# Acta Data — full reference > Acta Data Ltd builds the operational data layer that AI needs, then the AI on top of it. Every system a business owns is landed in its own Google BigQuery environment as immutable events, with modelled metrics, board and regulatory reporting, a real-time balanced-scorecard Summary Page, safe PII-restricted self-service analytics through Claude, and AI agents on the repetitive operational work. Live in weeks, handed over Terraformed inside twelve months. Acta Data Ltd is registered in England & Wales, company number 14182372, ICO registration ZB502441. Registered office: Chester House, Lloyd Drive, Cheshire Oaks Business Park, Ellesmere Port, Cheshire CH65 9HQ. Contact: info@actadata.co.uk, +44 7749 407562. LinkedIn: https://www.linkedin.com/company/acta-data/. Serves the United Kingdom. This is the full-text reference. The shorter index is at https://www.actadata.co.uk/llms.txt. ## The model - **The stack, in plain terms**: Google BigQuery is where the data lives, shadcn is what you look at, Claude is who you ask. All Google underneath, inside the client's own secure VPC. - **Recorded once**: every activity is recorded once and never rewritten, carrying four atomic units — cost, revenue, conversion and time — plus its business context. That is what makes value-stream leaks visible and what makes the data usable by AI. - **Who it is for**: mid-market and SME operators with enough operational complexity to need a data function and not enough scale to justify five hires for one. - **The second brain**: the layer keeps every activity, every decision and what happened next, so it can be asked in real time. Recommendations are generated inside the client's own encoded policy, the decisioning priors are learned from decisions that business has actually made, and each recommendation is scored against the outcome so the next is better. - **The team**: data and AI people who have held C-suite positions across marketing, operations, technology and product. ## How it works — How Acta Data builds your operational data layer Clean, model, alert, act: how Acta Data takes every system you own into one place in Google BigQuery, then puts reporting, exceptions and agents on top of it. 1. **Clean.** Every system in, deduplicated and reconciled, with personal data obscured at ingest so none of it travels into the layer. 2. **Model.** Recorded once in BigQuery and never rewritten — every activity carrying its cost, revenue, conversion and time, with definitions written down as code. 3. **Alert.** A real-time view of the business focused on the next best action — the Summary Page scorecard and the one thing that needs attention now. 4. **Act.** Agents work the exception off the back of it — chasing, flagging and routing — rather than just noticing it, keeping your people on the decisions. Delivery runs over roughly twelve months: a real number in week one, the layer taking shape in the first month, the Summary Page and reporting in months two and three, self-service and then agents from months four to twelve, and a structured Terraformed handover after twelve months. Optional maintenance follows at £2k per month on a rolling contract. ## Pricing Priced by how much of the business is in scope, not by feature tier. For scale, one area a year is less than half a fully-loaded senior data hire at around £143k. - **Discovery — £15k one-off.** Value stream mapping and an AI readiness review, delivered as a prioritised build plan the client owns outright and can execute themselves or hand to another supplier. Credited in full against a subsequent build. - **One area — £60k per year (£5k per month).** A single problem solved end to end, usually operations. - **Whole business — £120k per year (£10k per month).** Every value stream on one layer with one set of definitions, the full reporting suite including regulated reporting, plus safe PII-restricted self-service analytics through Claude. - **Enterprise — priced against the outcome delivered.** Adds autonomous agents running operational workflows. - **Maintenance — £2k per month**, optional, rolling, no notice period. ## Sectors — full detail ### Consumer Credit — https://www.actadata.co.uk/sectors/consumer-credit Consumer Duty evidence, vintage performance and one agreed contribution number. Lending businesses usually have the data and no way to stand behind it. Risk, finance and marketing each hold a version of the truth, the bureau data sits in exports, and Consumer Duty evidence gets assembled by hand as the deadline approaches. We give you one timeline per customer and the history to trend it. Metrics that matter here: Approval rate, Funded volume, 30+ DPD, Roll rates, CAC payback, Vintage loss. Where it hurts: - Risk, finance and marketing each have their own version of the truth - Consumer Duty evidence assembled by hand, close to the deadline - Systems hold today's position and no history, so there is no trend to report - Decisioning data locked inside bureau exports What Acta builds: - Application-to-funded-to-collections single timeline per customer - Consumer Duty outcome reporting, with a full audit trail behind every number - History rebuilt so vintage cohorts and roll rates exist at all - Affordability and arrears views the credit committee can act on What you get: One number for net new contribution — agreed by risk, finance and growth. Common questions: - **How do you produce Consumer Duty evidence from our data?** We land every event — application, decision, funding, contact, payment — in your own BigQuery as an immutable history, then build outcome reporting on top with a full audit trail behind each number. The evidence regenerates itself rather than being assembled by hand as the deadline approaches. - **Can you rebuild vintage cohorts and roll rates if our system only holds today's position?** Yes. Most lending systems store the current balance and no history, so cohorts and roll rates do not exist to report. We reconstruct the position over time from the events, so vintage loss, roll rates and arrears trends become reportable — usually further back than you expect. - **How much does it cost for a consumer credit lender, and how fast?