Video summary

БАЗА OZON - РАНЖИРОВАНИЕ ОТ А ДО Я

Main summary

Key takeaways

Business

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:

  1. Increasing product visibility, and
  2. 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%
  • 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”)

Original video