At risk / mo
£6,980
Session intelligence, unified
See what's silently costing you sales.
TraceForge records every session, catches the errors and friction that lose orders, and puts a number on the revenue each issue puts at risk. One platform — replay, heatmaps, issues, revenue — plus a team of analysts who tell you what to fix first.
<script src="https://cdn.traceforge.com/traceforge.min.js">
One script. Live in minutes.
What TraceForge sees
Session started · desktop
add_to_cart · network 200
Reached checkout
Rage click x5 on promo field
Error - deduped, 142nd session - critical
Resolved in v4.2.1
£0/mo at risk
networkrage clickerrorresolved
◀ drag the timeline — watch one session become a priced issue ▶
Recreated from a demo-store session — masked, like yours would be. Sample figures.
One connected record
Stop stitching four tools together.
A replay tool, an error tool, an analytics tool, and a spreadsheet guessing at revenue. TraceForge is one thread — nothing gets lost between them.
Session
Replay + heatmap
what the user did
Error
JS · network · CSP
what went wrong
Issue
Deduped + ranked
how often, how bad
£ at risk
Revenue impact
what it costs
Session replay
The whole visit, one timeline.
A session isn't one tab. TraceForge stitches the full visit — every tab the user opened — with console, network and performance on the same timeline, and masking you can see.
$129.00
✓ Added to cart
Add to cart
pageview add to cart
Visit & tab stitching
One “session” is the whole visit — across every tab the user opened, not a fragment.
Context, not just video
Console logs, network calls and page performance line up on the same timeline as the replay.
Masking is visible
Inputs render masked in the replay itself — privacy you can see, not just a promise.
Heatmaps
Every click, every scroll, mapped.
Click, scroll and zonal maps over a pixel-accurate snapshot of your real page — the map never drifts from the layout. Flip to mobile and the story changes; compare two periods and see what a release did to attention.
Map
Device
Snapshot & live modes
Freeze the page as a pixel-perfect snapshot, or rebuild it live on the page — your call per device.
Compare two periods
Last 30 days against the previous — same page, same map, what changed.
Segments
Filter the map to a device, source, or converted-vs-not cohort.
Funnels & paths
The exact step where you lose them.
Define the journey — landing, product, cart, payment, confirmation — and TraceForge shows the drop at every step, the issues those dropped sessions hit, and the sessions themselves, replay-ready. Paths branch the way users actually move, not in a straight line.
Checkout funnelOverall conversion 8.1% · 12,400 entries
Use arrow keys to move between steps · sample data
Checkout funnel · 8.1% overall · 12,400 entries
Landing12,400
Product8,93072% continued
Cart3,21536% continued
Payment1,61250% continued
Confirmation1,000−612 · unusual drop
Payment→Confirmation lost 612 sessions — card tokenization fails on Safari (critical). Every dropped session is replay-ready.
Issues at every step
Each drop lists the errors and symptoms the lost sessions actually hit — the why, attached to the where.
Branching paths
Journeys fork. Nested path trees show the real routes — including the detour that kills conversion.
Drop-off to replay in one click
Open the sessions behind any step and watch exactly what stopped them.
Sample funnel · demo-store data · last 30 days
Revenue impact
Bugs, ranked by what they cost you.
Not a vanity score — a statistical estimate of the revenue each issue puts at risk, so you fix what matters first and watch “£ protected” climb as you ship.
| Issue | Sessions / mo | Status | Revenue at risk |
|---|---|---|---|
| Checkout — card tokenization fails on Safari | 142 | critical | £3,180 |
| Add-to-cart returns 500 on /api/cart | 318 | critical | £1,940 |
| Promo code field — dead clicks (mobile) | 690 | high | £1,120 |
| Shipping estimator — slow API > 4s | 254 | high | £740 |
Protected / mo (resolved)
£4,310
Annualized at risk
£83,760
Sample data. Estimates use a pooled statistical method (Mantel–Haenszel with Newcombe–Wilson intervals) and always show confidence ranges — honest numbers, not promises. We switch it on with you during onboarding, using your own order data.
TraceForge AI · grounded analysis
AI that cites its evidence — or admits it can't.
TraceForge AI reads the error together with the replay, console and network trail. Every diagnosis quotes its exact evidence and deep-links to the replay moment — and when it can't prove a cause, it says “insufficient evidence” instead of guessing.
TraceForge AI · high confidence
Token read on undefined during Safari checkout
“cannot read 'token' of undefined” · replay 0:58 ↗ network: POST /tokenize 200, empty body
41 const t = resp.data.token;
42 // resp.data is undefined when Safari blocks 3rd-party cookie 43 pay(t);
On issues without a source map, TraceForge won't invent a file — it asks you to upload one. Honest beats confident.
Source-mapped
Real file:line + a code frame, decoded from your uploaded maps.
Pre-error timeline
The last ~12 things the user did before it broke, fed to the analysis.
TraceForge AI, built in
Grounded analysis runs out of the box — nothing to connect, no keys to manage. Teams with strict data rules can optionally run it on a model they host, down to their own GPU.
MCP · your data in your AI tools
Your data, answered in your own AI tools.
