B2B SaaS · multi-agent market analysis

You know who bought from you.
You don't know why everyone else walked away.

FACTUM sees both sides. A B2B SaaS platform for multi-agent market analysis, built for small and medium business. We capture your customer data and parse open market data from YouTube, Threads, TikTok and Google — and at the intersection of those two datasets, we surface what neither side reveals on its own.

Request access → See how it works
In production
a working product, live today
20+ customers
already paying for the platform
LIC / SALT II
Accelerator · Lithuania · 2026
[ 01 ] The problem

Why small businesses make decisions blind

DATA EVERYWHERE, SIGNAL NOWHERE

Every day, thousands of comments about your niche are posted across YouTube, Threads and TikTok. Reading them by hand is impossible. Processing them on standard CPU tools is too slow and too expensive. The data exists — it's just out of reach.

AI WITHOUT CONTEXT

ChatGPT and the like answer from their training data, not from your market. They don't know what your customers are saying today, what your competitors are doing, or which trends surfaced in your niche this week.

AGENCIES OUT OF REACH

A professional market study starts at €5,000 and takes a month and a half. Small businesses can't afford that — so they keep building strategy on gut feeling instead of data.

[ 02 ] How it works

A two-stream data architecture

FACTUM fuses proprietary brand data with open market data through a hybrid AI pipeline.

STREAM 01
Brand data

You upload the raw, unstructured material behind your business.

Customer interviews Product descriptions Competitor notes
MULTI-AGENT STRUCTURING →
Audience personas Competitive map JTBD breakdown
THE INTERSECTION
Semantic
synthesis

Your internal assumptions get tested against live market data.

⚠ Discrepancy = the real insight
The gap between what you believed and what the market actually says.
STREAM 02
Market data

The parser continuously collects from open sources across your niche and region.

YouTube Threads TikTok Google
PREPROCESSING PIPELINE
cleaning · dedup · tokenize · topic-tag
Structured market sentiment — quotes · sources · signal weight
OUTCOME
The REAL DATA brand document

A living strategic asset that updates with the market in real time.

[ 03 ] The product

Three independent tools.
One database.

01 AI assistant
ai.my-factum.com

Ten specialized AI agents, one for each analytical job: audience personas, competitor and product analysis, survey builder, prompt library, transcription, and review-and-comment analytics. Each agent works only with the user's own data — no guessing. The output of one becomes the input of the next.

100+ user sessions processed
02 Data parser
parser.my-factum.com

Automated data collection from social and web sources: YouTube, Threads, TikTok, Google, Yandex, Otzovik. Runs around the clock, with fresh data landing continuously. The dataset is automatically segmented into three groups: your brand, competitors, and the neutral market.

24/7 · real-time ingestion
03 Analytics platform
cabinet.my-factum.com

A structured path through the stages of market analysis. Every step adds to the REAL DATA brand document. The Real Data % metric shows how far your decisions rest on facts rather than assumptions.

step-by-step analysis path

The three tools run independently but share a single database. Brand profile, parser data, and AI-agent conversations stay connected — no API proxy in between.

[ 04 ] What makes it different

Real data only — no guessing

Every AI-agent finding is backed by a verbatim quote, with its source, reaction count, and date. If the answer isn't in the dataset, the agent says so plainly. Nothing is generated from the model's training data.

Conflicts flagged first

When market data contradicts a brand's internal assumptions, FACTUM flags it as a ⚠ Discrepancy. Not confirmation of what you already believed — the facts that change the decision.

Agents hand data to one another

Audience personas from the audience analysis feed straight into competitor analysis. The product breakdown gets checked against both. First-stream assumptions are tested against second-stream data. Each step builds the document — with no manual data transfer.

Segmented by meaning

Comments are always split: your brand / competitors / neutral market. Agents analyze each group separately, then compare across them — which keeps signals from blurring together and sharpens every conclusion.

[ 05 ] Under the hood

Technical architecture

FACTUM is built as a product-grade B2B SaaS platform with a multi-agent AI core and a proprietary pipeline for collecting and processing unstructured data — proven under continuous load.

AGENTIC AI

Multi-agent orchestration

Specialized AI agents run in parallel and sequential chains. Each agent owns a distinct scope, with validation on its inputs and outputs. Their artifacts accumulate in a single brand document.

REAL-TIME PIPELINE

Proprietary data pipeline

The parser collects continuously, consolidating a dataset of over 5 million units of mixed-type content from 1,000+ sources. Validated on live sources, architected to scale horizontally to thousands.

NLP

Market-sentiment analysis

Sentiment tagging, audience clustering, topic modeling and insight extraction across millions of units of unstructured content. Brand data is matched semantically against the external market dataset — and the intersection is where testable hypotheses form.

ROADMAP · NVIDIA STACK

Technology roadmap

Engineered to move its compute core onto the NVIDIA stack: NIM microservices for local inference, GPU acceleration for parsing, tagging and transcription, and a fine-tuning pipeline that adapts models to each client's niche — the AI colleague that knows your market.

STACK TODAY
AI LAYER
[ 06 ] Who it's for

Who FACTUM is for

SMB marketer

You handle several clients and need data fast — tailored to a specific niche and region. FACTUM gives you a tool you operate yourself: it knows your client's market and keeps updating long after the project ends.

Small business owner

You make the calls without an analytics team. FACTUM walks you through a structured path and hands you a concrete artifact at every step — not vague advice, but data with sources and quotes.

Marketing team

You need one verified database on the market, audience and competitors for the whole team — or for a client you work with. The REAL DATA brand document is a living source of truth that refreshes with every new data pull.

FACTUM is software the user operates themselves. Not analysis-as-a-service — a tool that stays and keeps working.

[ 07 ] Access

Platform access

RECOMMENDED
FACTUM Pro
€200/mo
free tier to test & trial new features
Request access →
PLATFORM METRICS
5M+
content units in the database
1,000+
sources collected from
24/7
real-time collection

Subscription model, complemented by a B2B track. Pricing scales with the volume of data processed.

[ 08 ] Team

The team behind FACTUM

AK
Aliaksandr Kavaliou
CEO & Product

Founder, product architecture, platform development. Designs the multi-agent system and the user journey.

D
Dmitry
CTO & ML Engineer

Parser, infrastructure, ML pipelines. Python, FastAPI, large-scale text processing, deployment on Hetzner VPS. Owns the platform's compute core.

I
Irina
Head of Marketing Strategy

Market analysis, positioning, validating product hypotheses against real SMB data.

I
Inna
Head of Methodology

User-journey structure, the logic of how data accumulates, and the content of each analytical stage.

REQUEST ACCESS

See both sides of your market.

A working product in production with 20+ paying customers. Get on the platform and turn your blind spots into decisions.

Open the request form →
my-factum.com · ceo@my-factum.com