Video summary
БАЗА OZON - РАНЖИРОВАНИЕ ОТ А ДО Я
Main summary
Key takeaways
What’s happening & why it matters (business framing)
- The speaker argues sellers struggle on Ozon because they don’t understand the search ranking algorithm structure and how organic + paid promotion interact.
- They claim the ranking system has stabilized recently: no major changes planned as of Feb 2026 (with the last major update noted as fall 2025, introducing an experimental 5th stage of paid promotion ranking).
- They position the update as more than “SEO”: it’s a compounding system where mistakes can cause momentum reversal (a “snowball effect”):
- If you hit the right query semantics, logistics, and market price + maintain conversion, visibility compounds upward.
- If you stumble, you may lose ad effectiveness/traffic, conversions drop, and rankings fall rapidly.
Ozon ranking algorithm: the 5-stage playbook (core structure)
The speaker describes five stages, with the business implication that different levers matter at different stages.
Stage 1 — Query form: indexed candidate selection
Key idea: your catalog content must be indexed for relevant queries.
- Indexed fields matter:
- Name
- Category name
- Brand
- Descriptions
- Characteristics
- Synonyms/semantic expansion:
- Ozon may index you for queries you didn’t intend, because it treats some queries as synonyms.
- Result: you can be “eligible” for non-target traffic.
Stage 2 — Base layer (textual relevance + filtering)
Key idea: Ozon selects promising candidates from a huge pool and can cut off low-performing products.
- Text relevance based on:
- proximity of query words to product attributes (name/brand/category/description/colors)
- considers dynamic attributes like order activity and delivery speed
- Filtering effect:
- If you have no orders at all at this stage, you may be “cut off” from visibility.
- Practical mitigation mentioned:
- Use targeted ultra-low-frequency/branded queries where you still have some relevance.
- If indexed but not converting, consider self-buyouts (presented as an edge-case/launch tactic), or “attach one card to another so it parasitizes traffic” (implying cross-card linkage/indirect order generation).
Stage 3 — Semantic blocks (machine-learning purchase probability)
Key idea: a ML model estimates probability a customer will buy.
- Purchase probability can vary by:
- region
- query semantics (same product → different probability depending on what is searched)
- Focal targeting warning:
- You should choose one focus word / one query point rather than buying out a wide range—otherwise you dilute conversion and damage overall performance.
- Personalization:
- Search results are said to have minimal personalization; some personalization exists on recommendation shelves, but search is mostly pattern-based on behavior.
Stage 4 — Boosting (contextual weighting multipliers)
Key idea: final organic quality estimates get multiplied by “boost” factors.
- Organic ranking factors influence an overall organic assessment in a thin scored range (speaker says 0–1 style).
- Main organic factor weights claimed (organic influence):
- Text relevance/semantic hit: up to roughly 20–40% (speaker gives “20–40 on average”)
- Product sales: discussed as an ~28 days window (uncertainty noted: could be 7–14 days or more recent orders weighted higher)
- Other major factor contributions cited:
- Popularity (funnel actions): CTR/cart/favorites journey; may be reduced due to suspected manipulation
- Price factor: 5–15% generally, but in categorical query contexts it can dominate
- Delivery speed: directly affects rank and conversion; speaker describes “supermultiplier” effect
- Reviews/rating: smaller and less clear; speaker notes lack of transparency compared to WB
Stage 5 — Paid promotion ranking (the “experimental fifth stage”)
Key idea: paid promotion now materially affects the final position via a formula, not just “ad above organic.”
- In addition to organic scoring, ads contribute to a final aggregate score shown in the Product Visibility tool.
- The speaker says the final ranking mixes:
- Organic score (from earlier stages + boosting)
- Promotion score (built from ad mechanics like CTR / conversion / payment-for-order)
- Important business implication:
- Effective ads can compensate for weaker organic positions because the final score is a weighted combination of organic and promotion.
Ranking factors & KPI-like mechanics (what to measure)
The speaker provides multiple factor definitions that function like KPIs.
1) Text relevance (SEO-like)
- Definition: match quality between query and product attributes
- Business action:
- Ensure titles/descriptions/categories contain the exact semantic intent for your focus queries
- Don’t rely on “general category SEO” if you’re competing on narrow intent queries
2) Product sales (conversion to orders + sales volume concept)
- Sales is computed from past sales, with conversion + orders over a time window:
- Speaker suggests 28-day window as Ozon’s common example
- Key difference vs Wildberries (WB):
- Ozon is framed as more about conversion/order probability, while WB is more about sales value in rubles (with recency weighting), enabling different “self-buyout” behaviors.
