Video summary

Как починить падение SEO позиций после роста и понять в чем причина?

Main summary

Key takeaways

Technology

Why SEO Rankings Drop After Growth (and How to Fix It with On-Site Experiments)

The speaker explains that even if a landing page initially ranks well, rankings can later fall because search engines accumulate behavioral signals. The core issue is often that users don’t engage enough to satisfy engagement/bounce criteria, which leads to weaker performance over time.

Bounce / Engagement Differences (Examples)

Different search engines may interpret the same behavior differently:

  • Google: may treat it as a “bounce” if a user effectively visits only one page, even if they scroll a lot.
  • Yandex: may treat it as a bounce if a user returns to search and spends less than 10 seconds on the site.

Common real-world failure mode: a redesign makes the page look “prettier,” but users stop:

  • clicking through to a second page,
  • registering,
  • converting,

…and then go back to search.

Real Examples of What Goes Wrong (and What to Change)

The speaker references cases where rankings dropped after changes such as redesigns—for example:

  • 6th → 10th
  • 4th → top 10

Typical problems include:

  • awkward UX that discourages further actions,
  • heavy/confusing interfaces or animations that reduce engagement,
  • layout changes that prevent users from going deeper into the site.

Fix Strategy: Measure Behavior Correctly, Then Run Hypothesis-Driven Experiments

The guide describes a practical workflow.

1) Correct Tracking Setup (Yandex & Webvisor)

Key steps:

  • Set up Yandex Webvisor / Metrica correctly so internal team activity doesn’t pollute the data.
  • Filter out staff traffic
    • If it’s a solo operator: stay logged in consistently using one browser.
    • If multiple people are involved: use a common corporate IP (or VPN concept) and add an IP filter in Yandex Metrica.
  • Privacy / legal compliance
    • Configure Webvisor so you still get useful outputs (e.g., click maps and recordings) without improper data collection.
    • The speaker mentions avoiding issues related to Federal Law 162 (Russia) and GDPR by configuring counters appropriately.

Goal: collect reliable click maps, scroll/behavior recordings, and session patterns—without violating privacy rules.

2) Measure Scrolling + Keep an Experiment Journal

  • Use Webvisor/recordings and scroll maps.
  • Maintain an experiment log/journal (screenshots + what changed + before/after results).
  • Example approach: shared documentation (e.g., Google Docs) for iteration tracking.

3) Form Hypotheses in a Measurable Way

Hypothesis format:

“If we change X, then metric Y will improve by Z (time/pages/conversions), resulting in measurable money impact.”

The speaker warns against testing random “design improvements” without a measurable business hypothesis (e.g., adding animations/fonts without knowing their effect).

4) Define Engagement Thresholds as Goals

Focus on surpassing time/page thresholds:

  • > 3 seconds: many users leave within the first 3 seconds.
  • > 10 seconds: target to reduce “failures” and keep users engaged.
  • > 60 seconds: leads/registrations usually come from users who stay longer than this.

5) Test UI/Layout Changes Beyond the First Screen

Priority guidance:

  • Don’t only fix above-the-fold.
  • Change:
    • the sequence of blocks,
    • overall landing page structure,
    • transitions that guide users to deeper pages.

Measure improvements in:

  • session length
  • pages/views per session
  • reduced early exits/failures

Conversion Tactics to Increase Pages/Session Depth

Recommended patterns:

  • Move forms (applications/registration) to separate pages rather than embedding them on the same page or inside popups that don’t change the URL.
  • Create menu paths that route users to another section/landing page (e.g., industry-specific pages) to improve session depth.

Example: What “Works” (Speaker Walkthrough)

The speaker praises landing pages that use:

  • multiple “risk-free” CTA buttons placed periodically (e.g., every couple scroll screens),
  • buttons with different offers (free demo, call, estimate, trial, signup),
  • popups triggered at the right time with relevant CTAs,
  • benefit-focused messaging instead of “regalia/self-praise.”

Prioritization & Cost Management of Experiments

Experiments cost money and require traffic, so the speaker mentions prioritization approaches such as:

  • the RICE model (explicitly referenced as Rй in subtitles)
  • plus cost-of-implementation / ownership considerations

Example cost-of-ownership consideration:

  • a chatbot with a live person supporting onboarding (noted as an ongoing monthly expense).

How to Run Tests Effectively

Rules of thumb:

  • Run tests on at least two landing pages to reduce randomness/representativeness issues.
  • Prefer pages already in top 10 (or close) so there’s enough measurable traffic; otherwise tests may take too long.
  • Use sufficient volume: aim for at least tens/hundreds of sessions.
  • Test multiple hypotheses per landing page (recommended: at least two per landing page).
  • Use PPC if you need faster traffic generation.

Recommended Tools & Final Workflow

Tools

  • Yandex Webvisor / Yandex Metrica (correctly configured)
  • Microsoft Clarity (mentioned as an alternative and described as “cooler” by the speaker)

Continuous Improvement Loop (“Heidi cycle” phrasing)

  1. measure
  2. formulate hypotheses
  3. implement changes
  4. keep tracking
  5. launch new tests
  6. repeat continuously

Main Speakers / Sources

  • Dmitry: main speaker; runs SEO services and a Telegram SEO channel; speaks throughout.
  • Lesha / Alexey: mentioned as an experiment/journal example; later invited to share results, but details aren’t fully visible in the provided subtitles.

Original video