case · 2025 role: product logic + automation duration: MVP cycle status: working tool

Yandex Maps Outreach
map → clean base → first touch

Turns geo search results into a clean B2B base and brings each lead to a useful first touch without manual chaos.

[ visual · map → table → first touch ]
01 / context

Why it
was needed

Cold outreach usually breaks before the message: if the source base is dirty, the offer becomes random.

Manual collection from maps quickly turns into an unreliable sheet: duplicates, empty contacts, inconsistent addresses, and no answer to who should be contacted first.

I built a pipeline that thinks like a salesperson: find the right segment, remove noise, explain the reason to reach out, then prepare the message.

02 / solution

How the outreach machine
works

The system moves a company card from geo source to action queue: niche, normalization, priority, context, and outreach draft.

  • I
    Card collection
    Work starts from segment, city, and demand signal, not an endless list of nearby companies.
  • II
    Cleaning
    Duplicates, empty contacts, odd records, and noisy matches disappear before a human spends attention.
  • III
    Enrichment
    A card becomes a work object: niche, city, reason, next step, source, and priority.
  • IV
    Prioritization
    The queue shows who to contact now and why there is a commercial reason.
  • V
    First touch
    The message draft is born next to the reason, so outreach does not sound like mass spam.
[ visual · niche search ]
[ visual · cleaned table ]
[ visual · scoring ]
[ visual · message prep ]
live / interactive

Live outreach console in demo mode

A real React/Vite build from the local project: Supabase and Telegram worker are mocked, but Today queue, Leads, Settings, onboarding, and panels use real components.

embedded demo
03 / result

What we could
measure

map → clean base → first touch. Here are verifiable outcomes and visual materials from the source projects.

1000+
cards / day
1
narrow funnel
0
manual copy-paste
B2B
focus
The value is not parsing as a trick. The value is turning geo search into a sales queue with a reason, priority, and next step.
- project breakdown
04 / how it is built

Stack,
and why

01
Parsing
Pulls source data faster than manual collection.
02
Normalization
Turns messy cards into a table you can work with.
03
Filtering
Removes noise before it reaches communication.
04
Scoring
Helps start with the best-fit companies.
05
Export
Moves the base into a real workflow: Telegram, CRM, CSV, or outreach queue.
05 / next work
Source Hub CRM
Lead source console
06 / contact

Similar task? 30 minutes, no brief.

Describe what you need. I will say honestly whether I take it, how much it costs, and whether AI can speed it up.

book 30 min message on telegram