TraceForge ships a hosted MCP server. Connect the assistant your team already uses — Claude, Cursor, ChatGPT, any MCP client — and ask plain questions about your sessions, issues, funnels, performance and revenue. Answers come from your live TraceForge data: read-only, isolated per customer, live today.
Your AI assistant · connected to TraceForge MCP
No dashboard hunting
No exports, no SQL. Ask in plain language; the answer is pulled live from your sessions, issues, funnels and revenue.
Read-only. Isolated. Enforced server-side.
Every key sees exactly one customer's data — scoping is derived server-side from the key, never claimed by the client. Nothing can be written.
17 tools, live today
Issues, evidence, sessions, funnels, paths, events, performance, symptoms, revenue — the whole platform is queryable. No waitlist.
Performance · Core Web Vitals from real users
Slow is a bug too.
TraceForge scores Core Web Vitals from your real visitors — LCP, INP and CLS per page, with FCP and TTFB behind them. No lab robot: the same sessions you replay grade your pages, so when a page turns slow, you can watch what slow felt like.
4.1 s
poor
260 ms
needs improvement
0.18
needs improvement
pass rate
% of real sessions rated “good”
FCP 1.6 s · goodTTFB 620 ms · good
Viewing /product/:id · 3 204 sessions · last 14 days · sample data
| Page | LCP | INP | CLS | Samples | 14-day trend |
|---|---|---|---|---|---|
| All pages | 2.4 s | 210 ms | 0.06 | 18 240 | |
| / | 1.9 s | 140 ms | 0.04 | 6 410 | |
| /product/:id | 4.1 s | 260 ms | 0.18 | 3 204 | |
| /cart | 2.2 s | 190 ms | 0.27 | 2 105 | |
| /checkout | 2.9 s | 460 ms | 0.02 | 1 832 |
Core Web Vitals · real users · last 14 days
/LCP 1.9sgood
/product/:idLCP 4.1spoor
/checkoutLCP 2.9sneeds work
/product/:id turned poor 9 days ago — 74% of real sessions still rate good overall. Sample data.
Per page, not site average
A green site average hides the one red template. Every page is scored separately, on real traffic.
Trends that catch regressions
A fourteen-day trend per page — see the day a release, or a third-party script, made a template slow.
Vitals with receipts
Every score comes from sessions you can replay — watch the slow load a user actually sat through.
Provider health · third-party scripts
When a third-party script fails, it fails on your site.
Payments, analytics, ads, chat, reviews — your pages run code you don’t control. TraceForge watches every third-party host your visitors actually hit: request volume, error rate, latency percentiles and status codes, per host. So when checkout drags at 2 a.m., you know whose script did it.
| Provider | Requests (24 h) | Errors | p50 | p95 | Status mix | Trend |
|---|---|---|---|---|---|---|
| Payments js.pay-••••••.com | 41,200 | 0.2% | 180 ms | 420 ms | stable | |
| Analytics cdn.tag-••••.io | 128,400 | 0.0% | 45 ms | 110 ms | stable | |
| Ads ads.••••••.net | 96,800 | 3.8% | 310 ms | 1,240 ms | stable | |
| Chat widget.••••.chat | 22,100 | 0.1% | 95 ms | 240 ms | stable | |
| Reviews degraded reviews.••••••.app | 18,700 | 11.4% | 610 ms | 2,900 ms | error rate rising |
5 third-party hosts watched · trailing 24h · sample data
✓ Payments · Analytics · Chat · Ads — healthy
degraded Reviews widget — 11.4% errors · p95 2.9s
The kind of slow leak that never shows up in your own error logs.
Per-host truth
Volume, error rate, p50–p99 latency and status mix for every third-party host your pages call — your own endpoints too.
Catches silent degradation
A tag that slows to three seconds throws no error — it just quietly costs conversions. Percentiles catch what logs miss.
Tied to real sessions
Failing requests link back to the network issues and the sessions where visitors felt them — evidence, not just a graph.
The reviews widget is failing one request in nine and adding ~3 s at p95 on product pages — the kind of slow leak that never appears in your own error logs. Sample data · trailing 24 h.
Included · software + expert analysis
We tell you what to fix. Not just what happened.
TraceForge comes with a team, not just a login. Our analysts review what the platform surfaces on your account — issues, revenue at risk, heatmaps, funnels and page performance — and send you a short, prioritised note: what to fix, why, and what it's worth. Grounded in your data, not generic best practice.
Advisory notesample
This week: 3 of your top-5 revenue-at-risk issues sit on mobile checkout — £5,040/mo combined.
Fix order:
- Card tokenization failing on Safari (£3,180/mo) — smallest fix, biggest protection. Tokenize server-side; Safari blocking the third-party cookie is the trigger.
- Promo-code dead clicks on mobile (£1,120/mo) — the click map shows 690 taps on a field that looks disabled. Restyle it or drop it.
- Product pages render slow — LCP 4.1s, “poor” on 3,204 samples. Provider health pins it on your reviews widget (11.4% errors, p95 2.9s); defer it below the fold — nothing else on the page needs to change.
