Trusty
Consultation
Marketplace Platform Optimization

A new marketing discipline that cracked marketplace algorithms

MPO is a platform that manages visibility, sales and card quality. We lift SKUs in search, turn conversion into revenue and point out what to fix in content, price and reviews — all in one loop.

Skolkovo resident · Innovation priority 5.1.4.5 / 5.2.2.7.3

Platform architecture
1
Data Layer — collect & normalize
2
Modeling Layer — market ML
3
Decision Layer — recommendations
4
Execution Layer — action orchestration
End-to-end loop
Data → Model → Decision → Action → Re-train

80% of marketplace SKUs are left unmanaged

As SKU counts grow, collecting and analysing data becomes impossible. Top-1% of sellers drive >70% of sales — everyone else is lost in the algorithms.

>70%
of sales done by top-1% sellers
500+
SKUs per seller (top-1%)
~₽600B
spent on assortment and ad management by top-1%
80%
of SKUs get no manager attention
External pressure
From market & platforms
↑ ×3
data and parameter volume in 2 years
+12 / mo
marketplace rule & fee changes
20+
per-platform conditions per seller
Seller squeezed
Internal pressure
Routine eats the team
1 : 500
manager-to-SKU — physical attention ceiling
80% of time
goes to spreadsheets, reports and "the way we’ve always done it"
no time
for new-product search, hypotheses and product decisions
Result
What the business gets
Reactive decisions
we react once sales already sagged
COST ↑ SALES ↓ LOAD ↑
unit economics breaks
7–90 days
reaction lag to MP changes

Marketplace Platform Optimization — behavioural marketing, not keywords

Traditional methods (SEO, PPC) work with technical signals and load the team with manual routine. MPO takes the routine off your plate: the platform watches the data, makes the small decisions and runs execution — so sellers and managers spend time on product, new-item discovery and creative work, not on rewriting spreadsheets.

SEO & PPC
Keywords and technical parameters
Budget grows with competition
Works only while paying
No direct organic lift
MPO — Trusty
Behavioural signals + analytics
Cost drops as organic grows
Compounding, long-lasting
Organic positions grow from avatar activity

Four layers working as a single loop

Trusty MPO is a cloud ML platform that unifies sales, demand and ranking data, predicts algorithm response and gives per-SKU recommendations in real time.

01
Data Layer
Collect & normalize
Positions, visibility, keyword dynamics
Ads, stock, prices, reviews
Competitor analysis (20+ params)
APIs across all three marketplaces
02
Modeling Layer
Adaptive market ML
Daily automatic re-training
Algorithm response forecasting
Uplift models
Explainability — "why"
03
Decision Layer
Per-SKU recommendations
Funnel bottleneck diagnosis
Action list with expected effect and risk
WHAT IF generator
Target function: margin / turnover / profit
04
Execution Layer
Automated control loop
Action orchestrator (scenarios, budgets, schedules)
In-house + external APIs
Execution control and reporting
Before/after effect monitoring
Closed loop
  1. Data
  2. Model
  3. Decision
  4. Action
  5. Measure
  6. Re-train

3 tools live, the rest — coming

All tools run in a single end-to-end analytics loop. Below — production-ready. The rest are in active development.

Ready
Buyouts for promotion

Orders and buyouts on target keys through real-account infrastructure. Volume, pace and distribution control.

Ready
Reviews & Q&A

Relevant reviews, questions and answers after real purchases. Scenarios for launch, recovery and conversion strengthening.

Ready
Conversion upkeep

Continuous work with behavioural factors — CTR, cart adds, session activity. We don’t break conversion.

In active development
Soon
Auto-reviews

Review text generation with uniqueness, sentiment and product-fit checks. In development — 2026 release.

Soon
ML training platform

In-house model-training platform: keyword ranking, activation-response forecasting, per-SKU growth curves.

Soon
Ad management

Marketplace internal-ads autopilot: bids, budgets, keyword distribution. Linked to organic via end-to-end analytics.

Soon
End-to-end analytics

Single pane: positions, organic, ads, buyouts, reviews, per-SKU margin. Everything decisions need — in one window.

Soon
Stock monitoring

Stock control across warehouses and FBO/FBS, out-of-stock forecast, automatic alerts before the card drops from results.

Soon
Demand forecasting

Per-SKU sales forecast at 7/30/90 days horizon, factoring seasonality, activations and ad pressure. For procurement and logistics.

Analytics knows where it’s thin — activations fix it precisely

Above we covered the 4 platform layers — they gather data, model the market and find each SKU’s funnel bottleneck. But data without action doesn’t move positions. Buyouts, reviews and behavioural signals are the platform’s lever: pinpoint influence on the algorithm exactly where the model sees a signal deficit.

Platform
Finds the bottleneck

Data + Modeling Layer see: CTR drops on the "kids shoes size 28" cluster — organic sinks.

Activations
Pinpoint the signal

Execution Layer runs buyouts + reviews on that exact cluster. Volume and pace calibrated for anti-fraud.

Platform
Measures and re-trains

In 3–7 days the model reads new positions, ACoS and organic — corrects the plan for the next iteration.

In execution
01
Scenario planning

Orchestrator assembles the plan: which keys to boost, volume, pace. Goal — launch / recover / strengthen.

02
Behavioural signal execution

Real-account infrastructure places orders and buyouts via search queries. Relevant reviews and Q&A follow.