** Solving one area — usually risk or collections reporting — is £60k a year, billed at £5k a month, in your own Google environment. You see a real number within the first week, and the reporting behind it lands over the following two to three months. ### Debt Management — https://www.actadata.co.uk/sectors/debt-management Collections performance, treatment outcomes and fair-value evidence in one place. Collections runs on operational data that rarely reaches the board intact. Contact outcomes live in the dialler, arrangements in the servicing system, and customer outcomes nowhere in particular. We join them so treatment effectiveness and customer harm are both measurable. Metrics that matter here: Cure rate, Breakage, Contact-to-arrangement, Cost to collect, Promise kept %. Where it hurts: - Contact, arrangement and payment data spread across three systems - Treatment effectiveness argued from agent anecdote - No evidence trail for fair value or vulnerable-customer outcomes - Cure and breakage rates recalculated by hand every month What Acta builds: - One customer timeline across contact, arrangement, payment and outcome - Treatment pathway performance, cure and breakage by cohort - Vulnerability and forbearance flags carried through to reporting - Regulatory outcome reporting with the working shown What you get: Evidence that a treatment strategy works, and where it quietly does not. Common questions: - **Can you measure whether a treatment strategy actually works?** Yes — that is usually the first thing we build. We join contact outcomes from the dialler, arrangements from the servicing system and payments into one customer timeline, then measure cure and breakage by cohort and pathway. You see which treatments work and where one quietly does not. - **How do you evidence fair value and vulnerable-customer outcomes?** We carry vulnerability and forbearance flags through from the operational systems into the reporting layer, so outcome reporting shows the working rather than a headline number. Every figure has an auditable trail back to the events behind it, which is what a fair-value review needs. - **Our contact, arrangement and payment data live in three systems — is that a problem?** No, that is the normal starting point. We connect each system and land its events in your own BigQuery, then join them into one timeline per customer. Nothing has to be migrated or replaced — we work in your environment and read from the systems you already run. ### Credit Unions — https://www.actadata.co.uk/sectors/credit-unions Member value, loan book health and social impact you can actually show. Credit unions carry the reporting burden of a bank on a fraction of the resource. The ledger holds the truth but not the history, member value is asserted rather than measured, and the board pack is somebody's weekend. We make the book and the membership legible without adding headcount. Metrics that matter here: Loan book value, Arrears 30+, Member growth, Savings per member, Cost-income ratio. Where it hurts: - Core banking system holds balances, not the history behind them - Board pack assembled manually from several exports - Member value and social impact claimed but not evidenced - Arrears trend impossible to see because only today's position is stored What Acta builds: - Loan-book position rebuilt over time, so arrears and provisioning trend properly - Member lifecycle view: joining, saving, borrowing, staying - Social impact reporting built from the ledger rather than estimated - A board pack that regenerates itself every month What you get: A board pack that stands up to a regulator and a member AGM alike. Common questions: - **Can a credit union afford this without adding headcount?** Yes — the point is to get bank-grade reporting without bank-grade resource. Solving one area is £60k a year at £5k a month, in your own Google environment, with no new hire to make or manage. The board pack that used to be someone's weekend regenerates itself. - **Our core banking system holds balances but not history — can you still trend arrears?** Yes. We rebuild the loan-book position over time from the ledger events, so arrears and provisioning trend properly instead of showing only today's figure. That history is what lets you see a problem developing rather than reporting it after it has arrived. - **Can you evidence member value and social impact for the AGM?** We build social impact and member-value reporting from the ledger itself rather than from estimates, alongside a member lifecycle view of joining, saving, borrowing and staying. It stands up to a regulator and to a member AGM alike because every number traces back to the data. ### B2B Services — https://www.actadata.co.uk/sectors/b2b-services Pipeline, delivery margin and account health from one definition. In B2B the deal and the delivery are measured by different teams, in different systems, on different definitions. Sales forecasts optimistically, delivery reports late, and nobody can say which accounts actually make money. We join the commercial and delivery sides so the answer is one number. Metrics that matter here: Pipeline coverage, Win rate, Delivery margin, Net revenue retention, Utilisation. Where it hurts: - CRM pipeline and delivered revenue never reconcile - Delivery margin per account unknown until the quarter closes - Renewal risk spotted after the renewal conversation, not before - Every board meeting starts with rebuilding the same numbers What Acta builds: - Opportunity-to-cash timeline joining CRM, delivery and finance - Account-level margin including delivery cost and time - Leading indicators for renewal and expansion risk - A commercial pack the sales and delivery leads both sign off What you get: One view of which accounts are worth more effort, and which are quietly costing you. Common questions: - **Why do our CRM pipeline and delivered revenue never reconcile?** Because sales and delivery measure different things, in different systems, on different definitions. We build one opportunity-to-cash timeline joining CRM, delivery and finance, so pipeline, delivered revenue and margin all resolve from the same events instead of three teams' spreadsheets. - **Can you show delivery margin per account before the quarter closes?** Yes. We build account-level margin that includes delivery cost and time, updated as the work happens rather than reconciled at quarter-end. You can see which accounts are worth more effort and which are quietly costing you, in time to do something about it. - **How much does it cost and how quickly do we see something?