— Senior analyst · TraceForge advisory
Analysts, not autoresponders
A person reads what the platform found on your account and writes the note — triage you can act on, not a digest.
A fix order, priced
Ranked by revenue at risk against fix effort — three things worth doing, not four hundred alerts.
The whole surface
Issues, revenue, heatmaps, funnels, performance, third-party scripts — if your data shows it, the note says what to do about it.
Trust
Replay sessions, not secrets.
Masking and block selectors, network sanitization, and PII redaction before any AI call — with an audit log. Consent is detected automatically from the CMP you already run.
OneTrustCookiebotCookieYesIAB TCF v2Google Consent Mode
Masking by default
30 sensitive field patterns and 9 headers sanitized; mask or block any selector you name.
GDPR/CCPA-aligned controls
Retention configurable per plan (60-day default), anonymized IDs, PII kept out of analysis.
PII never reaches the model
Redacted before analysis, with an audit trail of what was detected.
The platform
And everything around the core.
User paths
The real routes visitors take — branching the way they actually move.
Behavioral symptoms
Rage clicks, dead clicks and form abandonment, detected automatically.
Dashboards
Drag-and-drop widgets over your live data.
Release tracking
New issues and error-rate change per release.
Alerts & notifications
Slack, email or webhook when something spikes or breaks.
Jira & email reports
Push issues to Jira; scheduled digests to your inbox.
Revenue-events SDK
Cart, checkout and purchase events, 13 methods.
Identify users
Tie sessions to a user or account when you choose to.
Why one platform
One platform, or four tools and the gaps between them.
A replay tool here, an error tracker there, an analytics suite, and a spreadsheet trying to price the damage. Each one sees a fragment; none sees the thread. The expensive part isn't any of the tools — it's the seams between them.
| The job | With four point tools | With TraceForge |
|---|---|---|
| See what the user did | A replay tool — sessions in isolation, context elsewhere | Replay with console, network and performance on one timeline, stitched across tabs |
| Know what broke | An error tracker — stack traces with no user story | Every error linked to the exact second of the session where it happened |
| See where pages lose people | An analytics suite — numbers without the why | Heatmaps, funnels and paths built from the same sessions |
| Put a price on it | A spreadsheet and a guess | A statistical revenue-at-risk figure on every issue |
| Keep it fast | A lab test on one URL, once a week | Core Web Vitals from every real session, per page — plus health checks on the third-party scripts you can't control |
| Decide what to fix first | A meeting where four tools disagree | One list, ranked by cost — with an analyst's note on top |
| Run it all | Four scripts, four invoices, four places your data lives | One script, one platform, one place |
The gaps between tools are where revenue quietly leaks. We removed the gaps.
FAQ
Questions, answered.
TraceForge is a browser behaviour tracking and session replay platform for ecommerce and product teams. It records every session, maps clicks and scrolls, catches the errors and friction that lose orders, and puts a number on the revenue each issue puts at risk — one script, one platform, with TraceForge AI and a human advisory service included.
Session replay with full-visit tab stitching, click, scroll and zonal heatmaps, funnels and branching user paths, error and issue tracking, revenue-impact estimates, Core Web Vitals from real users, third-party provider health, behavioural symptom detection, dashboards, release tracking, alerts, Jira and email reports, a revenue-events SDK, a hosted MCP server and expert analyst reviews.
Each issue carries a statistical estimate of the revenue it puts at risk, calculated with a pooled method — Mantel–Haenszel with Newcombe–Wilson confidence intervals — and always shown with confidence ranges rather than a single promise. It is switched on with you during onboarding, using your own order data, so the figures reflect your store rather than industry averages.
TraceForge ships a hosted MCP server, so you can connect any MCP client your team already uses and ask plain-language questions about your sessions, issues, funnels, performance and revenue. It exposes 17 read-only tools, answers from your live data, and every key is isolated to a single customer with scoping enforced server-side — nothing can be written.
Masking is on by default — 30 sensitive field patterns and 9 headers are sanitised, and you can mask or block any selector you name. PII is redacted before any AI analysis, with an audit log of what was detected. Consent is detected automatically from the CMP you already run, data is stored on servers in Europe, and retention defaults to 60 days, configurable to your policy.
One script tag — add it to your site and sessions start recording in minutes. When you want revenue analysis tied to orders, a revenue-events SDK captures cart, checkout and purchase events with 13 methods.
Pricing is per session recorded, quoted straight. Replay, heatmaps, issue tracking, funnels and performance, revenue analysis, MCP access and expert reviews are all part of the platform — tell us your traffic and we will give you a number. Book a demo to get yours.
Every account comes with a team, not just a login. Our analysts review what the platform surfaces on your data — issues, revenue at risk, heatmaps, funnels and page performance — and send a short, prioritised note: what to fix, why, and what it is worth.
Pricing
Per-session pricing. Quoted straight.
You pay for the sessions we record. Replay, heatmaps, issues, revenue analysis, MCP access and expert reviews are all part of the platform — tell us your traffic and we'll give you a number.
Replay + heatmapsIssue trackingFunnels & performanceRevenue analysisMCP accessExpert reviews
Data retention
60-day retention by default — configurable to your policy.