3) Popularity (query-specific funnel actions)
- Popularity tied to request context, not overall store averages:
- impressions → view → add to cart → favorites
- They claim popularity can be gamed with bots in high-competition niches, which is why weight may be adjusted down.
4) Price and price dynamics
- Factor range stated: ~5–15%, but in categorical queries price can become decisive.
- They explicitly warn: price-to-query fit can largely determine visibility for high-frequency categorical queries.
5) Delivery speed / logistics
- Direct rank effect and indirect conversion effect (customers dislike waiting).
- Contextual weighting:
- For remote regions, delivery/logistics gets higher coefficients.
- Notable logistics/fulfillment insight cited:
- FBO vs FBS affects Ozon credit attractiveness/limits (see below).
6) Reviews / rating
- Based on product rating and number of reviews.
- Speaker notes uncertainty and differences vs WB (where recency may matter more, per their comparison).
Ad strategy framework: how to win under Stage 5
The speaker frames paid ads as both:
- Increasing product visibility, and
- Affecting final ranking through a compound score.
“Promotion assessment” mechanics (what ads feed)
- Promotion assessment is a sum of two ad tools:
- a CTR-like component (probability of click on product/order)
- an order conversion/payment component (payment for an order; conversion to ordering)
- They argue summation vs multiplication means effective ads can still dominate final score.
The “democracy coefficient” (organic vs ads weighting)
- They describe regulatory coefficients controlling organic vs promotion influence (conceptually called a “democracy coefficient”).
- Example impact described:
- If ad account/rating drops to ~0, the ad coefficient effectively collapses → you lose paid influence.
Practical consequence
- If your ads have strong CTR/conversion, the speaker claims ad effectiveness can outweigh organic by a multiple (they estimate ~3–4x influence in their example reasoning).
Concrete case examples / observations
Ozon vs WB issuance differences
- Ozon top placements: often cheaper-but-selling units (volume/unit sales and conversion).
- WB top placements: often higher average check products because ruble sales in recency matters more.
- They argue this conceptual split explains why sellers see results differ between marketplaces.
Coffee price-segment anecdote (metrics-ish)
- They cite public-like order volumes by price segment (weekly scale; exact reliability uncertain due to speech form):
- Around 2.5–3,000 rubles as the highest order-volume segment
- ~500 rubles segments have lower ruble volume but high unit counts (e.g., “100M” vs higher segment “120M/week” type statements)
Logistics & credit limits (execution and operations)
- They claim Ozon’s fintech credit limit for FBO is roughly:
- ~one month turnover
- For FBO + FBS mix, they claim FBS can reduce credit attractiveness (interpreted as “less reliable counterparty”).
- Actionable operational advice implied:
- If you want better credit/limits, reduce/disable FBS (or ensure faster delivery and avoid long assembly times via FBS).
Time windows & quantitative targets mentioned
- Last major algorithm change: fall 2025 (experimental 5th stage of paid ranking)
- Status: “no major changes planned” as of Feb 2026
- Organic sales KPI window: often treated as 28 days (speaker notes uncertainty: might be 7/14 and may weight recency)
- Organic influence share: “organic influence up to ~40%” (also gives average 20–40% for text relevance)
- Price factor:
- general: ~5–15%
- in categorical queries: may rise to around ~40% (speaker’s scenario-based estimate)
- Delivery boosting:
- speaker claims delivery boosting weight shifted historically:
- 2022: delivery boosting up to 32%
- current: delivery mentioned around ~15%
- speaker claims delivery boosting weight shifted historically:
- Promotional participation:
- If “automatic participation in promotion” is enabled, they mention an additional +10% (as an extra bonus/assessment input).
Actionable recommendations (from the speaker’s logic)
- Choose focus semantics:
- define one focus query/word for “buyout” or promotion; don’t spread across too many queries and dilute conversion.
- Fix indexing fundamentals:
- ensure key fields (name/category/brand/description/specs) are filled because synonyms can still expose you to off-intent queries.
- Compete on the right price segment:
- if you target high-frequency categorical queries, align price with market expectations; cheap competitive pricing may be required for visibility.
- Prioritize logistics speed:
- delivery affects rank and conversion; remote regions get extra logistics weighting.
- Don’t rely on organic alone if Stage 5 dominates:
- maintain ad health (CTR/conversion and account standing), because ads heavily influence the final score.
- Avoid “momentum breaks”:
- losing ads/conversion can cause a rapid rollback due to compounding algorithmic effects.
Presenters / sources
- Vladimir Blakhin (speaker)
- NeLex agency (source/affiliation mentioned: “this is the Neleks agency”)