** Solving one area — usually the commercial-plus-delivery view — is £60k a year at £5k a month, built in your own Google environment. A real number is in front of you within a week, with the full commercial pack live over the following months. ### Wholesale & Distribution — https://www.actadata.co.uk/sectors/wholesale Customer profitability, stock turn and true landed margin. Wholesale margin hides in the detail — rebates, drop sizes, carriage, returns and payment terms all move it, and none of them sit in the same report. We rebuild margin at the line level so pricing and range decisions stop being guesses. Metrics that matter here: Gross margin, Contribution per customer, Stock turn, Dead stock, Fill rate. Where it hurts: - True margin per customer buried under rebates, carriage and terms - Stock turn and dead stock reported too late to act - Trade pricing decisions made without knowing the current margin - Rep performance measured on revenue, not contribution What Acta builds: - Line-level margin with rebates, carriage and returns attributed properly - Customer and product contribution ranking that updates itself - Stock turn, ageing and availability in one operational view - Rep and depot performance on contribution rather than revenue What you get: Pricing and range decisions made on real margin, the same week. Common questions: - **Our true margin is buried under rebates, carriage and terms — can you untangle it?** Yes. We rebuild margin at the line level with rebates, carriage, returns and payment terms all attributed properly, so margin per customer and per product stops being a guess. Pricing and range decisions then run on real contribution rather than headline revenue. - **Can you show stock turn and dead stock early enough to act?** Yes — stock turn, ageing and availability land in one operational view that updates itself, rather than a report that arrives too late to do anything about. Dead stock becomes visible while there is still a decision to make on it. - **We're a distributor, not a tech company — is this really for us?** It is built for exactly this. You do not need a data team or a platform project: we work in your own Google environment, read from the systems you already run, and put real margin in front of you within a week. Solving one area is £60k a year at £5k a month. ### Manufacturing — https://www.actadata.co.uk/sectors/manufacturing Cost to make, yield and OEE joined to what you actually sold. Production data and commercial data rarely meet, so nobody can say which lines make money. Shop-floor systems know output, finance knows revenue, and the cost of a unit is an estimate everybody privately doubts. We connect them. Metrics that matter here: OEE, Yield, Scrap rate, Cost per unit, On-time in full. Where it hurts: - Standard costs stale, so product profitability is a guess - Yield and scrap reported weekly, decided daily - Downtime causes recorded but never analysed - Production and sales data never reconciled What Acta builds: - Actual cost to make per unit, updated as inputs move - Yield, scrap and downtime joined to shift, line and product - Order-to-despatch timeline across production and commercial - Product-level profitability the commercial team trusts What you get: A clear answer on which products and lines are worth running. Common questions: - **Can you tell us the real cost to make a unit?** Yes. We build actual cost to make per unit, updated as inputs move, rather than the stale standard cost most product profitability rests on. Joined to what you sold, it gives the commercial team a product-level profitability number they can actually trust. - **Our shop-floor and finance systems don't talk — can you connect them?** That join is the core of the work. Shop-floor systems know output, finance knows revenue, and the two rarely meet. We land both in your own BigQuery and build an order-to-despatch timeline across production and commercial, so which lines make money stops being an argument. - **How do you handle yield, scrap and downtime?** We join yield, scrap and downtime to shift, line and product, so causes recorded but never analysed finally get analysed. Reported weekly but decided daily becomes a live operational view, which is where OEE improvement actually comes from. ### Omni-channel Retail — https://www.actadata.co.uk/sectors/omni-channel-retail True margin by SKU, channel and customer, without the spreadsheet. Retail data is scattered by design — webstore, marketplaces, retailer EDI, ad platforms — and margin truth is buried under returns, promotions and shipping. We land it all and rebuild margin so the trading meeting runs on numbers rather than exports. Metrics that matter here: Net revenue, Gross margin, Return rate, AOV, Repeat rate, Contribution per SKU. Where it hurts: - Channel data scattered across webstore, marketplaces, retailer EDI and ad platforms - Margin truth buried under returns, promotions and shipping costs - Buying meetings run on stale exports - Customer value measured per channel, never end to end What Acta builds: - Unified order-line history with cleaned promo, returns and COGS attribution - True margin metric tree by SKU, channel and customer segment - Live Summary Page tuned for the trading meeting - Cohort and repeat-purchase view across every channel What you get: A live trading pack the buying team trusts more than the spreadsheet. Common questions: - **Can you unify webstore, marketplaces, retailer EDI and ad platforms?** Yes. We land every channel — webstore, marketplaces, retailer EDI and ad platforms — into one unified order-line history in your own BigQuery, with promotions, returns and COGS attributed cleanly. Margin truth stops being buried under the things that distort it. - **Will the trading meeting finally run on live numbers instead of exports?** That is the point of it. We build a live Summary Page tuned for the trading meeting, plus a true-margin metric tree by SKU, channel and customer segment. The buying team ends up trusting it more than the spreadsheet, because it is current every time they open it. - **Can you measure customer value across every channel, not per channel?** Yes. We build a cohort and repeat-purchase view that spans every channel end to end, rather than measuring a customer separately in each one. That is what makes AOV, repeat rate and contribution per SKU comparable across the whole business. ### Legal Services — https://www.actadata.co.uk/sectors/legal-services Case economics, WIP and cost per acquired case, settled. Case management, marketing and finance systems do not talk, so cost per acquired case is a guess and WIP value drifts between fee-earner and finance views. We build one case lifecycle from first touch to settlement. Metrics that matter here: Cases opened, Cost per acquired case, WIP value, Settlement value, Time to settle. Where it hurts: - Case management, marketing and finance systems don't talk - Cost per acquired case is a guess - WIP value drifts between fee-earner views and finance views - Panel and source performance unknown until a case closes What Acta builds: - Case lifecycle history from first touch to settlement - Funnel and lifetime value by panel, source and claim type - Shared WIP and pipeline view for partners and finance - Settlement and duration benchmarks by case type What you get: A partner pack that is current every time they open it, ending the 'whose number is right?' debate. Common questions: - **Can you tell us our true cost per acquired case?