SpotLift vs. The Competition: TVSquared, EDO, Realytics, iSpot & Recast
A technical, economic, and architectural comparison of SpotLift against legacy TV attribution vendors (TVSquared by Innovid, Realytics, iSpot.tv, EDO), modern Media Mix Modeling platforms (Recast, Google Meridian, Meta Robyn), and digital multi-touch attribution (MTA) tools.
1. Executive Master Comparison Matrix
| Feature / Capability | SpotLift (spotlift + Pro) | TVSquared by Innovid | Realytics (smartclip) | EDO (Search Intent) | iSpot.tv | Top-Down MMM (Recast / Meridian) |
|---|---|---|---|---|---|---|
| Delivery Model | Python Library & CLI (In-VPC) | Closed SaaS Web Portal | Closed SaaS Web Portal | Syndicated Enterprise Data | Closed SaaS Web Portal | Python / R or Managed SaaS |
| Methodological Transparency | ✅ 100% Open-Core Code & Published Math | ❌ Proprietary Black Box | ❌ Proprietary Black Box | ❌ Proprietary Black Box | ❌ Proprietary Black Box | ✅ Open Math / Whitepapers |
| Amazon Retail Media (AMC) | ✅ Native AMC Clean Room & Halo Engine | ❌ Blind (Cannot place pixels on Amazon) | ❌ Blind (Cannot place pixels on Amazon) | ❌ Blind to conversions (Search only) | ❌ Blind (Cannot place pixels on Amazon) | ⚠️ Coarse weekly aggregate correlation |
| New-To-Brand (NTB) Tracking | ✅ Native AMC NTB Units & GMV Ingestion | ❌ None | ❌ None | ❌ None | ❌ None | ❌ None |
| Paid Brand PPC Cannibalization | ✅ Mathematically audits Sponsored Brands tax | ❌ None | ❌ None | ❌ None | ❌ None | ❌ None |
| Signed Residuals (No Noise Bias) | ✅ Preserved (clip_negative=False) | ❌ Clamped to 0 (Upward Bias) | ❌ Clamped to 0 (Upward Bias) | ⚠️ N/A (Indexed Search Ratio) | ❌ Clamped to 0 (Upward Bias) | ✅ Signed residuals preserved |
| Prime-Time Spot Collisions | ✅ Shifted-Gamma EM Deconvolution | Opaque proprietary heuristic | Proportional / station-weight split | Syndicated single-spot indexing | ACR-derived household matching | ⚠️ Coarse weekly aggregation |
| Budget Portfolio Optimization | ✅ Constrained KKT Marginal Optimizer (mCPIV + MMM draws) | ❌ Naive average CPV sorting | ❌ Naive average CPV sorting | ❌ Benchmark indexing only | ❌ Naive average CPV sorting | ✅ Budget curve solver (Macro-level) |
| Two-Way MMM Integration | ✅ Causal lower bounds → MMM & MMM-anchored station optimizer | ❌ None (isolated portal) | ❌ None (isolated portal) | ❌ None | ❌ None | ⚠️ Macro channel-level only |
| Tactical Schedule Optimization | ✅ Spot, daypart & creative-level | ✅ Spot-level | ✅ Spot-level | ✅ Spot-level | ✅ Spot-level | ❌ Macro channel-level only |
| Data Privacy & GDPR | ✅ Zero 3rd-party tags; 100% inside your VPC | ❌ 3rd-party JavaScript tracker on brand.com | ❌ 3rd-party JavaScript tracker on brand.com | ✅ Privacy-safe (Syndicated outside-in) | ❌ 3rd-party tracking tag & ACR sync | ✅ In-warehouse aggregate data |
| Pricing & Contract Terms | MIT Open-Source / Flexible Pro Pass (€249) | €3,500 – €6,000+/mo (12m lock-in) | €2,500 – €5,000+/mo (Annual contract) | $80,000 – $150,000+/yr (Enterprise) | $60,000 – $150,000+/yr (Enterprise) | $3,000 – $10,000+/mo (Managed SaaS) |
| Procurement Speed | Immediate (pip install spotlift) | 3–6 months legal, DPA & IT review | 2–4 months contracting | 3–6 months enterprise vetting | 3–6 months enterprise vetting | 1–3 months onboarding |
2. Deep-Dive Provider Comparisons
SpotLift vs. TVSquared by Innovid
TVSquared (acquired by Innovid in 2022) is the incumbent platform used by traditional media holding groups (GroupM, Omnicom, Publicis).
