Pokročilá analytika

Pokročilá marketingová analytika ide za základné metriky. Zahŕňa attribution modeling, prediktívnu analytiku, Marketing Mix Modeling a rigorózne experimentovanie. Táto fáza je pre analyticky orientovaných marketérov.

Prečo Pokročilá Analytika?

Základná analytika:

• Čo sa stalo? (reporting)
• Koľko? (metriky)

Pokročilá analytika:

• Prečo sa to stalo? (diagnosis)
• Čo sa stane? (predikcia)
• Čo by sme mali robiť? (prescriptive)

“Without data, you’re just another person with an opinion.” — W. Edwards Deming

1. Marketing Attribution

Attribution určuje, ktoré touchpointy prispeli ku konverzii.

Attribution Modely:

Multi-Touch Attribution (MTA):

Customer Journey:
Google Ad → Blog → Email → Social → Demo → Purchase
   ↓         ↓       ↓       ↓       ↓
  20%      15%     25%     10%     30%   (Data-Driven)

Attribution Challenges:

1. Cross-device: User na mobile, konverzia na desktop
2. Offline impact: TV, print, word-of-mouth
3. Long cycles: B2B môže byť mesiace
4. Walled gardens: Facebook, Google majú limited data sharing
5. Privacy: Cookie deprecation, iOS ATT

GA4 Attribution:

GA4 používa data-driven attribution ako default:

Attribution Settings:
├── Reporting attribution model: Data-driven
├── Lookback window: 30/90 days
└── Conversion credit: Cross-channel

2. Marketing Mix Modeling (MMM)

MMM je ekonometrický prístup k meraniu marketing efektivity na agregovanej úrovni. V 2025 zažíva renaissance vďaka privacy regulations.

Čo je MMM?

Sales = f(TV, Digital, Price, Seasonality, Macro...)

Regression model určujúci contribution každého faktora.

Prečo MMM v 2025?

Privacy-First Measurement:

• NEVYŽADUJE user-level tracking ani cookies
• Funguje s agregovanými dátami
• GDPR/CCPA compliant by default
• Imúnne voči iOS ATT a cookie deprecation

Strategické Rozhodovanie:

• Dlhodobé plánovanie budget allocation
• Offline + online channel impact
• Brand building vs performance balance
• Incrementality measurement

MMM vs MTA:

MMM Output Example:

Channel ROI Analysis:
─────────────────────────────────
Channel     Spend     Incremental   ROI
TV          €500K     €1.2M         2.4x
Digital     €300K     €900K         3.0x
OOH         €100K     €180K         1.8x
Print       €50K      €40K          0.8x
─────────────────────────────────

Modern MMM Tools:

• Google Meridian (open source, AI-powered)
• Meta Robyn (open source, R package)
• Nielsen Marketing Mix (enterprise)
• Analytic Partners (enterprise)
• Mutinex (privacy-first platform)

Triangulation Approach 2025:

Najlepšie riešenie = kombinácia metodík, nie jedna vs druhá.

TRIANGULATION = MMM + MTA + Incrementality Testing

MMM                     MTA                    Incrementality
(Strategické)           (Taktické)             (Validácia)
│                       │                      │
├─ Long-term planning   ├─ Campaign optimization ├─ Holdout tests
├─ Budget allocation    ├─ Real-time insights   ├─ Geo experiments
├─ Brand + performance  ├─ User journey         ├─ Lift studies
└─ Offline + online     └─ Digital touchpoints  └─ Causal proof

           ↓                    ↓                     ↓
                    COMPREHENSIVE VIEW

Best Practice:

• Use MMM na steering the ship’s direction (quarterly planning)
• Use MTA na course-correction (weekly optimization)
• Use incrementality testing na validation (ad-hoc)

MMM 2025 Best Practices:

1. Combine with experiments: MMM + incrementality testing pre validáciu
2. Integrate systems: MMM + attribution + financial data
3. Real-time updates: Quarterly → monthly → continuous modeling
4. Clear recommendations: Translate model outputs do C-level decisions
5. Organizational adoption: Embed MMM do planning a budgeting workflows

3. Predictive Analytics

Customer Lifetime Value (CLV) Prediction:

Simple CLV:

CLV = Average Order Value × Purchase Frequency × Customer Lifespan
CLV = €50 × 4/year × 3 years = €600

Predictive CLV (ML-based):

Features:
├── Historical purchase behavior
├── Recency, Frequency, Monetary
├── Product categories purchased
├── Channel engagement
├── Demographics
└── Customer service interactions

Model: Predikuje budúcu hodnotu zákazníka

CLV Use Cases:

Churn Prediction:

Churn Probability Model
─────────────────────────
Input Features:
├── Days since last activity
├── Engagement trend
├── Support tickets
├── Payment failures
├── Product usage
└── Demographic risk

Output: Probability (0-100%)
Action: If >70%, trigger retention campaign

Propensity Modeling:

4. A/B Testing & Experimentation

A/B Test Basics:

Control (A)    vs    Variant (B)
50% traffic         50% traffic
Original            Change

Measure: Conversion rate, revenue, etc.
Statistical test: Is difference significant?