** Yes. We join marketing, case management and finance into one case lifecycle from first touch to settlement, so cost per acquired case stops being a guess. You can then see funnel and lifetime value by panel, source and claim type, and put spend where it actually pays. - **WIP value differs between fee-earners and finance — can you reconcile it?** Yes. We build a shared WIP and pipeline view that partners and finance read from the same definition, so the "whose number is right?" debate ends. The partner pack is current every time it is opened rather than rebuilt for each meeting. - **How much does it cost for a law firm and how quickly does it land?** Solving one area is £60k a year at £5k a month, in your own Google environment. You get a real number within the first week, with settlement and duration benchmarks by case type building out over the following months. ### Customer Service — https://www.actadata.co.uk/sectors/customer-service Cost to serve, repeat contacts and the root causes behind both. Service operations generate enormous amounts of data and almost no insight. Contacts sit in the telephony platform, cases in the CRM, and the reason people got in touch nowhere at all. We join the contact to the customer and the order behind it, so cost to serve and repeat contact become measurable — and fixable. Metrics that matter here: Cost per contact, First-contact resolution, Repeat contact rate, Average handling time, CSAT, Failure demand. Where it hurts: - Contact data split across phone, email, chat and social, with no single view - Cost to serve unknown, so nobody can price or staff it properly - Repeat contacts counted as new ones, hiding the real failure demand - Agent performance argued from call listening rather than outcomes What Acta builds: - One contact timeline per customer across every channel, joined to the order or account - Cost to serve per contact, per customer and per product - Repeat-contact and root-cause analysis, so failure demand is visible - Deflection and self-serve impact measured against real volume What you get: The reasons people contact you, ranked by what fixing them is worth. Common questions: - **Can you tell us our real cost to serve?** Yes. We build cost to serve per contact, per customer and per product by joining the contact to the customer and the order behind it. Once it is measurable you can price and staff service properly instead of guessing at it. - **How do you separate repeat contacts from genuinely new ones?** We join every contact into one timeline per customer across phone, email, chat and social, so a repeat contact is recognised as one rather than counted as new. That reveals the failure demand hiding in your volume — and its root causes, ranked by what fixing them is worth. - **Can you measure whether self-serve and deflection actually work?** Yes. We measure deflection and self-serve impact against real volume rather than against a vendor's claim, so you can see what genuinely reduces contacts. Agent performance moves onto outcomes instead of call-listening anecdote at the same time. ### Recruitment & Training — https://www.actadata.co.uk/sectors/recruitment-training Desk margin, time-to-fill and cohort outcomes on one definition. Placement, pipeline and margin data sit across ATS, CRM and payroll, so consultant productivity gets argued from memory. We join them, then wire course and cohort outcomes back to revenue. Metrics that matter here: Time to fill, Desk margin, Fall-through, Placements per consultant, Cohort completion. Where it hurts: - Placement, pipeline and margin data spread across ATS, CRM and payroll - Consultant productivity argued from memory - Course completion and outcome data disconnected from revenue - Fall-through absorbed quietly rather than measured What Acta builds: - Candidate and placement timeline from first contact to invoice - Desk-level margin, time-to-fill and fall-through rates - Cohort completion and outcome tracking wired to billing - Consultant performance on contribution, not activity What you get: Every desk and every cohort measured the same way, without a spreadsheet. Common questions: - **Can you measure consultant and desk productivity properly?** Yes. We build a candidate and placement timeline from first contact to invoice across ATS, CRM and payroll, then measure desk margin, time-to-fill and placements per consultant on one definition. Productivity stops being argued from memory and gets measured on contribution. - **Is fall-through actually measured, or just absorbed?** Most agencies absorb it quietly — we make it a number. Fall-through is tracked by desk and cohort so its real cost is visible, which is usually the fastest margin improvement available once you can finally see it. - **Can you connect training course outcomes to revenue?** Yes. We wire cohort completion and outcome tracking back to billing, so course and cohort performance is measured against revenue rather than sitting disconnected from it. Every desk and every cohort ends up measured the same way, without a spreadsheet. ### SaaS & Startups — https://www.actadata.co.uk/sectors/saas-startups Activation, retention and CAC payback that survive diligence. Product events, billing and CRM tell three different growth stories, and the board pack gets rebuilt by hand the week before each meeting. We give you one event history and one definition of every growth metric. Metrics that matter here: MRR, Net revenue retention, Activation rate, CAC payback, Logo churn. Where it hurts: - Product events, billing and CRM tell three different growth stories - Board metrics rebuilt by hand the week before each meeting - No activation or retention signal early enough to act on - Diligence questions that take a week to answer What Acta builds: - Every product event recorded once, joined to billing and CRM - Activation, expansion, churn and CAC payback from one definition - An investor-ready pack that regenerates itself - Cohort retention curves from the first cohort onwards What you get: The board pack builds itself, and the growth numbers survive diligence. Common questions: - **Our product events, billing and CRM tell three different growth stories — can you fix that?** Yes. We record every product event once and join it to billing and CRM in your own BigQuery, so activation, expansion, churn and CAC payback all come from one definition. The three conflicting stories collapse into one set of numbers everyone works from. - **Will the metrics survive investor diligence?