Now accepting design pilot partners Kuwalyst

Make Your Marketing Spend Work Harder
With Science, Not Guesswork

The agentic marketing platform that diagnoses, experiments, and optimises, autonomously. EU-sovereign. No data leaves your infrastructure.

Learn about our pilot program Book a Deep-Dive

You're spending millions. But you're flying blind.

If even 20% of your marketing budget is misallocated, and without causal measurement, you can't know it isn't, that's spend working against you every month. Most companies leave this on the table, not because they lack data, but because they lack the methodology.

  • Attribution Is Broken: Cookies are dying. Offline channels go unmeasured. Platform reporting is self-serving. Most companies can't prove which part of their spend actually works.
  • Budget Under Pressure: Consultancies charge €100k+ for one-off audits that gather dust. There's no ongoing steering, no experimentation, no closed loop between insight and action.
  • No Closed Loop: The gap between "what the data says" and "what to do next" is where money gets wasted. Most teams report on the past but never close the loop with experiments.

From Data to Decisions. Continuously.

A closed-loop system: most tools stop at reporting. Kuwalyst acts and learns. Four steps, running continuously, turning your marketing data into compounding advantage.

1. Connect

Plug into your sales, campaign, and customer data — online and offline. EU-sovereign, privately hosted. No data leaves your infrastructure.

2. Diagnose

Get an instant marketing health check. Know exactly where you stand in weeks, not months.

3. Steer

Weekly and monthly tactical reports that tell you what to do, not just what happened. Traffic-light summaries, anomaly flags, and prioritised actions.

4. Evolve

Kuwalyst works like a marketing scientist on your team, constantly forming hypotheses, running models in the background, and proposing experiments to improve performance.

See it in action

Watch how Kuwalyst can simplify the creation of marketing mix model

Mixed Marketing Model benchmark

We benchmarked the creation of a Marketing Mix Model with synthetic data reflective of a mid-sized European retail brand advertising across 7 countries.

Using Claude Opus 4.6 extended thinking mode, ROAS estimations were off by a magnitude of 10x

Claude MMM simulation ROAS

With the same data, Kuwalyst was able to get very close to ground truth ROAS, except for TikTok channel where it was off by a factor of 2x

kuwalyst MMM simulation ROAS

This was made possible thanks to Kuwalyst proprietary LLMs (fine tuned versions of open weights LLMs), trained to iterate over models and improving them incrementally

Kuwalyst MMM model iterations

FAQ

What data do I need to get started

The more data Kuwayst gets, the better it can hekp you optimise your marketing strategy. Your data does not leave your company and is not transferred to any third-party service provider.

  • Sales: the more granular the better. Ideally should include the "who" (customer), "what" (which products were bought), "where" (which store if your business also sells in brick and mortar stores). Kuwalyst can work with aggregated data though.
  • Marketing channel spend: Kuwalyst shines in formulating hypothesis in how to best spend your marketing budget. In order to map causal effects between marketing spend and slaes, it needs to leverage both online marketing spend (search, social, display) and offline marketing spend (TV, CTV, Radio, OOH)
  • Campaign-level data (optional): to get tactical reports about how your campaign are behaving and to suggest short term optimisations, Kuwalyst can leverage daily campaign performance data. This is can be done automatically by setting service accounts on your advertising platforms so Kuwalyst can automatically pull the data
  • Unstructured Marketing data (optional): in order to get to know your business and your marketing strategy, you can let Kuwalyst crawl your marketing campaign briefs as well as your strategy documents. Kuwalyst will use this to pinpoint possible discrepancies between your goals and your marketing implementation. It will also allow Kuwlayst to propose experiments to gain insights toward achieveing your marketing goals.
  • Customer data (optional): if you plan to do customer level analysis (CLV, segmentation), Kuwalyst can leverage customer level data such as demographics and buying history
How is Kuwalyst "EU-sovereign"?

Kuwalyst does not rely on closed models only accessible through APIs (e.g. Claude, chatGPT), Kuwalyst models are based on open weight models.

These models are hosted in France thanks to a partnership with Scaleway.

Scaleway-Logo-White

For added security, both Kuwalyst application can be hosted on a VPC or directly on your infrastructure.

How long until I get to see results?

As soon as you connect Kuwalyst to your data sources, Kuwalyst will be able to give you insights about marketing performance, what works well, and what works less well. For more complex projects requiring long running incrementality/lift tests, you can expect better performance in a matter of months.

The Founders.

Headquartered in Paris, France

Julien Bourdon-Miyamoto

Julien Bourdon-Miyamoto

CEO & Co-founder

20+ years in AI and ML engineering. 10+ years managing advertising ML products at global media companies. Ex-Meta Staff Ads ML Engineer.

Profile
Hajime Takeda

Hajime Takeda

Co-founder & Marketing Scientist

10 years as data scientist and marketing analyst in retail and fashion across the US and Japan. Recognised speaker at major data science and marketing science conferences.

Profile

Join the Pilot Programme.

We're selecting 3 design partners to shape the product alongside us. 4-month engagement, less than €10k, with a discounted annual licence afterward. Book a 45-minute deep-dive to explore your marketing data landscape. More details about the pilot program

What's your biggest marketing measurement challenge? (Multiple answers possible) *
What's your approximate annual marketing spend?
Do you have a data science or marketing analytics team? *
Would you be interested in a pilot programme? *