Sample Size Calculation:

Required Sample Size depends on:
├── Baseline conversion rate
├── Minimum Detectable Effect (MDE)
├── Statistical significance (α, typically 5%)
└── Statistical power (1-β, typically 80%)

Example:
Baseline: 3% conversion
MDE: 10% relative improvement (0.3% absolute)
→ Need ~35,000 visitors per variant

Statistical Significance:

p-value < 0.05 = Statistically significant
(95% confident result is not random)

Confidence Interval:
If CI doesn't include 0, result is significant

Common A/B Testing Mistakes:

1. Peeking: Stopping test early when results look good
2. Low sample: Not enough data for significance
3. Multiple comparisons: Testing many variants without correction
4. Selection bias: Non-random traffic splitting
5. External factors: Seasonal effects, promotions

Experimentation Culture:

Hypothesis → Design → Execute → Analyze → Learn → Iterate

Document everything:
├── What we tested
├── Why we tested it
├── What we learned
└── What we'll do next

5. Marketing Research Methods

Marketing research je základ data-driven rozhodnutí. Kombinuje kvalitatívne a kvantitatívne metódy pre pochopenie trhu a zákazníkov.

Kvalitatívne vs Kvantitatívne:

Best Approach: Kombinuj oba - kvalitatívne pre exploration, kvantitatívne pre validation.

Kvalitatívne Metódy:

1. Focus Groups

Čo: 6-10 účastníkov diskutuje tému s moderátorom.

Kedy použiť:

• Nový produkt concept testing
• Brand perception research
• Customer pain points exploration

Výhody: Rich insights, group dynamics Nevýhody: Groupthink, facilitator bias

2. In-Depth Interviews (IDI)

Čo: 1:1 rozhovor s respondentom (45-90 min).

Kedy použiť:

• B2B research (executives)
• Sensitive topics
• Deep behavioral understanding

Výhody: Detailed insights, no peer pressure Nevýhody: Časovo náročné, expensive

3. Observational Research (Ethnography)

Čo: Pozorovanie správania v prirodzenom prostredí.

Kedy použiť:

• Retail shopping behavior
• Product usage patterns
• UX research

Výhody: Real behavior (not claimed) Nevýhody: Observer effect, interpretation bias

4. Online Communities

Čo: Dlhodobá online group diskusia (days/weeks).

Kedy použiť:

• Continuous feedback
• Co-creation s zákazníkmi
• Trend monitoring

Výhody: Longitudinal insights, scale Nevýhody: Self-selection bias

Kvantitatívne Metódy:

1. Surveys (Dotazníky)

Čo: Štandardizované otázky pre veľký sample.

Kedy použiť:

• Market sizing
• Customer satisfaction (CSAT, NPS)
• Concept testing validation

Typy otázok:

• Multiple choice
• Rating scales (Likert)
• Ranking
• Open-ended (limitovane)

2. Experiments & A/B Testing

Čo: Kontrolované testy kauzálneho vplyvu.

Kedy použiť:

• Price testing
• Message testing
• Product feature validation

Výhody: Causal proof Nevýhody: Controlled environment ≠ real world

3. Analytics & Behavioral Data

Čo: Analýza digital footprint (GA, CRM).

Kedy použiť:

• User behavior patterns
• Funnel optimization
• Cohort analysis

Výhody: Large sample, real behavior Nevýhody: Correlation ≠ causation

Survey Design Best Practices:

1. Define Clear Objectives:

• Čo chceme zistiť?
• Komu sa pýtame?
• Ako použijeme výsledky?

2. Sample Strategy:

• Random sampling: Reprezentatívny sample
• Purposive sampling: Specific criteria (napr. existing customers)
• Quota sampling: Proportional representation (vek, pohlavie)
• Snowball sampling: Referrals (niche audiences)

3. Question Design:

• Jasné, jednoznačné formulácie
• Avoid leading questions (“Don’t you think…?”)
• Avoid double-barreled (“Quality and price?”)
• Logical flow (easy → complex)
• Include “Don’t know” option

4. Sample Size:

Confidence Level: 95% (standard)
Margin of Error: ±5% (acceptable)
Population: 10,000
→ Required sample: ~370 respondents

Online calculator: SurveyMonkey Sample Size Calculator

5. Response Rate:

• Email surveys: 10-30%
• Online panels: 2-5%
• Incentivized: 30-50%
• In-app: 5-20%

Statistical Analysis Basics:

Descriptive Statistics:

• Mean (priemer), Median (stred), Mode (najčastejší)
• Standard deviation (variabilita)
• Frequency distribution

Inferential Statistics:

• Hypothesis testing (t-test, chi-square)
• Confidence intervals
• p-value (< 0.05 = significant)
• Correlation vs causation

Segmentation Analysis:

• Cluster analysis (grouping similar respondents)
• Factor analysis (reduce variables)
• RFM segmentation

SK/CZ Research Agencies:

Research Process:

1. DEFINE PROBLEM
   └── Research objectives, questions

2. DESIGN RESEARCH
   └── Methodology, sample, questionnaire

3. COLLECT DATA
   └── Fielding, quality control

4. ANALYZE DATA
   └── Cleaning, coding, statistics

5. REPORT FINDINGS
   └── Insights, recommendations, action plan

6. IMPLEMENT & MEASURE
   └── Apply learnings, track impact

Research Ethics:

• Informed consent: Respondents vedia účel
• Anonymity: Protect personal data (GDPR)
• No deception: Honest about research purpose
• Right to withdraw: Can exit anytime
• Data security: Secure storage and handling

6. Dashboard & Visualization

Marketing Dashboard Types:

Executive Dashboard Elements:

┌─────────────────────────────────────────┐
│  MARKETING DASHBOARD - WEEK 47          │
├─────────────────────────────────────────┤
│  Revenue: €1.2M (+15% vs LY)            │
│  Pipeline: €3.5M (+22%)                 │
│  MQLs: 450 (target: 400) ✓              │
│  CAC: €85 (target: €100) ✓              │
├─────────────────────────────────────────┤
│  [Channel Performance Chart]            │
│  [Funnel Conversion Chart]              │
├─────────────────────────────────────────┤
│  Alerts: Campaign X underperforming     │
└─────────────────────────────────────────┘

Visualization Best Practices:

Do:

• Use appropriate chart types
• Highlight key metrics
• Show trends, not just snapshots
• Include context (vs target, vs LY)

Don’t:

• Overcrowd with data
• Use 3D charts
• Hide bad news
• Use inconsistent scales

Tools:

7. Data Infrastructure

Marketing Data Stack:

DATA SOURCES
├── Website (GA4)
├── CRM (HubSpot)
├── Ads (Google, Meta)
├── Email (Klaviyo)
└── Product (Amplitude)
           │
           ↓
    ETL / INTEGRATION
    (Fivetran, Stitch, Airbyte)
           │
           ↓
    DATA WAREHOUSE
    (BigQuery, Snowflake, Redshift)
           │
           ↓
    TRANSFORMATION
    (dbt)
           │
           ↓
    VISUALIZATION
    (Looker, Tableau)

CDP (Customer Data Platform):

CDP Functions:
├── Data Collection (all sources)
├── Identity Resolution (merge profiles)
├── Segmentation (audience building)
├── Activation (sync to channels)
└── Analytics (customer insights)

Examples: Segment, Bloomreach, Rudderstack

Data Quality:

Data Quality Dimensions:
├── Completeness: All required fields
├── Accuracy: Correct values
├── Consistency: Same across systems
├── Timeliness: Up to date
└── Validity: Correct format

8. Advanced Metrics

Marketing Efficiency Metrics:

Incrementality:

Incrementality = Conversions with marketing - Conversions without

Methods to measure:
├── Holdout groups (gold standard)
├── Geo tests (regional holdouts)
├── Ghost ads (PSA tests)
└── Matched market tests

Marketing Contribution:

Marketing-Sourced Pipeline: First touch = marketing
Marketing-Influenced Pipeline: Any touch = marketing

Attribution:
├── Marketing Sourced: €1M (30%)
├── Marketing Influenced: €2M (60%)
└── Pure Sales: €0.5M (10%)

9. SK/CZ Analytics Landscape

Lokálne špecifiká:

• GDPR compliance: Prísnejšie než US
• Consent management: Cookie banners required
• Local tools: SmartLook (CZ), Exponea/Bloomreach (SK origin)
• Language: Slovenské/české dashboardy pre stakeholders

SK/CZ Analytics Tools:

Praktické cvičenie: Attribution Analysis

Krok 1: Export Data

Z GA4 exportuj Multi-Channel Funnel report.

Krok 2: Compare Models

Porovnaj attribution podľa:

• Last click
• First click
• Linear
• Data-driven

Krok 3: Identify Discrepancies

Ktoré kanály sú under/over-valued podľa modelu?

Krok 4: Recommend Actions

Na základe data-driven attribution, ako by ste realokovali budget?

Checklist: Advanced Analytics Readiness

Základy:

• GA4 správne nastavený
• Basic attribution understanding
• Regular reporting cadence
• A/B testing capability

Intermediate:

• Multi-touch attribution implemented
• Basic predictive models (CLV)
• Dashboards pre stakeholders
• Experimentation culture

Advanced:

• Marketing Mix Modeling
• Propensity models
• CDP implemented
• Incrementality testing

Kľúčové pojmy

Ďalšie kroky

Po zvládnutí pokročilej analytiky pokračuj na:

• Fáza 14: Budúce trendy – AI v analytike
• Fáza 12: Marketing Leadership – data-informed decisions

Analytika je základ pre evidence-based marketing. Investujte do dát a skills na ich interpretáciu.