microsoft AdsmsclkidDetects, and blocks with an exclusion audience
Meta Adsfbclid, with a paid medium or ad parameterDetects, and blocks with an exclusion audience
TikTok, LinkedIn, X, Pinterest, Snapchat, Redditttclid, li_fat_id, twclid, epik, ScCid, rdt_cidDetects and reports

Without a click id, a link counts as paid when its utm_medium is cpc, ppc, paid, paidsearch, paid_search, paid-search, paidsocial, paid_social, paid-social, cpm, cpv, display, shopping or ads. The utm_source decides the platform; an unknown source is filed under Other paid.

Meta adds fbclid to ordinary link shares too, so fbclid alone is not enough. It counts as paid only alongside a paid utm_medium or one of Meta's ad parameters: ad_id, adset_id or utm_id.

Three witnesses#

Up to three sources describe each click. Spurn merges them into one click by click id or by page view, so a click seen three times is still one click and one credit.

  1. The tag (/s.js) runs in the visitor's browser. It reports browser signals (signs of automation, screen and hardware, timezone, language and a device fingerprint), engagement (time on page, scroll depth, pointer movement, interactions and pages viewed) and conversions.
  2. A server hit comes from your WooCommerce or WordPress plugin, the Next.js package, the Cloudflare Worker, your own server, or the tracking template that Google and Microsoft call when the ad is clicked. It records the IP address and user agent even when the page's JavaScript never runs.
  3. The ad platform confirms the click happened. Spurn imports Google Ads clicks every hour and matches them to landings by gclid, which adds the campaign, keyword and an estimated cost.

Spurn also looks up the IP address in local data: its network owner and country, hosting and cloud provider ranges, Tor exit nodes and VPN providers. Lookups run on Spurn's own servers; visitor IPs are not sent to an outside lookup service.

On top of that it counts repeats inside your click window (by IP, by visitor, by device fingerprint and by /24 network), checks whether a click id has been replayed, and checks the IP against your rules and your existing exclusions.

The score#

Each piece of evidence is a reason with a weight. Spurn combines the weights as independent evidence:

score = 100 × (1 − Π(1 − wᵢ / 100))

One reason on its own scores its weight. Each extra reason closes part of the gap that is left, so the score climbs towards 100 but never passes it. Two examples:

  • Data centre network (75) and No interaction (25): 100 × (1 − 0.25 × 0.75) = 81.
  • VPN or proxy (45) and Timezone mismatch (25): 100 × (1 − 0.55 × 0.75) = 59, rounded.

Every reason#

The reasons below are the full list, worst first. The code is what the REST API returns in a click's reasons array. A hard reason is conclusive on its own: a purchase or long engagement never clears it.

ReasonCodeWhat it meansWeightHard
Known botBOT_USER_AGENTThe browser announced itself as a crawler, scraper or monitoring tool.95Yes
Automated browserAUTOMATIONThe page was driven by automation software (headless Chrome, Selenium, Puppeteer or Playwright).90Yes
Tor exit nodeTORThe click came through the Tor network, which hides who and where the visitor is.85Yes
Data centre networkDATACENTERThe IP belongs to a hosting or cloud provider (AWS, Google Cloud, OVH and the like), where people do not browse from.75No
Repeat clicksREPEAT_IPMore paid clicks from this IP within your window than your rule allows.65No
Repeat deviceREPEAT_DEVICEThe same device clicked your ads more often than your rule allows, even if its IP changed.65No
Already blockedBLOCKED_NETWORKThis network is on your exclusion list but still reached an ad, through a campaign the list does not cover.60No
Outside target countriesOUTSIDE_GEOThe visitor's IP is in a country your ads do not target.60No
Click ID replayedCLICK_ID_REUSEThe same ad click ID arrived from more than one network, which a real click cannot do.55No
Network burstSUBNET_BURSTSeveral different IPs in one small network block clicked in a short time.50No
VPN or proxyVPNThe IP belongs to a VPN or proxy provider, which hides the visitor's real location.45No
Inconsistent deviceDEVICE_MISMATCHThe device's signals contradict each other, for example a phone browser with no touch screen.35No
Page never loadedNO_PAGE_LOADThe click reached your server or the ad platform, but the page never ran in a browser.35No
Timezone mismatchTIMEZONE_MISMATCHThe browser's clock is set to a different part of the world than its IP address.25No
No interactionNO_ENGAGEMENTLeft within seconds without scrolling, moving the pointer, tapping or typing.25No