03
Algorithm lifts positions

Card revenue per target keys grows → ranking algorithm lifts it. Effect spills to adjacent keys.

04
Monitoring & re-training

Before/after comparison by cluster. The model learns from the result and corrects the next scenario.

"MPO isn’t about "pumping positions." It’s a single loop: marketplace data → funnel model per SKU → recommendation → precise action → re-training. We built it so the seller stops guessing what to press and starts making decisions on data — across the whole catalogue, not one SKU at a time."

DS
Dmitry Surtsukov
CEO & Co-founder, Trusty

Trusty MPO vs analytics / ERP

The market has analytics (MPStats, MarketGuru) and ERPs (MoySklad). None covers the full loop: data → model → recommendation → action.

FeatureTrusty MPOMPStatsMarketGuruMoySklad (ERP)
External marketplace data✓ 20+ params
ML market modelling✓ daily re-train
Internal metrics (API) — real-time
WHAT IF generator✓ scenarios
Per-SKU recommendations✓ real-time autoManualManual
Automated execution✓ native + externalMP toolsMP toolsSupply only
SKUs per seller100–1,00010–5010–50
Response speed1–100 h7–90 days7–90 days
Target metric upkeep✓ with tolerance

CAC dropped from ₽378 to ₽179 in 6 weeks

Pilot on a card with ~₽200 AOV. External signals launched week 2. Conversion holds, organic orders grow, acquisition cost halves.

CAC by week
Start
378
After 6 wks
179
53% acquisition cost
taller bar = costlier purchase
378
Oct 20
399
Oct 27
262
Nov 3
282
Nov 10
229
Nov 17
179
Nov 24
WeekOrdersSpendRevenueCACExternal signals
Oct 20–267628,75274,482378
Oct 27–Nov 216465,436290,084399132
Nov 3–916543,230359,659262202
Nov 10–1617649,626170,63128213
Nov 17–2320546,960192,2782292
Nov 24–3029853,455279,520179

External signals run in pulses: week 1 is pre-launch, weeks 5–6 hold the effect without them.

Plan 2026–2028

2026
R&D
DataSet v1 (WB + Ozon): positions, visibility, ads, stock, prices, reviews, competitors
Diagnosis v1: funnel bottleneck
Recommendations v1 with effect & risk forecast
Enterprise-ready v1: roles, logging, base SLA
Grant funding, core software registration
2027
Testing & scale
DataSet v2 with Yandex Market
ML recommendations v2: uplift + explainability
Target function: margin / turnover / profit / stock-rotation
Execution Layer v1: semi-auto execution with approve
Patent filing
Seed round, Execution v1 rollout
2028
Autonomous AI loop & multi-MP
AI closed-loop v1: recommend → act → measure → re-train
Execution v2: action orchestrator (scenarios, budgets, schedules)
Multi-MP analytics: unified metrics, SKU comparison
Enterprise / on-prem: private contour, extended compliance
Scale-up round, strategic partnerships

We help strong cards grow, and help big brands stay in control of their SKUs

MPO makes marketplaces cleaner: it amplifies products the market actually wants, and gives large sellers real manageability over their own catalogue. We don’t "push at any cost" — we work where the product, content and price already deserve the buyer’s attention.

  • Support for strong cards

    We take on SKUs whose product, content and pricing are already on point. Our tools bring these cards the audience they deserve.

  • Control for large brands

    Big sellers watch dozens to hundreds of SKUs at once: where positions have sunk, where a competitor is advancing, where to reinforce the signal — and decide on data, not on gut feel.

  • Real buyers

    Every action runs through live accounts with real purchase history. Each buyout turns into a genuine experience and an honest review for the next customer.

  • Transparency at every step

    Clients see plan, pace and result in real time. Every number comes from marketplace APIs — no "magic multipliers," no grey reports.

FAQ

MPO without the marketing fog

How is MPO different from plain self-buyouts?
Self-buyouts are a one-off "lift one keyword at any cost" trick. MPO is a platform: we collect data from three-marketplace APIs, model the market, find the funnel bottleneck, then design the action scenario. Buyouts are just one execution tool.
Any risk of marketplace penalties?
Any activity done wrong carries risk. We work through real-account infrastructure with purchase history, keep pace and volume below anti-fraud thresholds, and monitor algorithm reaction in real time. 0 penalties across Trusty clients in 3 years (since 2023).
Who shouldn’t use MPO?
Cards with weak content, price 20%+ above market, awful reviews and zero margin. MPO accelerates what already works. If the product is weak, we recommend content and price hypothesis first.
How fast do results show?
First position shifts on days 7–10. Plateau on a keyword cluster in 4–6 weeks. ACoS drop and organic growth from week 3–4.
Do you need access to the seller cabinet?
No for basic diagnosis — article codes and links are enough. Yes for end-to-end analytics: API access (read+write for ads, read-only for finance).
How much does it cost?
The platform is still in active development, so there’s no public price list yet. For closed-beta waitlist members we lock in individual terms: first 3 months free, then a transparent model — commission on managed GMV plus an activation pack. Join the waitlist and we’ll send conditions the moment onboarding opens for you.
Waitlist

Be among the first on the new platform release

The platform is still in active development. Leave your contact — we’ll add you to the waitlist and invite you to the test among the first, the moment a new batch opens. No spam, no strings attached.

Priority onboarding
3 months free
Direct line to the product team
We reply within 2 business days. No spam — we only contact you about the MPO platform.