** That is the bar we build to. Because every metric derives from an immutable event history rather than a hand-built spreadsheet, the numbers are consistent and auditable, and diligence questions that used to take a week become a query. Cohort retention runs from your first cohort onwards. - **We're early — is it worth doing this now?** Doing it early is the advantage: the event history you capture now is what later cohort and retention analysis depends on, and it cannot be recovered retrospectively. Solving one area is £60k a year at £5k a month, in your own Google environment, and the board pack then builds itself. ## Questions and answers ### What does Acta Data actually do? We build the data layer a business needs to run on, and then the AI on top of it. Every system you own gets connected and landed in your own Google BigQuery environment as events that are recorded once and never rewritten. On top of that we build the modelled metrics, the board and regulatory reporting, and a Summary Page that shows the whole business against target. Then we wire in Claude so your team can ask their own questions, and put agents onto the repetitive operational work. ### How much does it cost? Four options, priced by how much of the business is in scope. Discovery is a £15k one-off: the value stream map, the AI readiness review, the watch-outs and a build plan you own and can hand to anyone — there is no obligation to use us afterwards — and if you do go ahead with a build, the whole £15k is credited against it. Solving one area, usually operations, is £60k a year, billed monthly at £5k. Mapping the whole business, with safe PII-restricted self-service analytics through Claude, is £120k a year at £10k a month. Enterprise adds autonomous agents and is priced against the outcome it delivers rather than from a list. After the twelve-month build you can walk away, or keep us on for £2k a month on a rolling monthly contract so nothing falls over. For scale, one area a year is less than half a fully-loaded senior data hire at around £143k. Each tier covers a stated number of source systems and legal entities rather than a revenue band, because a large simple business is a cheaper build than a small complex one: Whole business covers one entity, one country, up to 12 source systems, and anything past that — group structures, several countries, more integration — is scoped as Enterprise. ### We are a large or multi-entity group — which tier applies? Enterprise. The published tiers are bounded by scope rather than by turnover: Whole business covers one entity, one country, up to 12 source systems, and One area covers one value stream, up to 6 source systems. Past those limits — several legal entities, more than one country, or more integration than that — it is scoped and priced properly instead of squeezed into a tier. That cuts both ways: a large, structurally simple business often lands inside a published tier, because what drives the cost is the number of systems and the number of people who have to agree a definition, not revenue. ### Can we just buy the strategy and build it ourselves? Yes, and the Discovery tier exists for exactly that. For £15k you get your value streams mapped, the leaks quantified, an honest read on whether your data can support AI yet, the watch-outs, and a prioritised build plan. It is written to be acted on by somebody else — your own team, or another supplier. No proprietary format, no dependency, and nothing held back to protect a follow-on sale. And it is a down payment rather than a sunk cost: if you do come back for the build, the whole £15k comes off it. ### How long before we see something? Within a week. We connect the first source, land the events and put a real number in front of you — not a plan for a number. The layer takes shape over the first month, and the Summary Page and reporting behind it go live in months two and three. There is no six-week discovery phase that produces a document. ### Do we own the environment and the data? Yes. Everything is built in your own Google Cloud project, inside your own secure VPC. You own the environment, the data and the models throughout — we are working in your account, not hosting you in ours. At the end there is a structured handover with everything Terraformed and documented, so you can run it without us. ### What happens to personal data? Personal data is obscured at ingest, so it does not travel into the modelling layer or into any AI prompt. Self-service analytics through Claude is PII-restricted by design. We only ever need read access, with personal data excluded. ### Does this replace our data team? No — it means you do not have to build one before you get value. We work alongside whoever you already have, and the point of the engagement is that your team is self-sufficient well before twelve months. Then we get out of the way and you use us where it actually matters. ### What technology do you use, and why that stack? Google BigQuery for where the data lives, shadcn for what you look at, and Claude for who you ask. It is all Google underneath, in your own secure environment, because that stack scales, integrates with everything and lets us start immediately rather than spending a quarter on procurement and platform choices. ### Which sectors do you work in? We work where operations are complex enough to need a real data function: Consumer Credit, Debt Management, Credit Unions, B2B Services, Wholesale & Distribution, Manufacturing, Omni-channel Retail, Legal Services, Customer Service, Recruitment & Training, SaaS & Startups. The team has held C-suite positions across marketing, operations, technology and product, so the value streams in those sectors are familiar rather than newly researched. ### Can you take over an existing BI or reporting setup? Yes. Taking over reporting somebody else built is one of the most common ways engagements start — usually when the manual reconciliation behind it has become the bottleneck, or when the person who understood the workbooks has left. ### What do you mean by a second brain? A layer that remembers everything the business has done — every activity, every decision, and what happened next — and can be asked about it in real time. For an individual it answers what they would otherwise have to ask the colleague who has been there fifteen years. For the business it means that judgement stops living in a handful of heads: when somebody resigns the reasoning stays behind, and when somebody joins they start with the whole history. Recommendations are generated inside your own policy rather than filtered