Pangolin
Login
Welcome back! Please enter your details.
We built Pangolin because DTC brands were flying blind on their own budgets
Most marketing measurement was built for enterprise brands with unlimited data science budgets, or it's built on attribution models that break the moment a browser blocks a cookie. DTC brands spending up to £10M on paid media get stuck in between: too much complexity to run on gut feel, not enough resource to run a data science team. Pangolin closes that gap.
How it works
Bayesian MMM
Bayesian MMM builds a probabilistic model of your media response, incorporating what you already know about your channels (your priors) and updating that view as new spend and revenue data comes in. The output isn't a single number pretending to be certain. It's a range of plausible outcomes, with the uncertainty shown, not hidden.
Budget recommendations
Pangolin generates budget recommendations built around contribution margin, not ROAS. Revenue tells you what came in. Contribution margin tells you what you kept. That's the number that should be deciding your next £100k in spend, and it's the one most platforms don't optimise for.
Agentic execution
Where it makes sense, Pangolin's agentic layer can execute the reallocation directly, once you've approved it. Human judgement stays in the loop. The model doesn't get to spend your budget unsupervised - it gets to make the case for where it should go.
Why it matters
That's not from spending more. It's from spending what you already have in the right places.
The team
Pangolin was founded by people who've sat on both sides of the media spend problem: the marketer trying to prove what's working, and the data scientist trying to model it properly.
14+ years in performance marketing across agency and client-side roles, including time at Publicis on the agency's AI usage committee. Has run budgets across TikTok, YouTube, Pinterest, programmatic, Snapchat, CTV and digital audio.
Engineering experience from ASOS, Trainline, Tesco Bank, WPP Open and Publicis Marcel. Builds the infrastructure that turns a statistical model into a product marketers can actually use.
Lead researcher and contributor to PyMC Labs. Leads the Bayesian methodology at the core of Pangolin - the part of the product that has to be right before anything else matters.
Who we're built for
Pangolin is built for CMOs, Heads of Performance Marketing, Growth Directors and founders at DTC ecommerce brands who are done guessing. If you've got media spend history and a real question about where the next pound should go, that's the conversation we want to have.
"We're not going to tell you Bayesian is the only way to do MMM properly, or that our model replaces your judgement. We'll show you the model, the uncertainty in it, and let you decide what to trust."
Frequently Asked Questions
Do we need an in-house data science team?
No. Pangolin is a complete software-as-a-service solution. Our pipeline automatically cleans your data, fits the Bayesian algorithms, and presents the output in an intuitive interface. We handle the hard mathematics so you can focus on allocation decisions.
How often do Pangolin's models update?
Models update automatically every single week. We ingest daily transaction data, process baseline adjustments over the weekend, and deliver the final, validated attribution outputs and recommendations on Monday morning.
What data history is required to start modeling?
We recommend at least 12 months (ideally 24 months) of historical daily sales and advertising spend data. This history is crucial to train the model to understand seasonality and baseline organic performance levels.
Does Pangolin replace our existing analytics stack, or sit alongside it?
Pangolin sits alongside your existing tools. We don't ask you to rip out GA4, your ad platform dashboards, or your CRM. We ingest data from them and turn it into a single incremental-revenue view your team can act on.
How long until we see our first output?
Once your data sources are connected, initial model outputs are typically available within days, not the 3 to 6 months a traditional MMM consultancy takes. Full confidence intervals stabilise over the following few weekly refreshes as the model sees more data.
Revenue is vanity. Contribution margin is the number that matters.
Stop flying blind with platform ROAS. See what a profit-optimised budget would look like for your brand.
Book a demo