Technological Solutions
Shaping the Industry
0
SaaS Development · Growth Analytics
Which ad brings in the leads that actually buy? X3 measures every visit without cookies, attributes ad spend from Meta, Google, LinkedIn, and TikTok down to the single ad, and reports qualified leads back to the ad networks server-side. Built by CPI Technologies, operated by good vibes GmbH in Zug, live at x3growth.io.
4
ad networks with spend and conversions: Meta, Google Ads, LinkedIn, TikTok
0
cookies needed to measure — X3 keeps counting when the banner is declined
0–100
relevance score for every single session
Starting point
Ad accounts show clicks, the CRM shows inquiries. What happens in between is invisible: which ad brought in the lead who actually buys later? Since cookie banners and tracking protection swallow a large share of visits, the numbers in analytics and ad accounts keep getting patchier — and campaigns optimize for clicks instead of customers.
We had exactly this problem on cpitech.io: leads came in through a quiz, WhatsApp, and meeting bookings, spend sat in several ad accounts, and conversions ran through a mix of pixels and tags. The tracking engine we built to fix it became X3 — a standalone product that brings visit, ad spend, lead, and deal into one data model and tells you where to double down.
Approach
We built X3 as a multi-tenant platform on Next.js. A lean SDK captures visits, clicks, videos, and forms without cookies and sends them to the API through a first-party proxy on the site's own domain. Data lives in Postgres on Neon in Frankfurt, the application runs on Vercel. Google, Meta, LinkedIn, and TikTok ad accounts connect through Composio; X3 pulls spend down to the single ad and matches it to visits.
The Conversions API, Stripe billing, and an MCP server — through which Claude and other AI assistants run the analysis themselves — all sit on the same data model. We built X3 with AI-assisted development: agents write code and tests, we own architecture, review, and release.
Product tour
Every screen shows the demo clinic “Klinik Dr. Weiß” with made-up sample data — no real customers.
Every project's home page sums up the week in one sentence: leads, ad spend, cost per qualified lead. Below it are concrete recommendations — such as moving about €700 from an expensive Meta video to a cheaper Google campaign. Tasks check themselves off as soon as the numbers show it.
Screen from the X3 dashboard · sample data
Who is on the site right now, from which ad, on which page? The live view shows visitors as tiles per page and on a map — no names, no cookies. Agencies see every project on one shared map with a live feed.
Screen from the X3 dashboard · sample data
Any visit can be played back — only with consent and with masked text. Next to it: the relevance score and the source down to the campaign. Mass replay plays many visits at once across the live board's page tiles.
Screen from the X3 dashboard · sample data
The click heatmap doesn't just show where people click — it shows which elements lead to leads, with click share and lead rate per element. Plus scroll depth per page and device.
Screen from the X3 dashboard · sample data
Experiments run for a fixed duration with a clear decision rule. At the end, X3 rolls out the better variant automatically, stops harmful variants early, and shows the expected effect with its range — in this example +32% leads from showing the price in the hero.
Screen from the X3 dashboard · sample data
Agencies see every client project in one cockpit, sorted by what needs attention right now: leads, ad spend, cost per qualified lead, and budget forecast per project. Clients connect their ad accounts via a link, without an X3 login of their own.
Screen from the X3 dashboard · sample data
AI-native
X3 ships with its own MCP server. One click connects Claude — or Claude Code, Cursor, VS Code, and other MCP-capable tools — to your data, signed in via OAuth instead of copied keys. The AI agent pulls metrics, explains individual leads' journeys, compares campaigns, and writes its own SQL queries when needed.
It can act, too: create goals and funnels, mark conversions, or install X3 on a website. Write actions go through roles, confirmations, and an audit log — the agent works with you, not around you.
In use on cpitech.io
X3 has fully replaced the previous tracking on cpitech.io: the old tracking and conversion engine is switched off, and X3 is now the only path conversions take to Meta and Google Ads. Quiz inquiries, WhatsApp contacts, and meeting bookings land in X3 as leads with source, answers, and journey.
The site's A/B tests run on X3 as well — for example the quiz start button copy on the software landing page, or a two-step contact step in the quiz. Whatever X3 recommends to others, we test on our own funnel first.
Feature set
The feature set as listed in the current price list. Which feature is in which plan is on x3growth.io.
KPIs, recommendations, and notices on one page. Tasks resolve themselves once the numbers show it.
Every session gets a score from 0 to 100 from active time, scroll depth, clicks, and goals — configurable per project and versioned.
Leads with answers, fit, journey, and pipeline status. Revenue can be imported; X3 only reveals contact details on click.
Google, Meta, LinkedIn, and TikTok: spend, cost per lead, and cost per qualified lead down to the single creative, with preview.
Last click, first touch, linear, position-based, or time decay — switch models and X3 recalculates. Journeys work across devices.
Qualified leads go to Meta, Google Ads, LinkedIn, and TikTok server-side — with enhanced matching and automatic retries.
Click and scroll heatmaps, session replay only with consent and masked text, and many visits at once in mass replay.
Visitors in real time per page and on the map — for one project or all projects at once.
Tests with a fixed duration, protection against harmful variants, and automatic rollout of the better variant.
HubSpot and Pipedrive: leads into the CRM, meetings and won deals back — for ROAS based on real revenue instead of estimated lead values.
Second-by-second audience curve, drop-off, rewatch, and skip zones, and a frame preview when hovering the curve.
One click connects Claude to X3. The AI agent queries data, explains journeys, and can even install X3 on a website.
Every client project in one overview, plus a connect link for clients' ad accounts. White-label reports on a custom domain are planned.
A lean SDK for browser and server, a first-party proxy on your own domain, plus REST API and SQL access.
Tech stack
Model calculation
This calculation shows the monthly recurring revenue X3's real price list produces.
Model calculation
| Plan | Monthly price | Price billed annually (per month) | Customers in model | MRR share |
|---|---|---|---|---|
| Project Analytics | €28 | €22.40 | 150 | €3,780 |
| Project Growth | €58 | €46.40 | 100 | €5,220 |
| Project Pro | €178 | €142.40 | 50 | €8,010 |
| Agency Analytics | €196 | €156.80 | 125 | €22,050 |
| Agency Growth | €596 | €476.80 | 60 | €32,184 |
| Agency Pro | €1,596 | €1,276.80 | 15 | €21,546 |
| Total | 500 | €92,790 |
€92,790
MRR in the model iMRR = Monthly Recurring Revenue: the revenue that recurs every month from active subscriptions — the key metric of a software product.
€1,113,480
ARR in the model iARR = Annual Recurring Revenue: the revenue that recurs every year, i.e. MRR × 12. SaaS valuations are based on this number.
€7,237,620
Company value at 6.5× ARR
Model calculation, not a forecast: 500 paying customers in the plan mix above, 50% of them billed annually at 20% off. Net prices from X3's price list, excluding overage, extra projects, and add-ons, with no churn; free accounts don't count. Assumed ARR multiple of 6.5× (typical range for private SaaS companies: 2–7).
Next step
X3 shows how we build software today: AI-assisted, in days instead of months, with billing, AI interface, and operations from a single team. Tell us about your idea — we'll show you what's realistic and what it could be worth.
Tell us briefly about your project.
Dein Vorhaben
Maximilian Schmidt, Founder & CEO, has answers for anything
50.1104° N, 8.6538° E
