Data & Analytics

You already have the numbers. You just cannot see them.

Most businesses do not have a data problem. They have the data, in four systems that disagree, and somebody rebuilding the same spreadsheet every Monday. That spreadsheet is the project.

We start underneath, not on top A beautiful dashboard built on figures nobody trusts gets opened twice. Where the numbers come from is the whole job; the charts are the last afternoon of it.
We have been on the other side of this Our own retail software has been recording sales, stock and ledgers in real businesses since 2001 — so we know what operational data looks like when nobody was tidying it for a report.
Our own software

Five views of one week, and they cannot disagree

This is a reporting screen out of our own product. Summarized is the one the owner checks; the other four are the same week broken down when a figure looks wrong. They are one click apart and every view is built from the same transactions — which is the whole of what a reporting brief is asking for, whatever industry it arrives from.

Click the modes. The net profit is 59,916.40 in all five.

RetailWiz — Income Statement. Generated from one transaction ledger, so two views of the same week can never drift apart. See it in context →

First

Three questions that decide the shape of it

These separate a fortnight's work from a six-month one, and they are worth answering before anybody quotes you.

1

How many places do the numbers live in?

One system and this is a reporting job. Four systems, a bank export and a spreadsheet, and the real work is joining them — which is where the time goes and where the honest quotes differ from the optimistic ones.

2

When two systems disagree, which one is right?

Somebody in your business knows the answer to this, and it is never written down. Getting it written down is half of what makes a dashboard trustworthy rather than merely attractive.

3

What decision would actually change?

The most useful question and the least often asked. If seeing the number would not alter what anybody does on Monday, the number is interesting rather than valuable, and we would rather build the one that changes something.

Answer the third one first. A dashboard with six figures that change decisions beats one with sixty that impress visitors, and it costs a fraction as much. If what you need turns out to be a single weekly figure emailed to four people, we will build that and tell you to stop there.

What we build

Five kinds of data work

BI dashboards

the thing people ask for

One screen carrying the figures you would otherwise rebuild by hand every week, updating itself. Sales, margin, stock, receivables, whatever it is you currently ask somebody for on a Monday morning.

  • The handful of figures that matter, not everything measurable
  • Refreshed on a schedule, with the time it last ran shown
  • Filters that answer the obvious follow-up question
  • Different views for the owner and the branch manager
  • Readable on a phone, because that is where it gets checked
  • Every figure traceable back to where it came from

The one feature that decides whether it survives: you can click a number and see what it is made of. A figure nobody can take apart is a figure nobody argues with, right up until the day it is wrong.

Data engineering

where these projects are won or lost

The pipes underneath — pulling data out of each system on a schedule, making the names and dates and currencies agree, and putting the result somewhere the reporting can rely on. It is invisible, it is most of the cost, and skipping it is why so many dashboards get abandoned.

  • Extracting from each system, including the awkward ones
  • Duplicates, spellings and formats reconciled to one standard
  • One agreed definition per figure, written down
  • History kept, so last year is still there to compare against
  • Failures that shout, rather than a dashboard quietly showing stale numbers
  • Runs on its own, and tells you when it did not

If a proposal you have been given spends most of its words on charts, ask what happens the week a system changes a field. That question is the difference between a dashboard that lasts and a screenshot that once looked good.

Predictive analytics

what the history suggests happens next

Using what has already happened to estimate what is coming — which lines will move, what to reorder and when, which customers have quietly stopped buying. Useful, and worth being plain about: these are estimates with a range, not prophecies.

  • Demand and reorder timing from real sales history
  • Seasonal patterns separated from ordinary noise
  • Customers whose buying has dropped off, flagged early
  • Stock likely to sit, and cash likely to be tied up in it
  • Every forecast shown with how far out it has been before
  • Built to be checked against what actually happened

Two years of clean history beats a clever method every time. If the history is thin or the business has changed shape, we will say so rather than fit a model to it and let you find out later.

Data science & analysis

the question behind the question

Not a dashboard and not a forecast — a specific question answered properly, once. Why did margin fall in the second quarter when volume did not. Which branches behave differently from the rest and what they have in common. Which of two things is actually driving the third.

  • One question, defined precisely before any work starts
  • The workings shown, so the conclusion can be argued with
  • What the data cannot tell you, stated as plainly as what it can
  • Customer and product segmentation from behaviour, not assumption
  • Pricing and margin analysis across lines, branches and periods
  • A written answer a person can act on, not a folder of charts

The deliverable is a decision, not a document. Sometimes the honest finding is that the data does not support the conclusion you were hoping for — and that is worth paying for too, because the alternative is acting on it anyway.

Web scraping & data collection

when it is not in a system yet

Gathering information that exists publicly but not in any form you can use — competitor pricing, listings, directories, published figures. Collected on a schedule and delivered as something you can sort, filter and compare.

  • Public pages collected on a schedule you set
  • Structured into a spreadsheet or straight into your database
  • Changes tracked over time, not just today's snapshot
  • Run gently, so nobody else's site suffers for it
  • Rebuilt when a source site changes shape, as they do
  • Rules a site publishes about automated access, respected

We will ask what you intend to do with it before we build it. Where a site's terms forbid collection, or the data is personal, the answer is no — we would rather lose the work than hand you somebody else's lawyers.

FAQ

Questions, answered

My software already has reports. Why would I need anything else?

If one system holds everything you need, you may not, and we will say so — our own products carry a great many reports for exactly that reason. The case for a dashboard begins when the answer lives in more than one place: sales here, purchases there, bank somewhere else, and a person joining them up in a spreadsheet every week. That weekly rebuild is the thing worth removing, not the reports.

Our data is a mess. Do we need to tidy it up first?

No — and waiting until it is tidy is how these projects never start. Every business we have opened the books on has duplicate customers, three spellings of the same supplier and a column somebody repurposed years ago. Sorting that out is the work, not a prerequisite for it. What we do ask is that you tell us where the bodies are buried, because you already know and it saves weeks.

Do you build the dashboard, or set up something like Power BI?

Either, and the honest answer usually depends on who maintains it after we leave. A ready-made BI tool is quicker to stand up and your own people can change it. A dashboard built into your own application makes sense when it must sit inside something you already use, or when the licences start costing more than the build. We will tell you which way the arithmetic points for your case.

Can you work with the data already inside RetailWiz?

Yes, and more easily than with most systems, because we wrote it. RetailWiz and Zeemovil are our own software and have been running in businesses since 2001, so we know exactly where the figures live and what they mean — which is normally the slowest part of a data project and, in this one case, is already done.

Is web scraping legal?

It depends entirely on the site and what you do with what you collect, and it is not a question to wave away. We work from public pages, respect the rules a site publishes about automated access, and keep the rate low enough that nobody else notices. Where a site's terms forbid it, or where the data is personal, we say no — and we would rather lose that piece of work than hand you a problem with somebody else's lawyers in it.

Not sure this is you?

The rest of what we do

Tell us which spreadsheet gets rebuilt every week

Fifteen minutes on a call, no charge. That one spreadsheet usually tells us more about what you need than a specification would.

💬