How it works
Four layers in. One explanation out.
Each layer answers something the others cannot. The edge knows every request that arrived. The browser knows what happened inside the page. Your server knows what actually converted. Your finance system knows what was collected. Reveliqo reconciles the four into one account of the day.
Today
- People
- 2,842
- +14%
- Conversions
- 94
- +21%
- Revenue
- €18,420
- +17%
Three things worth knowing
1Organic traffic is growingContext
Most of the increase arrived through two product guides, but their trial conversion stayed below the site average — so the extra visits have not yet turned into proportional signups.
2AI referrals carry a disproportionate share of new revenueWorking
AI assistants delivered 2.8% of human visits and 7.4% of new subscription revenue. The gap between those two shares is the finding.
3Mobile pricing conversion fellCheck this
The decline begins shortly after yesterday's deployment and is concentrated on Safari. Causation has not been established.
Demonstration briefing. Three headline metrics for today: 2,842 people, up 14 percent; 94 conversions, up 21 percent; 18,420 euros of revenue, up 17 percent. Then three explanations: organic traffic is growing but its trial conversion stayed below the site average; AI assistants delivered 2.8 percent of human visits and 7.4 percent of new subscription revenue; and mobile pricing conversion fell shortly after a deployment, concentrated on Safari, with causation not established. All figures are demonstration data.
The four layers
Why one source was never going to be enough.
Every analytics argument in this category is really an argument about which of these you are willing to do without.
Edge Intelligence
Traffic, pages, referrers, campaigns, geography, status codes, origin latency and traffic quality, collected from proxied HTTP traffic where it arrives.
Alone it misses: Anything that happens inside the page — a single-page navigation produces no document request at all.
Smart Beacon
Soft navigation, engagement, Core Web Vitals, clicks, forms, client-side events and content-asset exposure, reported from the browser.
Alone it misses: Any request a browser blocked it from reporting, and the origin timing it never observed.
Business Events
Server-side outcomes your application knows are real: leads, trials, accounts, orders, subscriptions, upgrades, cancellations and refunds.
Alone it misses: How the visitor got there, and what happened on the pages before the event fired.
Commercial Data
Invoices, payments, recurring value, customer worth and margin, read from the systems that bill and recognise them.
Alone it misses: Everything before the sale — which is most of what marketing can actually change.
Getting connected
What setting this up actually involves.
- 01
Connect the site
Already proxied through the edge? Core request analytics start with no snippet. Otherwise deploy Smart Beacon from a first-party path on your own domain.
- 02
Say what success means
Leads, trials, orders, subscriptions, bookings, registrations, calls or something custom. The product reorganises around the answer.
- 03
Choose a collection mode
Minimal, Visit or Customer. The lowest setting is a genuinely useful analytics product rather than a degraded one.
- 04
Add browser signals
Smart Beacon where you need single-page navigation, Core Web Vitals, engagement or form behaviour the edge has no way to observe.
- 05
Send server events
POST the outcomes your application knows are real, with an idempotency key so a retried webhook cannot double-count a sale.
- 06
Connect the money
Your finance and accounting systems, Stripe, Shopify or your own API — so a channel becomes customers and invoices rather than sessions.
- 07
Annotate releases
Deployment markers from GitHub, so the question after a conversion drop has somewhere to start.
- 08
Read the brief
What moved, what it is worth, what was ruled out, and the evidence one click underneath every sentence.
Layer one
Collection that does not depend on the browser cooperating.
A request proxied through the edge is observable there whatever the browser does with scripts afterwards. The analytics record is written aside from the request rather than inside it, so nothing in the pipeline can slow a page down.
To be exact about what that does and does not mean: this is proxied HTTP traffic. DNS resolution does not carry a path, a referrer or a status code, so nothing that only resolves a name can produce web analytics.
Edge collection
Visitor request
Arrives at the global edge
Edge analytics
Path, source, campaign, status, latency, geography
Traffic quality
Humans, crawlers, bots and security traffic separated
Durable spool
Written aside from the request path
Your origin
Answers the request, unaffected
No analytics call enters the live request path. An outage in the analytics pipeline slows nothing your visitors touch.
Diagram: a visitor request reaches the global edge, where path, source, campaign, status, latency and geography are recorded and traffic is classified into humans, crawlers, bots and security traffic. The record is written to a durable spool alongside the request rather than inside it, and the origin answers the request unaffected.
