This is an example of ad-hoc, on demand reporting, where a manager or team needs the data as-at right now, but they don't want to spend 20 mins formatting and calculating a report.
Instead of waiting on someone in admin to prepare the WIP report from a raw CSV export.. We drop that CSV into a "data" folder and hit refresh in Excel. 20-mins becomes less than 20-secs.
Something like this we'd build in an afternoon, with enough time left over to build example 2.
The main purpose of this example is to demonstrate the List report and generally how we save you time. Also includes a basic pivot report for things like Changes in invoiced amounts VS prior export
We take an "opportunities" export to CSV and build a Won-Lost-Pending Sales dashboard.
Aswell as some Targets VS Actuals analytics off a quick input "Targets" table.
The main purpose of this example is to demonstrate the simplicity, dynamism & power of a Pivot report. A list report is also included for filtering & inspecting row-level data.
There's less studies available on analytics, than there is on time-spent in Excel (top of page) though complementary to automated data pipelines, analytics is our biggest value-add.
Most SMB's want to be more data-driven. With enhanced clarity, decisions are made faster & with greater confidence.
"We're data rich, but between systems that don't communicate, raw exports & manually tracked spreadsheets, we're often operating with reporting delays and lacking visibility."
Many office-based tasks are at their core, moving data from point A to point B. Often with approvals & processing steps along the way.
"We've got workflows that we know could be automated, but we're not sure how, or where to begin. Most of our processes sit between Spreadsheets, our CRM/ERP, Outlook & some file templates."
AI isn't an everything fix. There were automations available to you long before AI, many of which are more impactful to your business.
"We'd like to leverage AI, we know there's upside potential, but we're weary of unknown-unknown risks of a new technology. We'd like best-practice training, and projects that make practical sense."
We build a data-model off your system exports (csv/Excel) and any kind of manually-tracked spreadsheets in your business.
It's difficult to describe the outcomes, it's far better to show, not tell. Hence the example video above.
We build a data-model of your business. Which, if you've never hired a dedicated data-analyst before, delivers a level of clarity that you didn't realise was possible.
For whichever datasets are deemed important, we build a data-model of the business. This could mean:
We tag & enrich your data in simple, practical ways for the end-users of any report or dashboard.
This could mean:
There's often a data-integrity/cleanliness aspect to begin with, that most businesses aren't expecting, but are often very grateful for.
Analytics is tightly woven with the Automation side of our business, given that most of the time we save clients, is by automating their manual data-processing & reporting workflows.
This is the core of our business, we do it very well.
Taking system exports from Accounting/ERP Software to deliver superior Financial Reporting, Variance Analysis & Time-Series Forecasting.
If what you're looking for isn't provided by your software (common), we solve that by exporting the raw data to csv/excel and building everything tailored to your needs.
For example, Slice on Customer vectors like;
Most business software is a relational database with a pretty user-interface. As a result, your business operates off distinct operational ID's (job_number, case_code, order_ID, project_ID, opportunity_number, supplier_ID, or asset_ID etc..).
Operational analytics revolves around the tracking of those unique ID's, building data pipelines that ingest regular system exports to build out your businesses data model & workflows.
The ID's change from business to business, and so do the requirements, but the approach remains the same.
By tracking these items over time in an automated, improved way, the whole organisation operates more proactively and efficiently when compared to individual teams/dept's running their own reports.
With a well-defined data-model, we can do two things:
CRM exports, with Historical Data Mining and Database Enrichment.
Things we work on:
One data model, multiple lenses. Board sees trends and risk. Managers see KPIs and comparisons. Teams see today’s priorities.
Definitions are fixed, so when the numbers change between meetings, it's because the numbers changed. Drill down from totals to the line item in seconds. Only one person is required for your businesses reporting & analytics tasks, their task now done in minutes, not hours or days.
Without a doubt, the most time we've saved for clients is by automating previously manual reporting & data-processing inside of Excel.
We broadly classify anything that starts & ends in Excel as "Reporting Automation".
Everything else, we'll group as:
But all automation, we like to think of as a data-pipeline, with three main parts:
This is our bread & butter for saving you time.