afterwards, the priors come from decisions your business has actually made rather than from industry averages, and every recommendation is scored against what happened next so the following one is better. ### Can you prove our data retention policy is actually being followed? That is the work. Most retention policies exist as a document and nothing else, because nobody knows which of a dozen systems hold a given record and deleting it from the CRM does nothing about the copy in the warehouse, the reporting database or a spreadsheet on somebody's desktop. We map where personal data actually lives across every connected system, express your retention rules as code against that map, run them on a schedule, and keep an audit trail of what was deleted, when and under which rule. You set the policy — we make it happen and produce the evidence that it did. ### Can you support data destruction and right-to-erasure requests? Yes. An erasure request is the same problem as retention, one record at a time: the difficulty is not deleting, it is knowing everywhere the record exists. Once personal data is mapped across your systems, a request can be actioned once and evidenced across all of them, with a record of what was removed and when. That record is usually what an auditor or the ICO actually asks for. ### Our marketing list is old — can you tell which contacts we can lawfully email? We can show you which contacts have an evidenced lawful basis and which do not. Every contact gets traced back to the event that created it — the form submission, its timestamp and the wording that was on the page at the time — or to nothing at all, which is the common case for anything more than a few years old. The list then splits into contacts you can stand behind and contacts to suppress. The decision about what basis you rely on is yours and your counsel's; we produce the evidence to make it with. ### What is the Summary Page? One page showing the whole business in real time: a balanced scorecard across finance, customer, operations and people with every measure against target and RAG derived from the metric tree rather than typed in, the quarter's objectives tracked against pace, and the three things that most need attention today. Clear one and the next moves up. ## Writing — full text ### From SQL Server and spreadsheets to a live data platform Case study, published 2026-08-21, by Shaun Adams. https://www.actadata.co.uk/blog/point-in-time-reporting-in-three-months A digital-first debt resolution agency with no way to see its book as it stood last month. Three months later: a nightly feed into BigQuery, 30+ governed definitions, point-in-time history, and seven live reporting products the team run themselves. #### “As of now” is not a reporting position The Digital DRA manages customer accounts at scale for energy, telecoms and consumer-finance clients. All of the operational data sat in SQL Server, and every report came out of it into Excel, by hand, every month. The bigger problem was not the assembly work. It was that every figure was as of now. There was no way to see the book as it stood at a past month-end — so nothing could be tracked month on month. Not arrears movement, not collections performance, not whether a treatment was working. That is a reporting gap that quietly becomes a regulatory one. Consumer Duty asks you to demonstrate outcomes over time. If your systems only hold today, you cannot show a trend, because the trend was never stored. > A system that holds today's position holds no history. There is nothing to trend, because nothing was kept. #### The definitions were the real work Definitions varied between spreadsheets. So every new question — from the board, a client, or the FCA — started from scratch, and the answer depended on which workbook you asked. This is where reporting projects actually fail, and it is not a tooling problem. We encoded the definitions once, as 30+ governed views: revenue, collections, contact, service, Consumer Duty. One place where “an account in arrears” means one thing. Then the part that made it trustworthy on day one: every figure was validated against the client's existing board numbers before go-live. Not reconciled afterwards — matched first, so nobody had to take the new platform on faith. #### Point-in-time history for the whole book A nightly feed lands the data in their own BigQuery environment, with snapshot history behind it, so any figure can be reconstructed at any month-end. That single capability is what turned a monthly assembly job into a platform. Once the past is stored properly, “how did this look in March” stops being an archaeology project. Seven reporting products now run on top of it: an FCA Consumer Duty board report, a monthly client pack with a full audit trail, an operations pack that generates its own board PDF, and per-client reports — all in DRA's own brand. #### It was built with their team, not around them DRA's team built the nightly feed and supplied the operational knowledge. We designed the data model, encoded and reconciled the definitions, and shipped the reporting suite iteratively. That split matters. The operational knowledge — why a number moves, which exception matters, what a client needs to see — was already in the building. It usually is. What was missing was somewhere to put it. #### Then we gave it to Claude A governed layer with the definitions written down and personal data stripped out is exactly what an AI assistant needs and almost never gets. Point a model at a folder of spreadsheets and it answers confidently from whichever one it was handed. Point it at a layer where “an account in arrears” has one meaning, and it reasons over the business. So the reporting suite became prompt-driven. Tom Hill, their COO, changes his own reports — no ticket, no developer, no waiting for us. Claude is now used across the organisation rather than in one corner of it. The technology was available to them before we arrived. What was missing was the layer underneath it. #### What we handed over We showed our work throughout, validated every number against what they already trusted, and handed over a platform they own outright in their own cloud environment. That is the model. We set it up properly, and you take it as far as you want — with us, or on your own. The measure of the engagement is not how long we stay. If your regulatory reporting is assembled by hand every month, and your systems only hold today's position, the problem is not the spreadsheet. It is that nothing underneath it was ever built. Three months is a realistic timeline for fixing that. --- ### Your data team is a folder on a shared drive Insight, published 