- The TVSquared Approach: Ingests broadcaster log files via portal upload and correlates them against web traffic captured by an Innovid/TVSquared tracking tag pasted into the client's website header.
- Where TVSquared Falls Short:
- Retail Media Blindness: Zero visibility into Amazon, Otto, or retail media marketplaces. If you sell consumer goods or health products where viewers purchase via the Amazon mobile app, TVSquared records zero lift.
- Contractual Lock-In: Requires multi-year enterprise contracts (€3,500 to €6,000+ per month) with steep implementation fees.
- Naive Reallocation Trap: Recommends moving budget simply to the station with the lowest historical average CPV, ignoring audience wear-out, diminishing marginal returns, and contractual agency group commitments.
- The SpotLift Advantage: Local Python execution with zero 3rd-party tag latency, native Amazon Marketing Cloud (AMC) clean-room integration, and a mathematically rigorous constrained portfolio optimizer.
SpotLift vs. EDO (Entertainment Data Oracle)
EDO pioneered measuring Search Engagement Volume (SEV)—tracking surges in Google search activity within 5 minutes of a TV commercial airing.
- The EDO Approach: Based on the sound premise that mobile search behavior is the immediate proxy for TV ad intent. When an ad airs, viewers pull out their phones and search.
- Where EDO Falls Short:
- Search Index, Not Business Conversions: EDO tracks search volume ratios on Google, but has zero visibility into actual cart checkout, GA4 conversion events, or Amazon sales. High search engagement does not guarantee high order value or profitable CPA.
- Syndicated Outside-In Pricing: Charges $80,000 to $150,000+ per year for syndicated category benchmark data.
- Blind to Marketplace Sales: Cannot tell you whether TV-stimulated searchers bought on Amazon or were intercepted by expensive competitor conquesting ads.
- The SpotLift Advantage: Connects broadcast logs directly to first-party conversions (GA4 revenue, purchases, and AMC marketplace GMV) for a fraction of the cost, while auditing branded search PPC cannibalization.
SpotLift vs. Realytics (smartclip / RTL Group)
Realytics is the leading European TV-to-web attribution and programmatic TV platform headquartered in France and popular across DACH.
- The Realytics Approach: Employs direct baseline interpolation (evaluating clean pre/post windows 20 minutes before and after spots) to calculate Cost-Per-Visit (CPV) across European broadcast networks.
- Where Realytics Falls Short:
- Collision Failure in Prime-Time: Standard linear interpolation breaks down during synchronized prime-time commercial breaks (where 20% to 50% of airings collide within 8 minutes).
- SaaS Dashboard Silo: Results live inside a closed web portal. Data scientists cannot integrate the attribution logic into automated dbt models, Airflow DAGs, or Python analytics workflows without manual exports.
- Zero Marketplace Halo: Completely ignores Amazon retail conversions and New-To-Brand customer acquisition.
- The SpotLift Advantage: Native support for European broadcast schemas (
Sendetermin,Sender,Netto) with Shifted-Gamma EM deconvolution for overlapping collisions, open-core Python automation, and AMC clean room support.
SpotLift vs. iSpot.tv
iSpot.tv is a US media measurement giant specializing in smart TV Automated Content Recognition (ACR) and business outcome attribution.
- The iSpot Approach: Second-by-second ACR fingerprinting from millions of opted-in smart TVs matched to digital web visits via household IP graphs.
- Where iSpot Falls Short:
- Eroding Identity Spine: Matching household smart TVs to second-screen mobile visits via IP addresses is increasingly degraded by Apple Private Relay, VPNs, IPv6 rotation, and cookie deprecation.
- Heavy Procurement Barrier: Minimum annual contracts start at $60,000 to $150,000+, with 3–6 months of InfoSec and legal review.
- US-Centric: ACR panel infrastructure is heavily concentrated in the US; in Europe (GDPR), ACR-to-household matching faces severe regulatory hurdles.
- The SpotLift Advantage: Causal and econometric-first, not panel-dependent. Works globally on any broadcast schedule (Schaltplan) without transmitting sensitive IP addresses across external networks.
SpotLift vs. Top-Down Media Mix Modeling (Recast, Google Meridian, Meta Robyn)
Modern econometricians often express deep skepticism toward tactical TV "spike models." Recast and leading media scientists have published prominent critiques of legacy TV attribution:
- The "Phantom Conversions" Critique: Legacy SaaS platforms non-linearly clamp negative minute differences to zero (
). By Jensen's inequality, this turns random background traffic noise into strictly positive phantom lift. - The "Ignoring the Baseline" Critique: 5-minute spike models only capture immediate direct response, missing the weeks of brand equity, word-of-mouth, and retail baseline lift that TV advertising builds.