Layer two
Then the things only a browser can see.
Smart Beacon is optional and deliberately small. It reports what the edge has no way of knowing, and nothing the edge already does.
The edge sees the request
- Page and document requests
- Source and referrer
- Campaign parameters
- HTTP status and errors
- Origin latency
- Country and device class
- Traffic quality classification
Smart Beacon sees the experience
- SPA and soft navigation
- Engagement and time on page
- Core Web Vitals
- Clicks and custom events
- Form starts and submits
- Client-side conversions
- Content asset exposure
Layers three and four
The conversion your server knows about, and the money your finance system does.
A browser event says somebody clicked. A server event says the subscription started, the payment cleared and the refund three weeks later reversed it. Reveliqo takes the second kind as authoritative, and reads realised revenue from the systems that own it rather than keeping a ledger of its own.
A business event
POST /v1/events
Idempotency-Key: ord_2026_09_21_5f8a1c
{
"event": "subscription.started",
"site_id": "site_123",
"customer_id": "cust_hash_abc",
"occurred_at": "2026-09-21T17:45:00Z",
"value": 149,
"currency": "EUR",
"properties": { "plan": "business", "billing_period": "monthly" }
}The idempotency key is not optional on revenue. A webhook that retries without one turns a single €149 subscription into two, and the mistake is invisible until finance disagrees with you.
From visit to recognised revenue
- VisitEdge and beaconReveliqo
- Lead / Trial / OrderServer eventYour application
- CustomerAccount createdFinance
- Invoice / SubscriptionBilled and collectedFinance
- RevenueRecognisedAccounting
- MarginContributionAccounting
Reveliqo does not keep its own ledger. The last two steps are read from the systems that own them, so a revenue figure here reconciles against the one finance already trusts.
How an answer is produced
- 1Your question
- 2AI analysis planner
- 3Semantic metrics layer
- 4Analytics query service
- 5Deterministic result
- 6AI explanation
Steps 3 and 4 compute every number. The model plans which question to ask and explains what came back — it never does the arithmetic, and it cannot reach data the asking user is not entitled to see.
Metrics
- people
- visits
- pageviews
- conversion_rate
- revenue
- revenue_per_visit
- average_order_value
- trial_to_paid
Dimensions
- site
- page
- landing_page
- channel
- source
- campaign
- country
- device
Diagram: a question goes to the AI analysis planner, which calls the semantic metrics layer and the analytics query service. Those compute a deterministic result, which the AI then explains. The model never calculates the numbers itself.
And then
You read the answer, not the report.
The brief names what moved and what it is worth. Behind every sentence sits the metric, the segment, the period and the comparison that produced it — plus the explanations considered and rejected, and an explicit statement of what has not been established.
Evidence
Mobile pricing conversion fell 27%.
Why Reveliqo says that
Traffic volume remained within its normal range. The decline is concentrated on iPhone Safari and begins 18 minutes after deployment #842.
- Before deployment
- 4.1%
- After deployment
- 3.0%
- Visits analysed
- 3,481
- Difference
- −26.8%
- Metric
- conversion_rate
- Segment
- device_class = mobile · browser_family = Safari · page = /pricing
- Period
- Yesterday 14:00 – today 09:00, compared with the preceding seven days
Strong correlation with deployment #842. Causation has not been proven — a controlled comparison would be needed for that.
In full
The six steps, end to end.
Collect
Requests at the edge, behaviour in the browser, conversions from your server and commercial outcomes from the systems that own them.
Classify
Separate people from verified search crawlers, AI crawlers, other automation and security traffic before anything is counted as a visit.
Resolve
Every metric has one canonical definition in the semantic layer, so the brief, the funnel, the API and the scheduled email cannot disagree.
Investigate
Decompose what moved across channel, campaign, page, country, device, browser, performance and release, and rank the strongest explanations.
Explain
State what changed, what it is worth and what was ruled out — with the metric, segment and period it used attached to every sentence.
Connect
Follow the outcome through to the invoice, the recurring value and, where the finance system supplies it, the margin.
Connect a site today. Read tomorrow's brief instead of building it.
Connect a site and the first brief arrives with the day's changes already explained — traffic separated from bots, conversions attached to revenue, and the evidence behind every sentence one click away.
Real people separated from bots Every answer shows its evidence Reveliqo runs on Reveliqo