When the self-serve platform isn't the right fit,we run the analysis for you.
Long-form video. Full campaign sweeps. Raw camera files. Multi-language SA creative. Custom audience cohorts. Spatial and immersive environments. Same Buyology neuroscience foundation — without the platform's self-serve constraints.
Self-serve platform limits — and why they exist
The self-serve platform is engineered for the 90% case — single assets, standard formats, ≤ 30 seconds, English-dominant creative. The constraints aren't arbitrary; they're engineering trade-offs that keep the platform fast, browser-based, and zero-friction. When your work falls outside that 90%, concierge picks up.
| Constraint | Self-serve | Why the limit exists | Concierge |
|---|---|---|---|
| Video duration | First 30 seconds | Whisper transcription model runs in your browser. Longer clips exhaust browser memory on most consumer devices. | Any length — full TVCs, long-form, multi-cut campaigns |
| File size | ≤ 100 MB | Vercel serverless function memory ceiling. Larger files break the upload pipeline mid-transfer. | Raw camera files, ProRes, uncompressed 4K — no ceiling |
| Video formats | MP4, MOV, WebM | FFmpeg.wasm in-browser decoder supports a fixed codec set. Exotic codecs require native FFmpeg. | Any format your editor exports — AVI, MKV, MXF, ProRes, DNxHD |
| First-load time | ~3 min cold fetch | Whisper speech model (~40 MB) downloads to your browser the first time. Cached forever after. | Zero wait — model runs on our infrastructure |
| Batch processing | One asset at a time | Each analysis ties to a single user session. Sequential uploads only. | Parallel sweeps — 50 assets, 200 assets, full campaign portfolios |
| Language model | Auto-detect, defaults to English | Browser-side Whisper-base balances size vs accuracy for typical English creatives. | Language-specific Whisper-large variants tuned for Afrikaans, Zulu, Sotho, Xhosa |
| Audience cohort modelling | 19 verticals, SA-LSM 1-10, urban/rural | Pre-built cohorts cover 90% of self-serve use cases. | Custom cohorts — your CRM segments, panel data, syndicated audience definitions |
| Spatial / immersive analysis | Not in self-serve yet | AR/VR/3D environment neural mapping requires GPU compute and bespoke calibration. | Walk-through retail environment heatmaps, hologram displays, in-store activations |
What concierge unlocks
Capabilities running on our infrastructure that don't fit the self-serve UI — either because they need more compute, longer pipelines, custom calibration, or tooling we're still proving out before shipping to every user.
30-second platform cap doesn't exist on concierge. Send us your 90-second hero spot, your 3-minute brand film, your 6-minute investor video — analysed start to finish, every cut, every beat.
Brief us once, send the full campaign — print, OOH, social cutdowns, TV, radio, in-store. Returned as a unified Campaign Intelligence report with variant rankings, audience-level winners, and the recommended cut for each channel.
Same campaign in Afrikaans, Zulu, Sotho, Xhosa, English? We run each language through a tuned Whisper-large variant and apply the Rueckl et al. (2015) L1-cost differential by province. Discover which language version actually wins in KZN, Limpopo, Western Cape.
Your panel data, CRM segments, or syndicated audience definitions become Buyology-compatible cohorts. We've modelled Massmart shopper segments, Discovery Vitality tiers, MultiChoice viewing cohorts.
AR experience walk-throughs, VR retail prototypes, hologram displays, in-store activation paths. We map gaze, dwell, and approach behaviour onto 3D scenes using methods that don't fit the self-serve UI.
Test creative before it ships. We run pre-flight neural sims against your draft cuts, flag the loss-aversion fails, the cognitive overload moments, the brand-recognition gaps. Iterate three rounds in the time a traditional panel would take to recruit.
How a concierge engagement runs
One paragraph, 5-minute call, or a fully-spec'd RFP — we work to whatever level of structure you bring. Tell us what you're testing, the audience, the channel, and what decision the analysis needs to inform.
Any format, any size, any channel. Raw camera files, finished cuts, work-in-progress edits, competitor pieces for benchmarking. Encrypted upload link provided.
Same neuroscience foundation as the self-serve platform — Whisper transcription, Rueckl L1-cost modelling, NeuMa purchase-intent calibration, 50-principle behavioural audit. Plus the capabilities that don't fit the self-serve UI.
Kantar-grade PDF report with neural scores, heatmaps, cohort comparisons, and prioritised optimisation recommendations. PowerPoint export for stakeholder presentation. Raw data file (JSON/CSV) for your own analytics team. Live debrief call with the analyst who ran the work.
Re-test after edits in days, not weeks. Roll the engagement into a monthly retainer for ongoing campaign sweeps. Build a benchmarking corpus across your entire brand portfolio.
What you get back
Full neural analysis report (PDF)
Kantar-grade structure: executive summary, neural scores per dimension, audience-level breakdowns, heatmaps, cohort comparisons against industry benchmarks, prioritised optimisation recommendations.
PowerPoint export
Stakeholder-ready slide deck with the headline findings, charts, and key visuals. Drop into your board pack or client presentation as-is.
Raw data file (JSON / CSV)
Every neural score, every cohort breakdown, every segment timestamp. Pipe into your own dashboards, BI tools, or analytics stack.
Live analyst debrief
45-minute call with the analyst who ran the work. Walk through findings, stress-test interpretations, agree on the action plan.
Common questions
Traditional studies recruit 30-100 participants, take 6-8 weeks, and cost R4M+ for a single asset. Our concierge runs the same virtual neuromarketing methodology as the self-serve platform — AI models trained on published neuroscience datasets (DEAP, NeuMa, GSR, Visual Saliency) — without the recruitment, lab, or six-week wait. Typical turnaround: 5 business days for a single asset, 2-3 weeks for a full campaign sweep.
The neuroscience foundation is identical. Same NeuMa model (AUC 0.746), same GSR cognitive workload model (AUC 0.70), same 50-principle behavioural audit, same SA-LSM and province-language calibration. Concierge unlocks capabilities that need infrastructure the self-serve UI can't ship — longer videos, batch parallel runs, custom cohorts, spatial analysis — not different science.
You do. The analysis is delivered to you under NDA, and your asset data is not used to train any downstream model without explicit written permission. Anonymised aggregate benchmarks (industry quartiles, vertical baselines) may be derived from the analysis structure, never from your specific creative or audience data.
We run single-asset pilots (often free for the first one, to prove the methodology), campaign sweeps of 5-50 assets, and ongoing monthly retainers for brands testing every flight before market. Brief us on what you're trying to test — we'll scope from there.
Yes. Agencies running concierge under their own brand is a standard arrangement. The deliverables ship with your branding; the methodology cites Buyology Labs' published neuroscience foundation as a third-party validation reference.
Tell us what you're trying to test.
One paragraph or a full RFP — whatever level of detail you have. We'll scope the engagement, agree turnaround, and (for first-time clients) often run the pilot for free to prove the methodology against your existing metrics.