The first thing it showed me was that 31% of our “visitors” were crawlers. Our conversion rate had not dropped; it had never been that high. Now the numbers I forward are numbers I trust.
RK
R. K.
Store owner · DTC fashion
For site owners & developers Is the snippet seeing real visitors?
One asynchronous script covers analytics and chat. Five layers of bot rules sit in front of the numbers you report, and a data-quality page tells you when a release broke collection.
Questions you answer every day
- How much of yesterday’s traffic was crawlers, and which ones?
- Did the release break collection on any page?
- Which AI crawler reads us the most, and does it send anyone back?
- Can analytics run without a cookie banner on the EU site?
Visitors · humans
48,213
▲ 12.4%
Pageviews
162,940
▲ 8.1%
Filtered as bots
31%
▼ 2.2 pts
Live now
128
Visitors · humans vs all traffic
Humans onlyAll trafficSep 18Sep 25Oct 1
Bot layers · 14 days
Humans onlyAll trafficFive layers, 197+ rules. False positives can be marked as human.
The daily friction
Three things that make you distrust the dashboard
You installed it; you get asked when the numbers look wrong. These are the usual reasons.
A third of “visitors” never had a browser
Crawlers, scrapers and AI bots inflate visits and crush conversion rate. The marketing team reports the inflated number for a quarter.
A release broke collection, silently
A CSP change or a renamed container drops events on one template. Charts shrink; nobody gets an error.
Two scripts, two consents, two bills
Analytics from one vendor, chat from another. Two tags, two cookie notices, two sets of identifiers that never match.
Numbers you report
What the install looks like in numbers
Facts about the snippet and the sample site used across this website.
- Snippet, gzipped, async
- ≈2 KB
- Snippet, gzipped, async
- Bot rules in five layers
- 197+
- Bot rules in five layers
- SDK platforms + server API
- 7
- SDK platforms + server API
- Cookies required for analytics
- 0
- Cookies required for analytics
Visitor counts on this page are sample data. Snippet size and rule counts are current product facts.
Built for site owners & developers
Six things you check once, then trust
Install, filtering, privacy, quality, SDKs and alerts. After the first week most of this runs on its own.
One snippet for analytics and chat
Paste once before , or use the plugin for WordPress, Shopify or Webflow. Analytics, autocapture and the chat widget load from the same async script.
- ≈2 KB gzipped, async, never blocks render
- Plugins for WordPress, Shopify, Webflow and more
- Server API for orders and other events from your backend
index.html
<!-- analytics + chat, one script -->
<script async src="https://app.tapcub.com/w.js"
data-site="YOUR_SITE_KEY"></script>
// server: the order, with an id for de-duplication
tapcub.track('purchase', { order_id: 'A1024', revenue: 42 });Bots filtered in five layers
User-agent, headers, client signals, ASN and rate. 197+ rules split search engines, AI crawlers and scrapers from people. Toggle between humans and all traffic on any report.
- Known bots and AI crawlers by name: GPTBot, ClaudeBot, PerplexityBot and more
- Headless and no-JS traces, hosting ASNs, request bursts
- Not a 100% claim: a false positive can be marked human
Example scenario: 31% of “visitors” were crawlers. Conversion rate did not drop — it had never been that low. The number reported upward changed, the business did not.
Filtered by layer · 14 days
HumansAllThe AI content ledger lists which crawler read what, and who sent visitors back.
Cookieless by default, consent when you add more
Analytics counts visitors without a cookie or a persistent identifier. If you turn on chat or a custom identifier, the consent module and the privacy page tell you exactly what changed.
- Cookieless visitor counting for analytics
- Data region choice and a retention period per plan
- A privacy page you can link from your own policy
| Collected | Analytics | With chat |
|---|---|---|
| Cookie | None | Session cookie for the thread |
| IP address | Truncated before it is stored | Same |
| Identifier | Daily rotating hash | Visitor ID shared with analytics |
| Form fields | Not captured by default | Message text only |
| Retention | Free 6 months · Pro 12 · VIP 24+ | Per chat plan |
Know when collection breaks
The data-quality page aggregates intake warnings, collector rejects, blocked events and high-cardinality keys by hour. A release that drops events shows up as a reason, not as a shrinking chart.
- Rejects by reason: domain mismatch, exclusion rule, quota, bad payload
- Stale selection rules that stopped matching after a redesign
- Quarantine for events that failed validation
Data quality · today
- Collector rejects +412 · domain mismatchstaging.example.com sending with the production keyWebhook
- Rule “Add to cart” staleSelector matched 0 times since the 14:00 deployEmail
- Intake warnings back to baseline12 per hour after the fixResolved
Monitors for traffic drop, error spike, crawl anomaly and quota are built in.