2026-08-10, by Shaun Adams. https://www.actadata.co.uk/blog/your-data-team-is-a-folder It is called Reports, it has 240 files in it, and every one of them is a question the business asks every week. That folder is not a failure — it is a specification. Your data team is a folder on a shared drive. It is called Reports, it has about 240 files in it, and it is the reason the business works. I have been in enough builders merchants, wholesalers and family manufacturers to know that folder by heart. Change the company and the file names barely change: - Stock take March FINAL v4.xlsx - Debtors chase list (Dave's copy).xlsx - Margin by branch — DO NOT EDIT.xlsx - Quotes outstanding wk32.xlsx - Price list 2026 (new) (2).xlsx - Van costs Sheet1.xlsx - Credit limits master.xlsx - Supplier rebates Q3.xlsx Every one of those is a question the business asks every single week. None of them talk to each other. None of them have any history. All of them depend on somebody remembering to update them. #### Nobody says the next part out loud Those spreadsheets are not a failure. Somebody built each one because the system would not answer the question, and the business has run on them ever since. That person — usually one person — is doing a data engineer's job by hand, on a Sunday, and has been for years. The risk is not that the workbooks are wrong. Mostly they are right, because whoever maintains them knows exactly where the edges are. The risk is that all of that knowledge is undocumented, unversioned and resident in one head, and the day it walks out of the door the business loses the ability to answer its own questions. #### What each of those files is actually costing Taken one at a time, the same findings come up in merchant after merchant: - The stock take tells you what was on the shelf in March. It does not tell you which lines have not moved since, or what that cash would be worth doing something else. - The debtors list is somebody's personal copy, so the chase happens when they remember rather than when an account crosses a limit. - Margin by branch is gross margin, so the branch that absorbs the deliveries and the returns still looks like the good one. - Quotes outstanding has no win rate in it, and no record of which quotes went cold or why. - Two price lists are in circulation and the trade counter has the older one. - Van costs sit in their own file, so cost to serve per drop — where the margin actually goes — is never in the same place as the margin. - Credit limits were set once, years ago, on customers who have since doubled or halved. - Rebate thresholds get hit or missed without anyone noticing until the quarter closes. Not one of those needs AI to fix. They need the eight files to be one thing. #### Why it has become urgent rather than annoying Manual reconciliation was survivable while a report was the endpoint. It stops being survivable the moment you want software to act on the numbers. An agent has no way of knowing that the third tab is the one to believe. It cannot see that Dave excludes inter-branch transfers, that the March file was never finished, or that the margin column carries a manual adjustment somebody typed in during a stocktake two years ago. Point a model at that folder and it will answer confidently and wrongly, at speed, to more people than the workbook ever reached. > You cannot put AI on top of a process that gets settled by human judgement in a workbook every month. #### What replaces it Not a dashboard. The eight questions in that folder become one layer: every source connected, every activity recorded once with the time it happened, and the definitions written down as code rather than remembered. After that the questions are queries against one thing, instead of eight files that have to be reconciled before anyone can answer anything. The person who currently maintains the folder does not lose their job. They stop being the pipeline and start being the person who says what to do about what the numbers show — which is what you hired them for in the first place. #### And you are not too small If you are a merchant or a distributor reading this and thinking data and AI is not for you, it is more for you than it is for the tech companies. You are the one still doing it by hand. You are also the size where it pays back fastest. A large corporate needs a multi-year programme and a steering committee, because it has forty systems and nine countries to reconcile first. A hundred-person merchant needs a few months, because the whole operation genuinely fits in one layer — one stock system, one finance system, one CRM if you are lucky, and the folder in between. And the payback is not a nicer report. It is one dead product line cleared, one bad account repriced, one rebate threshold hit that you would otherwise have missed. > That folder is not evidence of a problem. It is a specification. Every file in it is a question the business already decided was worth answering every week, and somebody has already done the hard thinking about what matters — they just had to do it in Excel. Build the layer that answers those eight questions properly and you have not started a data project. You have finished one that has been running by hand for years. --- ### Nobody is drowning in data. They're drowning in reconciliation. Insight, published 2026-08-08, by Shaun Adams. https://www.actadata.co.uk/blog/drowning-in-reconciliation Month-end isn't slow because the numbers are hard. It's slow because five people are proving to each other that their versions agree. Every operations team we walk into has the same Monday. Someone exports a file. Someone else exports a different file. A third person joins them in a spreadsheet, notices the totals are out by a few hundred, and spends the morning working out which one to trust. By Wednesday there is a number everyone accepts. By Thursday it is out of date. The instinct is to call this a data problem. It usually isn't. The data exists — it is in the loan book, the CRM, the payment provider, the call system. The problem is that nobody has agreed, once, in one place, what a customer is, what a sale is, or which timestamp counts. So every report re-litigates it from scratch, by hand, in a workbook one person really understands. #### What that actually costs - The obvious cost: senior people spending days a month producing numbers instead of acting on them. - The cost nobody books: decisions made on the newest figure someone happens to trust, rather than the right one. - The cost that compounds: when the analyst who owns the workbook leaves, the reporting leaves with them. - The cost that stops you dead later: you cannot put an AI agent on top of