# Fast Data Science vs Twistag

> Comparing AI Development options: London vs Lisbon. Side by side: published rates, team size, service and industry coverage, and the curator's assessment of both — pairings are never sold.

Canonical page: https://www.aiagencies.eu/compare/fast-data-science-vs-twistag
Generated: 2026-09-17 · Curated by Gabor Kiss · No paid placements — inclusion, score, and rank are never sold.

## Facts
| | Fast Data Science | Twistag |
|---|---|---|
| Editorial score | 68/100 | 73/100 |
| Hourly rate | €175-260/hr (directory-listed) | €45-85/hr (directory-listed) |
| Team size | 2-9 | 10-49 |
| Location | London, United Kingdom | Lisbon, Portugal |
| Founded | 2018 | 2016 |
| Last reviewed | 2026-09-07 | 2026-09-09 |
| Services | AI Development, AI Consulting | AI Agents, AI Development, AI Automation |
| Industries | Healthcare, Public Sector, SaaS & B2B | SaaS & B2B, Manufacturing, FinTech |

## Our take on Fast Data Science
Reviewed 2026-09-07 by Gabor Kiss against published evidence.

The Clinical Trial Risk Tool is the strongest evidence a shop this size can offer: commissioned by the Gates Foundation, built over more than a year as an ensemble of machine-learning and rule-based models that extract sample size, phase, effect size and statistical-analysis-plan presence from protocols running to 200 pages, deployed publicly, open-sourced under MIT, and written up in Gates Open Research with a DOI.

The consulting record is public-sector and pharma—the Information Commissioner's Office email classifier trained inside an isolated environment on data that could not be retained, Boehringer Ingelheim's factory error-report clustering, Skills Development Scotland's causal analysis using DoWhy and instrumental variables, plus the WHO and the Office of Rail and Road.

What the cases do not carry is outcome numbers: the Skills Development Scotland write-up concludes that interventions appeared to be well-targeted, which is a research finding rather than a delivered result.

At €175-260 an hour this is the most expensive listing in London, so the question is what a named principal's time buys that a larger team's does not.

### Key strengths
- The Clinical Trial Risk Tool is peer-reviewed (Gates Open Res 2023, doi:10.12688/gatesopenres.14416.1), publicly deployed and MIT-licensed
- Sensitive-data constraints are handled explicitly—the ICO model was trained inside an isolated environment with no training data retained
- The client list is institutional and verifiable: Gates Foundation, WHO, NHS, ICO, Boehringer Ingelheim, Wellcome Trust

### Good to know
- No case study publishes an outcome number—ask what changed for the client after delivery
- At €175-260/hr against a 2-9 headcount, ask who does the work and what happens when they are unavailable
- The published strength is NLP and research-grade analysis—ask for evidence before commissioning a production platform

## Our take on Twistag
Reviewed 2026-09-09 by Gabor Kiss against published evidence.

The best-documented AI portfolio on this page, and the detail is where the value is. For the manufacturer Aralab, an accounts-payable agent reads any supplier PDF with Claude while deterministic code independently recalculates every line total, unit price and tax figure against the matched purchase order: the model proposes, the validation engine verifies. It shipped in six weeks with one engineer, handles more than 2,000 invoices a month, and moved three finance people off transcription.

The operational choices are published too. LangFuse traces every model call with prompt versions and cost per invoice visible in real time, and the review dashboard sorts each document into automated, partially matched or flagged so attention goes only where the model was uncertain.

Two more agent systems run in production: a compliance platform for a European regulatory-technology startup, cutting time per customer inquiry 75 percent and live at three of the top ten European cosmetic brands, and four chained agents for PepTalk over 3,000 speaker profiles indexed as vectors. The PepTalk write-up states plainly that no measured figure for volume, turnaround or conversion has been published.

Best for teams buying an agent system that has to be auditable and cheap to run, from engineers who work inside your stack and hand it back.

### Key strengths
- The hybrid pattern is published, not implied: the language model interprets and deterministic code independently recalculates every figure against the purchase order
- Observability ships with the system—LangFuse traces every call with prompt versions and per-invoice cost visible in real time, so the workflow cannot get quietly expensive
- Models sit behind a provider abstraction and were swapped twice on one engagement without the product changing shape

### Good to know
- The largest engagement, the regulatory compliance platform, is anonymized—ask for a speakable reference at one of the brands running it
- The PepTalk system publishes no measured outcome by the agency's own admission; ask what the volume and turnaround figures are now
- No headcount is published against a decade of trading—ask how many engineers would be forward-deployed on your build and for how long

## Which to choose
Choose Fast Data Science for:
- closer collaboration at a smaller team scale
- disclosed specialization in Healthcare / Public Sector
- broader service offering — also covers AI Consulting

Choose Twistag for:
- a more cost-effective engagement (~235% lower rate)
- larger team capacity for multi-stream or enterprise-scale programs
- disclosed specialization in Manufacturing / FinTech
- broader service offering — also covers AI Agents and AI Automation

## Questions buyers ask

### Which is cheaper, Fast Data Science or Twistag?
Twistag is cheaper by roughly 235% — €65/hr median vs €218/hr at Fast Data Science. Published rates cover different things at different agencies — seniority mix, discovery and included revisions are the usual variables — so treat the gap as a question to ask, not a verdict.

### Which has more SaaS & B2B experience, Fast Data Science or Twistag?
Both agencies show documented SaaS & B2B work and have similar industry breadth. Compare directly on the agency profiles: Fast Data Science and Twistag.

### Which scores higher overall, Fast Data Science or Twistag?
Twistag scores 73/100, 5 points higher than Fast Data Science at 68/100. See our methodology for how scores are calculated.

### When should I consider both Fast Data Science and Twistag?
Consider running parallel discovery briefs with Fast Data Science and Twistag if your project spans multiple workstreams, you want competitive proposals to compare scope and approach, or you're undecided between the specialization angles each brings (see the Best for cards above). Most engagements ultimately go with one — but the parallel-brief phase is a low-cost way to validate fit.

---
Source: AIAgencies.eu — the curated register of European AI agencies. Rate provenance: "agency-confirmed" = disclosed via submission or claimed profile; "directory-listed" = Clutch or two agreeing B2B directories (lone unverified sources are never written). Full method: https://www.aiagencies.eu/methodology · Rate index: https://www.aiagencies.eu/rates