SDKs for every surface, one person
Web, iOS, Android, HarmonyOS, mini-programs, Flutter and React Native, plus uni-app and a server API. The same login ID merges a person across all of them.
- 7 client SDKs and a server API with the same event shape
- Mini-program scenes and app versions as built-in properties
- Identity merge on login; anonymous history is kept
| Platform | Install | Identity |
|---|---|---|
| Web | <script async> or plugin | Anonymous → login ID |
| iOS / Android / HarmonyOS | Package manager | Device ID → login ID |
| Mini-programs | npm or copy | OpenID → login ID |
| Flutter / React Native / uni-app | Package manager | Wraps native SDKs |
| Server | HTTP API | Login ID, insert_id de-dup |
Monitors that message you, not a dashboard
Traffic drop, error spike, crawl anomaly and quota are built in. Add a monitor on any saved metric with a threshold, a period change or a forecast band.
- Hour, day, week or month grain
- Channels: webhook, WeCom, DingTalk, Feishu, email
- Scheduled daily or weekly report for the people who never log in
Error rate · 14 days · forecast band
Error rateExpectedSep 18Sep 25Oct 1
A day in the role
Install once, then a quiet dashboard
The first week is set-up. After that the job is reading alerts and answering “is this number real?” with a link.
-
Day 11
Install and verify
Paste the snippet, add the site, confirm the first pageview in the live view. Turn on cookieless mode for the EU domain.
Live view
-
Day 22
Check the bot split
31% of traffic is filtered. Open the AI content ledger: GPTBot read 1,204 pages and sent 412 visitors back.
Bots & AI crawlers
-
Week 13
Confirm the events
The event inbox proposes six names from autocapture. Confirm four, send purchase from the server with an order ID.
Event inbox
-
Any Tue4
A monitor fires
Collector rejects jumped: staging is sending with the production key. Fix the env variable; rejects fall back within the hour.
Data quality
-
Any Fri5
Answer “is this real?”
Marketing asks whether the spike is people. Toggle humans only, share the link. The answer is the chart, not an argument.
Shared dashboard
The same snippet serves the chat widget, so adding support later is a toggle in the console, not a second tag.
Working with other roles
You own the install; everyone else owns a question
Give each team the view they need and keep the keys, costs and members to yourself.
Marketing & paid media
Reads channels with bots already removed, so return is computed on people.
OpenProduct managers
Confirms events in the inbox; you only step in for server events and selectors.
OpenSupport teams
Gets the chat widget from the same snippet, with Agent-only seats you can scope.
OpenMembers · example.com
Invite- OWYou (site owner)Owner
- MKMarketingAnalyst
- PMProductAnalyst
- FIFinanceViewer
Up to 10 sites on the free plan, 1 member; Pro 3 members; VIP 1 ten; VIP 2 unlimited.
Owner and Admin hold site keys, members, costs and notification channels. Analyst edits events, rules and reports. Viewer reads. Agent-only is chat-only.
Templates & solutions
Packs tell you which events to send from the server
Every pack marks which events must come from the backend and which properties are required. Read it before you write the integration.
Templates that fit this role
Solutions worth reading next
In their words
What a site owner notices first
Two scenes from the job, written as the person doing it might tell them.
One script, one consent note, one visitor ID shared by analytics and chat. I removed two tags from the site and the page got faster.
TN
T. N.
Developer · agency
Illustrative scenarios written from the perspective of people who do this job. Not real reviews.
FAQ
Questions site owners ask
Still have a question?
Chat with the team behind TapCub — we usually reply within a few hours on business days.
How big is the snippet and does it block the page?
About 2 KB gzipped, loaded asynchronously. Analytics, autocapture and the chat widget share it; the widget loads lazily when needed.
How are bots handled?
Five layers: user-agent, headers, client signals, ASN / IP and rate. 197+ rules are built in. Every report can switch between humans only and all traffic; a false positive can be marked human.
Do I need a cookie banner for analytics?
Analytics can run without cookies, using a daily rotating hash for visitor counting. Chat uses a session cookie for the thread. The privacy page lists exactly what each mode stores.
Which platforms have SDKs?
Web, iOS, Android, HarmonyOS, WeChat and other mini-programs, Flutter and React Native, plus uni-app and a server API. The same login ID merges a person across them.
How long is data kept?
Free 6 months, Pro 12 months, VIP 24 months or more. Raw events can be slimmed before the retention deadline from the data-quality page.
Paste one line and see the first pageview in a minute
Free plan, up to 10 sites, no credit card. Humans-only toggle on every report.