A working data function. Without building one.
Not a tool you have to staff. A working data function — the connections, the history, the metric tree, the reporting people actually open, and the AI layer on top.
Six capabilities. One coherent data function.
Built on a full event history in BigQuery — everything recorded once and never edited afterwards, which makes every later question cheaper to answer, whether that is a report, an attribution model or an AI one.
Integrations & Data Sources
Every system that matters, landed in one place. How we connect is source-dependent — API, export, replica or event stream, whatever yours exposes.
Data Pipelines & Data Layer
BigQuery as the back end and services on Cloud Run, inside your own secure, scalable environment. Where a system only stores its current state, we rebuild the history behind it, so you get a timeline instead of a snapshot.
Business Mapping & Metric Trees
We map how revenue actually moves through your business, then build the metric tree — the agreed map of how every number rolls up — that your operators can argue with, and settle arguments from.
Operational & Risk Intelligence
Dashboards your team actually opens — finance, marketing, ops, risk — wired to the metric tree, with thresholds that raise the work rather than just colouring a cell red.
The Summary Page
One trading-pack-grade page that answers the C-suite question before it gets asked. Built for boardrooms, updated in real time.
Self-Service Analytics with Claude
Claude wired into your data in your own enterprise account. Your exec team asks in plain English and gets answers straight from the metric tree — and starts spotting opportunities nobody had time to look for.
A second brain for your business. And for everyone in it.
The layer does not only report. It remembers — every activity you have recorded, every decision anyone made, and what happened next. That is what turns reporting into recommendations: something that has seen your business work, is still watching, and tells you what to do about what it is seeing right now.
For everyone in it
Ask it what you would ask the colleague who has been here fifteen years. Why do we decline these? What did we do last time this happened? Which of my accounts is about to go wrong? It answers from your own history, at whatever level of detail the person asking wants — and it never gets tired of being asked.
For the business itself
The judgement that currently lives in a handful of heads becomes something the company owns. When someone resigns, the reasoning stays behind. When someone joins, they start with the whole history rather than a handover document and a folder of spreadsheets.
Aligned to your policies and the way you actually decide.
A recommendation engine that does not know your rules is a liability, and one trained on somebody else's business is a guess. Three things keep this one yours.
Your policies, as code
Recommendations are generated inside your policy rather than filtered after the fact. If something cannot be offered to a customer in that position, it is never suggested in the first place — which is also what makes the output safe to put in front of a regulator.
Your decisioning priors, not generic best practice
It learns how your business actually trades off risk, cost and service from the decisions you have already made — including the ones where a person overrode the system and turned out to be right. Your appetite, not the industry average.
The loop closes
Every recommendation is scored against what actually happened next, so the following one is better. That is the difference between a model trained once and a second brain that is still learning in December from what it got wrong in March.
And it runs while nobody is watching.
A monthly pack tells you what you should have done. A second brain is looking at the business continuously, which means the recommendation arrives while the decision is still live — the case still open, the customer still on the phone, the limit not yet breached.
That is also the point at which agents become worth having. Once something can see the whole business, knows your policy and knows what usually works, it can be trusted to do the repetitive part itself and bring you the rest.
The transparency and the speed the Acta Data team work at is refreshing. Being able to query our own data and build our own reports is the biggest step forward we've had in five years — and it has only been three months.
The compliance work nobody is doing. Built in, not bolted on.
Most businesses have a retention policy in a document and no way to show it is being followed. That is rarely negligence. It is that nobody knows which of fourteen systems hold a given record, and deleting it from the CRM does nothing about the copy in the warehouse, the reporting database, the backup, or the spreadsheet on somebody's desktop.
No personal data in the analytical layer
Personal data is obscured at ingest, so it never travels into the modelled layer or into an AI prompt. That is data minimisation built into the architecture rather than configured on top of it, and it is what makes self-service analytics safe to open up: a model cannot leak what was never there. Read access with personal data excluded is all we ever ask for.
Retention and destruction that actually happens
We map where personal data actually lives across every connected system, express your retention rules as code against that map, and run them on a schedule. Then the part that decides whether any of it counts: an audit trail showing what was deleted, when, and under which rule — so a retention policy becomes evidence rather than an intention. Erasure requests get the same treatment, once, across every system that holds the record.
Marketing lists you can stand behind
Every contact traced back to the event that created it — the form submission, with its timestamp and the wording that was on the page — or to nothing at all. Contacts with an evidenced lawful basis get separated from contacts without one, so a list can be suppressed rather than hoped over. The finding is usually the same and usually uncomfortable: a large share of the list has no traceable basis, and nobody knew, because the CRM stores the current state and not how it got there.
We are not lawyers, and none of this is legal advice.
Your DPO, your counsel or your compliance function decides what the policy says and what lawful basis you rely on. We are the engineering underneath it: finding where personal data actually is, doing what the policy says, and producing the evidence that it happened.
Acta Data Ltd is registered with the Information Commissioner's Office under ZB502441, and every environment we build is your own — so the data, and the responsibility for it, never leaves your control.
Compliance stops being an annual project and becomes a by-product of the layer being built properly in the first place.
Choose how you work with us.
Same operators, same stack, same standards. Four ways in — the only question is where you're starting from.
From Scratch
Months, not yearsNothing to build on, or nothing you trust. We start from zero: first real report inside a week, and a working data layer your business runs on within the first few months — regulated reporting included.
BI Takeover
MigrationA tired setup and a team that can't get ahead of it. We take the whole thing over and turn it into something AI can work with, without the lights going out on the way.
Claude Enterprise Onboarding
Safety firstClaude in your own enterprise account, done properly: who can see what, limits on what it can do, testing against real questions, and a data policy your risk and compliance teams will actually sign.
Agent Activation
Operational AIAgents that run real operational work, with one agent coordinating the rest. Your data stops being something you report on and starts being something that does the work.
Acta means acts. We build action data — not archive data.
Bring one decision your data should be helping you make.
30 minutes, no deck. We'll tell you on the call whether we're the right partner, and exactly what we'd ship in the first 30 days.