Verdicts and sensitivity#

The score becomes a verdict through your site's sensitivity, set in the Rules tab:

SensitivityInvalid atSuspicious at
Relaxed75 or more45 or more
Balanced (default)65 or more35 or more
Strict55 or more25 or more

A score below the suspicious cut-off is valid. Take the VPN example above, which scores 59: it is suspicious at Relaxed and Balanced, and invalid at Strict.

Only invalid clicks are blocked. Suspicious clicks are reported so you can see them, and they count in your stats, but Spurn takes no action on them. For what each reason does on its own at each sensitivity, see Monitor and protect modes.

Mitigation#

Real shoppers sometimes look odd: a buyer on a work VPN, a traveller whose laptop clock is still set to home. So good behaviour counts in their favour. When a click leads to a purchase, or the visitor stays at least 30 seconds and interacts with the page at least 5 times, Spurn halves the score.

Hard reasons are the exception. If a click has Known bot, Automated browser or Tor exit node among its reasons, the score stays as it is, purchase or not.

Two examples:

  • Data centre network alone scores 75. If that visitor buys, the score drops to 37.5: valid at Relaxed, suspicious at Balanced and Strict.
  • Known bot scores 95. A purchase changes nothing.

Allowlist#

IP addresses and ranges on your site's allowlist are always valid, with no reasons, whatever else Spurn sees. Use it for your office, your agency and anyone else who clicks your ads for work. See Allowlist and rules.

Rules you control#

Each site has its own rules in the Rules tab. They change what Spurn counts as a reason:

RuleDefaultWhat it does
Target countriesEmpty (any country)A click from outside these countries gets Outside target countries
Click window24 hoursThe period over which repeat clicks are counted
Clicks per IP3More than this from one IP in the window gets Repeat clicks
Clicks per device3More than this from one device in the window gets Repeat device
Minimum engagement3 secondsLeaving sooner with no interaction gets No interaction
Block data centre networksOnCounts Data centre network against a click
Block VPNs and proxiesOnCounts VPN or proxy against a click
Block TorOnCounts Tor exit node against a click
Block automated browsersOnCounts Automated browser against a click

Two more rules shape what gets blocked rather than how a click is scored: subnet grouping and how long an exclusion lasts. Both are covered in Google Ads exclusion lists, and every rule above is explained in Allowlist and rules.

Ghost clicks#

A ghost click is a click Google Ads recorded that never reached your store. Spurn finds them by importing your Google Ads clicks every hour: a Google click with no matching landing two hours later is a ghost click.

Google's click report has no IP address in it, so a ghost click cannot be scored or blocked. Spurn counts ghost clicks separately and includes them in refund claim packs. One ghost click is not proof of fraud: a shopper on a slow connection who gives up before the page loads leaves one too. A pattern of them is worth a claim.

The tracking template turns many would-be ghost clicks into scored clicks, because it records the IP when the ad is clicked even if the page never loads. See Tracking template.

Pending verdicts#

When the tag reports a landing, Spurn answers straight away with an instant verdict: valid, suspicious, invalid or pending. Some clicks are clear at once. A known bot does not need to be watched. Others wait for engagement.

While the page is open, the tag reports engagement at 10 seconds, at 30 seconds and when the page is hidden or closed, and Spurn rescores the click each time. The verdict becomes final when the visitor closes the page, or three minutes after the landing, whichever comes first. Spurn finalises clicks every minute, so the final verdict arrives within about four minutes at most.

Only final verdicts lead to blocking. Until then the click shows as pending in your click log.

If your monthly clicks and click packs have both run out, new clicks are recorded but not analysed. They have no score and no reasons, and the dashboard marks them as unchecked.

IP address data powered by IPinfo.