About Kleyr

Marketing measurement without the notebook sprawl.

Kleyr is a UK growth-sciences company. We build the platform marketing-science teams used to assemble from Python notebooks, Excel templates, and R scripts. Same methods. Same rigour. One workflow.

Why we exist

The methodology gap is real.

Sophisticated MMM exists at one end of the market — Recast, Mutinex, Liftlab, Keen — at enterprise pricing. Lightweight modelled attribution exists at the other — Triple Whale, Polar Analytics, Lifetimely — built around Shopify, methodologically unrigorous.

In the middle: a Shopify-native DTC mid-market that spends £2–20m / year on marketing, doesn't have an in-house marketing scientist, and is making meaningful budget decisions on data they can't fully trust.

Kleyr fills that gap. Recast-grade Bayesian MMM, geo-experiment calibration, and proper customer analytics — productised, integrated, priced for the segment that actually needs it.

How we work

Quietly technical.

Voice. We don't say "AI-powered", "revolutionary", or "supercharge". We state the number, show the credible interval, and recommend the experiment that would change the answer.

Methodology. Every method we ship traces back to a paper. PyMC-Marketing, Google Meridian, GeoLift, EconML, PyMC. All open-source, all referenceable. The Kleyr layer — data ingestion, run management, the agent skill surface, the app — is what we build.

Defaults. Diagnostics shown by default. Calibration plots in product. Reproducibility (lockfile + seed + content-addressed dataset) built in from day one. EU data residency by default.

What we believe

Six things we'll defend.

Bayesian by default
Posterior intervals are honest about uncertainty in a way frequentist confidence intervals are not — especially for non-statistician readers.
Experiments calibrate models
Geo lift estimates as Bayesian priors on MMM channel coefficients. The loop is what makes the answers credible.
Show the diagnostics
Reliability diagrams, AVM charts, residual plots — by default. The diagnostic suite is the trust differentiator.
Calibration > AUC
A churn model with 0.85 AUC and bad calibration tells you nothing useful. Show the reliability diagram, every time.
Honest agents
The AI agent says "I don't know" when the data can't support an answer — and recommends the experiment that would. This is the differentiator.
Customers own the artifacts
Every run produces an immutable, exportable bundle: dataset, model, posterior samples, lockfile, seed. Reproducible to the bit.
Who we are

Founded out of the work.

Kleyr was founded in 2026 in the UK after years of running marketing-science engagements in agency and in-house environments. The platform productises an analyst workflow that already worked — same R / PyMC methodology, same diagnostic discipline, same output schema — built for a wider market than enterprise pricing reaches.

We're bootstrapped on services revenue while we ship, with a default plan to raise on productised wedge once early customers prove pull. Pricing, packaging, and product decisions are made commercially — not by what's technically interesting, but by what changes a customer's next budget meeting.

Talk to the team →
✦ Get started

Bring your spend, leave with a model you can defend.

Book a 30-minute working session with the Kleyr team. We'll model your last 12 weeks of marketing spend on the call — no slides, no abstract demo.

Book a demoTry the benchmark simulator →