The approach is one of PowerQuery for automated data-processing (And excels data-model for analytics). We do very little in what most people think of when they hear the word "Excel". Excel is the best end-user interface for business data-workflows, it's simple, ubiquitous & dynamic.
For perspective, we work with a few accounting firms, none of them, nor most other businesses we work with, understood what can be done to improve spreadsheet-related workflows until we started working together.
Point being, if accounting firms are having us help with their financial reporting & excel-workflows, we're doing something right.
Ties in with Operational analytics as we track the movement of operational ID's.
Pipelines can be simply transferring data from point A to point B.
Ie, an event is triggered by the arrival of an order confirmation email, after savng it down, data needs to be extracted from that email/attachment, then updated into either a shared Excel spreadsheet or into your ERP. Sometimes requiring an internal email to notify another department that said fields have been extracted & updated, so that the next step in the process can commence.
This is the type of thing we look to automate. Or at the very least, improve.
Outside of automated reporting, field-processing automation can be done with either custom python scripts or the standard suite of workflow automation tools (Zapier, Make, Power-Automate etc).
For the standard workflow automation tools, we generally suggest that clients have some staff members learn the basics. It's a waste of our time and your money for us to be building drag & drop automations.
From experience it's usually a younger, tech-savvy team member who's keen on filling that role, sometimes they've already built a few automations for themselves or the business. We can help them here if they get stuck.
Any time your business receives a system-generated email, requiring someone to monitor, search the inbox, manually extract information from the email (or attachment), maybe save down said attachment and then action the data that was contained within...
Then that's an opportunity to not only free up time, but significantly improve the timeliness & consistency of that whole process.
Most commonly, emails from suppliers about order updates. System generated or otherwise.
On the flip side, it's less common, but if your business is manually generating emails from templates, with only the fields of data or attachment changing from client-to-client or case-to-case.
Then the process of populating, sending & monitoring for responses in a more structured way, is something we can help with.
PDF parsing can sometimes require AI, it depends on the variability of the input docs.
Ie, A highly structured govt/compliance document, same headers/formatting each time, no AI Necessary. But for a Semi-structured pdf, with greater variation between them, then AI might be the best option. An example of this might be a legal doc like a lease, sure the lease is different each time, but there's enough similarities that they can be grouped together. Maybe from that you want to query in natural language, information about the lease.
The most common type of docs we've done these projects for are actually docs required in Onboarding procedures. But also complex PDF's with multiple pages and a table of contents, contracts for example.
On the web-scraping side of things, it's often for scraping prospect lists, but it depends on your business. There's been some really creative ways people have gotten value from our web-scraping projects.
Start Small, Keep it Simple, Practical & Low-Risk
Our approach for everything is to keep it simple, same goes for AI.
For context, whenever somebody says 'AI' nowadays, they almost certainly mean a Large-Language-Model (LLM).
The concept of a Large-Language-Model (like ChatGPT) only entered into existence in 2017 with the release of the research paper, "Attention is All You Need". Link to paper here.
And most businesses really only started hearing about, or using AI around 2022 / 23 onwards.
Point being; in the timeline of our glorious universe, LLM's are in the pre-conception phase. There's "unknown unknowns" even to the people creating the models. Let alone the businesses attempting to leverage them to gain a competitive advantage.
It's a very common occurrence when a business asks us, "Can we automate this with AI?"
9 times out of 10, our answer is "No, but we can automate it without AI."
Many businesses still lack a unified understanding of:
Tom's conducted over 100 AI Basics training sessions to businesses ranging from small one-man-band realestate agencies in Perth, to $1b plus ASX listed entities.
The combined feedback from those sessions has been extremely helpful, not to mention that most recipients find the training itself immensely valuable.
Tabular data (database / csv / excel) and other forms of highly-structured data are not things we need AI for.
It's the semi-structured & unstructured end of the spectrum where AI dominates.
For extracting structured data from PDF's for example, AI's not always required, and we'll tell clients if that's the case. Suggesting instead, a much more reliable, deterministic pure-python script (No AI).
But sometimes those python scripts will require an AI step or two. Simply put, any semi-structured or un-structured text data in the form of:
... these are the types of things we consider using AI for.
Often times, the most involved part of those pipelines is in building the pre & post-processing steps.
