Operations manager
Orders in progress by stage, late deliveries flagged, and capacity against demand for the next seven days.
Data analytics · Glasgow
Dashboards, reports and forecasting you can act on
Most businesses have the data they need, spread across five systems and a spreadsheet. We bring it together and keep its history, then turn it into dashboards, reports that build themselves and forecasts you can check, so decisions rest on today’s numbers.
Est. 2019 · Glasgow · You own the code and the data
The Friday report, before and after
Before
- Export from the CRM and the accounts system
- Paste it all into Excel and format it
- Check the numbers and send it by five
- Out of date by Monday, and missing when its builder is on holiday
After
- The data arrives from every system without anyone touching it
- Each load is checked, and nothing is dropped silently
- The report is ready on Monday, with unusual changes flagged
- A person reads it and decides what to do
A person checks
Dashboards built around the person looking at them
What a dashboard shows depends on who’s using it, so we start with the person and the decisions they make, not a generic template.
Sales director
Deals at each stage, the value closing this month, progress to target, and deals that have stalled.
Warehouse manager
Stock at each site, the pick queue, dispatches due today and low-stock alerts.
Managing director
Cash in the bank today, revenue this week against last week, and your top three goals, each with a traffic light.
Live screens, trend reports and the weekly pack
- Operational dashboards answer “what’s happening right now, and what needs my attention?” They refresh within seconds or minutes and alert you by email, Slack or Teams when a number goes the wrong way.
- Business intelligence answers “what happened last quarter, and what should we do next?” It updates overnight and shows patterns across months or years of history, the kind that belong in a board pack.
- Scheduled reports replace the one someone builds by hand every Friday. The figures are collected from every system and checked, then sent on a schedule, with each week compared with the weeks before.
Every chart answers one question, with the answer in the title, drawn to true scale so a small change never looks like a big one. Click through to the detail behind any number, or export any view to Excel.
Live
Client scorecards and weekly KPIs that flag themselves
For a warehouse analytics platform used by third-party logistics companies, we built a scorecard for each of their clients. Each client’s week is compared with the weeks before it, and unusual changes are flagged automatically, so the team looks at the exceptions instead of every number.
Read the case: Client scorecards and weekly KPIs that flag themselves
Data pipelines that keep the history
Every dashboard and forecast rests on a pipeline that collects the data, cleans it and is honest about where it came from.
- From anywhere. APIs, public registers, bulk files, exports and the odd spreadsheet.
- History kept. When a source only shows today, we keep yesterday, so you can see what changed and when.
- Reconciled. Every load is checked: counts match, nothing dropped silently, and rejects are logged.
- Matched. The same company or customer linked across sources, with a method and a confidence score for each match.
- Watched. Alerts when a feed stops, slows down or changes shape.
Public doesn’t mean free of rules. We check the licence and the data-protection position of every public source before we build on it.
Pipelines we’ve built
A national register, monthly
The full Companies House company snapshot (about 5.7 million live companies) is loaded every month and swapped in so nobody reading it is ever blocked.
Live
ERP to warehouse, every 15 minutes
A client’s Business Central orders are collected every 15 minutes, converted for the warehouse system, and confirmed back once shipped.
Forecasting: what can be forecast
If you’ve recorded it over time, there’s a good chance it can be forecast.
Demand
How much you’ll sell, and when you’ll need to restock, by product, site or week.
Churn risk
Which customers are likely to leave or not buy again, so you can act while there’s still time.
Prices
Where a price is likely to land, based on your own sales and the factors that move it.
Time to sale
How long a deal, a job or a property is likely to take to close.
How we test a forecast
A forecast is only useful if you know how far to trust it. Before you rely on one, you see how it did on your own history.
Tested on held-back history
We hide the most recent stretch of your data from the model, forecast it, and compare the forecast with what really happened.
Honest ranges
Every forecast comes with a range, not a single confident number, so you can see how sure it is.
Features explained
We show which factors drive each forecast, in plain words, so you can check it against what you know.
A simple baseline to beat
Every model is compared with a plain average or last year’s figure. If it can’t beat that, we tell you.
When a simple average beats a model
If your sales are steady, last year’s figure with a seasonal adjustment may be as accurate as anything cleverer, and much easier to explain, so we test the simple option first. And chat-style AI isn’t built to forecast numbers: it can explain a forecast in plain English, but the number should come from a model tested on your data.
Ask your data a question
Once your data is in good shape, you can ask it questions in plain English, such as “which clients had no orders last month?” The answer comes from the same functions that draw your screens, so the number matches the screen, with a link to it. The assistant can read your data but can’t change it.
AI can also add a plain-English summary of what changed this week to the Monday report. It can get things wrong, so it links to the figures it describes and you can check every line in a click. See assistants for your data.
How we build it
-
Pick the person
Which decisions do they make every day, and which numbers would change them?
-
Map the data
Accounts, CRM, online shop, support desk and spreadsheets, such as Xero, Sage, HubSpot, Shopify and Zendesk. We connect to each one.
-
Build, show, refine
Working with your own data from week two, and refined every fortnight.
-
Run it alongside
New reports run next to your manual ones until you trust them.
-
Host it properly
On AWS, mostly in London, usable on a phone, and each person sees only what’s theirs.
What it costs
-
Dashboards and data
From £5,000
The price depends on how many sources we connect and how many people need their own view. Fixed, and agreed before we start.
A forecast usually needs a clean data feed underneath it, so the quote covers both. We estimate the payback before you commit, and show you the sums.
Prices exclude VAT where VAT applies.
Related pages
- Assistants for your data
Ask in plain English and get an answer with a link to its source.
- Integrations
Connect the systems you already use.
- AI and workflow automation
When the numbers should trigger the next step.
- Running the business from spreadsheets?
When the spreadsheet has become the system.
Questions
How much does a data project cost?
Dashboards and data projects start from £5,000. The price depends on how many sources we connect and how many people need their own view. You get a fixed quote after the first call, so there’s no hourly billing.
How long does it take?
It depends on how many sources we connect and how many screens you need, so we agree the timeline with your fixed price. You see working screens with your own data from week two, and every two weeks after that.
Do we need a data warehouse?
Usually not at first. We often connect to your live systems directly, or keep a small copy where a source is too slow to query. If you outgrow that, we’ll build a warehouse when you need one, not before.
Do we need Power BI or Tableau?
No. We build in Python and open tools, so there’s no per-seat licence for a BI product you half use. If you already have Power BI or Looker, we’ll feed it properly instead.
How accurate will a forecast be?
We can’t promise a figure before we’ve tested on your data. We measure accuracy on history the model hasn’t seen and show you the range, so you know before you rely on it.
How much history do we need?
It depends on the question and how much your numbers move. We look at what you have before quoting, and if there isn’t enough to forecast reliably, we’ll say so.
Get in touch
Tell us the question your numbers should answer.
One paragraph is enough: which systems hold the data, and what you’d like to see or forecast. We’ll reply with honest thoughts on what’s possible and what it would cost.