- The "Diminishing Returns" Critique: Naive tools assume linear scaling and tell marketers to dump budget into whatever channel had the lowest average CPV last week.
How SpotLift Answers the Recast Critique:
- Unbiased Signed Residuals: SpotLift preserves signed lift (
clip_negative=Falseby default) so background noise cancels out symmetrically across spots and stations. - Empirical Holdout Proof: In regional campaigns (
generate_us_state_dataset), unexposed control states receiving zero TV spend evaluate to exactlymean lift. - Diminishing Marginal Returns:
spotlift optimizemodels concave power-law response curves (), equalizing marginal CPIV ( ) rather than historical averages. - Two-Way SpotLift ↔ MMM Bridge (
export_causal_lower_bound&optimize_tv_with_mmm): SpotLift does not compete with MMM; it operates bidirectionally with MMM. Immediate 0–8 min lift exports as a causal lower bound (or latent immediate-share likelihood viaattach_pymc_lower_bound) with support-matched priors inexport_mmm_priors, while macro MMM posteriors feed back intooptimize_tv_with_mmmanddisaggregate_mmm_contributionsto anchor station/creative optimization:
python
import spotlift as sl
# Direction 1: Export causal lower bound & support-matched priors for MMM
evidence = sl.export_causal_lower_bound(attributed_spots, mmm_channel_name="tv_broadcast")
priors = sl.export_mmm_priors(attributed_spots, default_elasticity=0.70)
print(priors["google_meridian_priors"])
# Direction 2: Anchor SpotLift station budget optimization to macro MMM posterior
mmm_opt = sl.optimize_tv_with_mmm(attributed_spots, mmm_results=mmm_posterior_df)3. The 3 Architectural Moats of SpotLift
┌──────────────────────────────┐
│ SPOTLIFT THREE-PILLAR │
│ ARCHITECTURAL MOAT │
└──────────────┬───────────────┘
│
┌────────────────────────────────────┼────────────────────────────────────┐
▼ ▼ ▼
┌──────────────────────────────┐ ┌──────────────────────────────┐ ┌──────────────────────────────┐
│ 1. OMNICHANNEL & AMAZON │ │ 2. LOCAL IN-VPC ZERO │ │ 3. CONSTRAINED MARGINAL │
│ AMC CLEAN ROOM SUITE │ │ DATA EGRESS & GDPR │ │ PORTFOLIO OPTIMIZER │
├──────────────────────────────┤ ├──────────────────────────────┤ ├──────────────────────────────┤
│ • Ingests AMC pre/post DPV │ │ • 100% Python library / CLI │ │ • Models diminishing returns │
│ • Captures 36h retail halo │ │ • No raw customer data leaves│ │ curves R(s) = α·s^β │
│ • Uncovers New-To-Brand (NTB)│ │ your cloud warehouse / VPC │ │ • Equalizes marginal CPIV │
│ • Audits Sponsored Brands │ │ • Zero 3rd-party web pixels │ │ • Respects contractual agency│
│ PPC cannibalization tax │ │ • Eliminates 6-month InfoSec │ │ broadcaster commitments │
└──────────────────────────────┘ └──────────────────────────────┘ └──────────────────────────────┘4. Summary Recommendation: When to Choose SpotLift
┌──────────────────────────────────────────────────────────────────────────────────┐
│ Choose SpotLift When: │
│ • You have a data scientist, analytics engineer, or quantitative media planner. │
│ • You sell products on Amazon and need to capture retail halo, NTB, & PPC waste. │
│ • You demand peer-auditable mathematical transparency without black-box claims. │
│ • You refuse to pay €3,000–€8,000/mo on 12-month lock-in vendor contracts. │
│ • You operate in the EU (DACH, UK, France) under strict GDPR data requirements. │
│ • You run Media Mix Modeling (PyMC/Robyn/Meridian) and want two-way integration.│
└──────────────────────────────────────────────────────────────────────────────────┘Ready to audit your campaign?
- Try the Open-Core Engine: 5-Minute Quickstart
- Connect Spot Attribution & MMM: Two-Way MMM Integration
- Inspect the Mathematics: Econometric Methodology
- Explore Budget Optimization: Budget & MMM Optimizer
- Explore Prime-Time Deconvolution & Amazon AMC: SpotLift Pro