Small bets, big wins

The most successful systems we’ve developed began with a Visionary.

Not always someone with a grand title, just someone doing their job who saw a better way. Not just better for themselves, but for their team, and the organisation.

We call them Visionaries because plenty of people notice problems, but very few actually do something.  But then a lot of ideas aren’t worth the effort. The hard part is knowing when one is.

From idea to system: the process

When a Visionary comes to us with an idea, we follow a simple discipline:

1. We understand what they want to achieve — and whether we’re the right people to help. Not every problem needs a system, and not every system needs us.
2. We design a high-level architecture. A good system with a poor architecture will almost always fail.
3. Then we build a prototype — a deliberately small part of the system. This is often to test out a high-risk area, but it could also be part of the UI so people can see how it would work. But it’s always limited, and always designed to move the thinking forward.

The power of the prototype

Today this is often called a proof of concept. Sometimes it’s just “let’s see if this works”.

Names don’t matter. What matters is that the prototype tests whether something actually works — technically, operationally, or commercially. (In theory there’s no difference between theory and practice. In practice, there is.)

Because it’s a small bet, it can be thrown away. We often throw several prototypes away, because each one is small. But each time, we’ve learned something and the next step is clearer. We might refine a prototype or build something else. Each is a small bet, designed to reduce risk and build confidence.

Eventually, the prototype stops being a prototype and becomes a system. The Visionary can get budget to develop it further because everyone can see how it will work. 

Do you have a visionary idea?

Many of the systems we’ve built have gone on to have long, productive lives. Mostly well after the original Visionary has moved on. They didn’t know at the start whether their idea was large or small; they only found out by making small bets.

If you have an idea that might be a better way, and want to explore it further, let’s have a conversation.

A light-weight way to solve heavy data problems

Most people want more from their data, but knowing how best to use it can be difficult. There’s always a question mark over whether the development effort will have a big enough pay-back. Smart leaders know that everything has a cost – in management attention and budget.

But doing nothing is not the answer. Data-driven organisations are out-performing those less skilled in data analysis. So, what’s the answer? 

At Anatec AI we are big fans of the proof of concept or prototyping. The idea is to do a relatively small amount of work that does a lot of heavy lifting. It might focus on analysing a limited amount of data to assess the data quality. Or it might be to test out a theory on how two data sources could identify patterns. Or it might be a difficult interface that absolutely has to work for the solution to be a success. Each project will be different, because each problem is different. The approach, however, is consistent: do a small amount of high-risk work that tells you whether or not you are on track. 

The cloud provides a great way of doing light-weight small projects – you don’t need to buy servers or install operating systems. You don’t need to commit up-front. You can just use the cloud’s in-built flexibility to do just as much as you need to do, then decide what to do next. This is a lot safer, and easier to justify than ambitious projects that can end in mediocre results.

Azure containers are a great fit for that way of working. They are reliable, whether running locally or in the cloud. And they can securely access data that’s held in databases, cloud storage, queues or APIs. Of course, containers are just one example of a light-weight approach to our small-project ethos, but they are a good example of how smart Azure technology can produce big results at a small cost. 

The Anatec AI design approach underpins everything we do when architecting modern cloud systems. Our solutions need to be reliable, portable, and consistent across environments - and developed one bit-sized step at a time. Because our clients need to be in charge of the direction of the project and the budget, not the technology.

Do you want to get more from your data without breaking the bank? As qualified Microsoft developers, we’ve got a whole toolbox full of Azure tools to engineer data-driven solutions. If you’ve got an idea for your data that might be a game-changer, get in touch to find out more about how we work.

Keywords: Azure, containers, proof-of-concept, cloud native, data analytics, Azure AI

Three steps to optimising learning at work

Whilst everyone agrees that learning at work matters, not all organisations are good at knowing whether their learning programs actually work. That’s because optimising learning is messy. It needs data from multiple systems, departments, and stakeholders.

Historically, L&D hasn’t rushed toward that complexity. But that’s changing.

Three things have shifted the landscape:

1. Learning management systems (LMS) are now widespread, and they hold large amounts of learning data.
2. Cloud platforms enable data to be combined and analysed without the need for huge data warehouses. That makes learning analytics projects much more cost effective.
3. AI tools are available off the shelf and viable for small, focused experiments.

The result is that learning analytics is no longer theoretical or reserved for large enterprises. It’s affordable, practical, and increasingly hard to ignore.

So where do you start?

What is learning analytics?

It’s always useful to start with a definition, and learning analytics is pretty straight forward:

Learning analytics improves learning outcomes by analysing data.

It may not be complicated, but it’s still useful. And it also gives you the correct order to do things - an order many organisations still get wrong.

Step 1: Define what “better” looks like

You don’t start with data. You start with outcomes. What are you actually trying to improve?

• Fewer errors?
• Higher sales?
• Faster onboarding?
• New skills that are being used in day-to-day work?

There is no universal answer. It will depend on the business context and the problems you’re trying to solve. But until you can clearly describe what success looks like, analytics will only produce dashboards, not insight.

Step 2: Decide how you’ll measure improvement

If you can’t agree on a measure, you can’t tell whether learning made any difference.

Choose metrics that reflect the outcome you care about, and that you can actually measure. That might include:

• Knowledge gained
• Reduced error rate
• Skills demonstrated
• Business performance indicators

Your metrics don’t need to be perfect they just need to be useful. The goal is better decision-making, based on data.

Step 3: Identify the data that supports those metrics

Most organisations already collect far more data than they use: LMS activity, course attendance, assessment results, feedback surveys, HR data, performance metrics. The problem isn’t that the data doesn’t exist, it’s that it’s siloed. 

When data from multiple systems is brought together, patterns start to emerge. Training volume alongside retention. Learning pathways alongside performance. Timing of learning alongside error rates.

Putting it into practice

This approach works best when treated as iterative. Small proof-of-concept projects using modern business intelligence tools and AI can deliver fast feedback on existing learning initiatives.

You don’t need a multi-year transformation. You need a focused question, a clear metric, and enough data to test whether your learning is actually helping.

If you are looking at getting more from your learning data, and improving learning outcomes, get in touch for a no-obligation chat. 

Keywords: learning analytics, data analytics, learning management system, LMS, Power BI, L&D dashboard, Azure AI