a process that gets settled by human judgement in a spreadsheet every month. That last one is why this has moved from annoying to urgent. Manual reconciliation was survivable when a report was the endpoint. It is not survivable when you want software to act on the numbers, because an agent has no way of knowing that the third tab is the one to believe. #### The fix is boring You define the events once — a customer was created, an application was submitted, a payment was taken, a call was answered — and you store them immutably, with the time they happened. Then every report, every dashboard and every agent reads the same events. > Two people can still disagree about what to do. They can no longer disagree about what happened. That is not a clever piece of engineering. It is a decision about where the truth lives, taken once and then enforced. What follows is well-trodden: BigQuery to hold it, a modelled layer on top so the definitions are written down as code rather than remembered, and one page that shows the business. #### How you know you are in the trap - The month-end number changes depending on who produces it. - Someone maintains a workbook whose logic is not written down anywhere. - You have been asked for last year's figures and had to rebuild them. - Board packs are assembled by hand, not generated. - Your best analyst's week is full of extracting rather than thinking. If three of those are true, the reporting is not your bottleneck — the absence of an agreed layer underneath it is. Remove the manual reporting entirely and the analyst you already employ becomes the person telling you what to do about the numbers. --- ### Quality data first. Then the agents. Insight, published 2026-08-08, by Shaun Adams. https://www.actadata.co.uk/blog/quality-data-before-agents An AI output is only ever as good as what sits underneath it. Most failed AI projects are data projects that got skipped. There is a version of AI adoption that goes: buy the licences, connect the tools, tell everyone to use it, wait for the productivity. It generates a lot of activity and very little operational change, and the reason is almost never the model. A language model is very good at reasoning over what you give it, and completely indifferent to whether that thing is right. Point it at a warehouse where “active customer” means four different things and it will answer confidently, four different ways, and none of the answers will arrive flagged as suspect. You have not bought intelligence. You have bought a confident guessing machine — and one your team will believe for a while. #### What an agent actually needs When we put an agent into an operation, it needs three things that have nothing to do with AI: - Definitions written down as code rather than held in someone's head, so the same question always resolves the same way. - Events with reliable timestamps, so “what changed this week” is a query and not an argument. - A boundary around what it can see, so personal data does not end up in a prompt because nobody thought about it. With those, an agent is reasoning over facts and becomes genuinely useful. Without them, every output needs a human to check it — which is exactly the cost you were trying to remove. #### The order matters more than the ambition We build the data layer first, every time, including when the client came to us for the AI. That is sequence, not caution. The layer is what makes the agents trustworthy, and it is also the part that keeps paying off when you change your mind about which model or which vendor to use. > Models will keep changing. A clean, well-defined event history will not. #### Then put your people where it counts The point of the agents is not headcount. It is that repetitive work stops consuming people who are good at judgement. Someone spending their week pulling reports, chasing exceptions and re-keying between systems is someone not spending it on a customer who needs a decision made properly. So the sequence is: quality data first, then agents on top of it, then your people deployed where a human genuinely does it better. That last part is not the consolation prize. It is the whole return — nobody ever won a customer because their reporting reconciled. --- ### Cost, revenue, conversion, time Insight, published 2026-08-08, by Shaun Adams. https://www.actadata.co.uk/blog/cost-revenue-conversion-time Four atomic units are enough to describe almost any operation — and to show you where the value is leaking out of it. Most reporting is organised by department, because that is how the org chart is organised. Marketing has its numbers, operations has its numbers, finance has the ones that count. Each set is internally consistent, none of them join up, and so nobody can answer the question that actually matters: where is effort going in and value not coming out? We model operations differently. Every activity in a business, whatever the sector, moves at least one of four things: - Cost — what this step consumes. - Revenue — what it brings in. - Conversion — whether the thing progressed, and to what. - Time — how long it took, and how long it sat waiting. Those are the atomic units. Everything else — channel, product, region, underwriter, adviser, tier, cohort — is context layered on top. It sounds like a simplification and it is the opposite: once every activity carries those four measures plus its context, you can cut the entire business the same way and the totals still agree. #### Why this finds the leaks Value streams leak in the places no departmental report looks, because the leak happens between two departments. A quote that converts brilliantly and then takes nine days to fund. A channel with excellent conversion and a cost per completed case that makes it the worst one you run. A step nobody owns, where a fifth of cases wait a week for a document. You can only see those when cost, revenue, conversion and time are measured on the same activities, in the same units, against the same definition of what progressed. That is what the layer is for. #### What you do with it The output is not a bigger dashboard. It is a shortlist. When the whole stream is visible in one place, the argument stops being whose number is right and becomes which of these three things we fix this quarter — and you can put a value on each one before committing anybody to it. > The output is not a bigger dashboard. It is a shortlist. It also gives you somewhere sensible to point the agents. Something chasing the documents that hold up a fifth of your cases is worth more than a chatbot on the front page — and you only know that is the bottleneck because you took time as seriously as you took revenue.