This is where support from the Estonian state comes in: several grants from Enterprise Estonia (EIS) can help move your development project to the next phase without your company bearing all the risk and cost. One grant is suitable for testing an idea, another one for creating a working solution, and the third one for conducting more extensive applied research. But which of them best fits your idea?

In this article, we provide a practical overview of three support instruments offered by EIS: the innovation grant, the development grant, and the programme for applied research. We explain which development phase each grant is suitable for, what distinguishes them, and how STACC can support you as a development partner.

Grants as stages of a development project

These grants form a kind of development ladder in terms of both scale and purpose. It makes sense to kick off your initial development with the innovation grant, which provides up to 7,500 euros in funding, complemented by company self-financing. If the result is promising and the need for development increases, you can continue with the development grant, which offers up to 35,000 euros in funding and can cover costs such as creating a working solution or minimum viable product (MVP). If the first version proves viable but achieving commercial success requires more in-depth technology research and product development, the programme for applied research offers up to 2 million euros in support, typically requiring 40–50% co-financing.

Since 2015, STACC has completed 27 projects funded by EIS under the innovation grant, development grant, or programme for applied research, delivering R&D services totalling 4.83 million euros. We have supported clients both in applying for funding and in implementing the projects. In the application phase, we help companies draft the development plan and describe the required technology level and methodology – provided the company already has a defined business problem or a need for a technological solution.

What is the innovation grant?

The innovation grant is designed for small and medium-sized enterprises that want to purchase research and development services to begin developing a new product or service idea.

For example, the grant can be used to conduct a feasibility study, carry out a technological analysis, map and assess existing solutions, perform a preliminary analysis of workflows or datasets, build a data-driven prototype, or test a selected technology in a pilot project. The support allows companies to test early ideas safely and evaluate whether they have potential for further development.

This is especially helpful when a company has an innovative idea but lacks the resources or technical capacity to move forward independently. For instance, you may want to evaluate the feasibility of a new AI solution, analyse your existing data, or test whether a specific technology is suitable before investing in a full-scale development project.

To be eligible for the innovation grant, the company must have a novel idea that falls within one of Estonia’s national RDIE focus areas, and the cooperation partner must be a service provider approved by EIS – for example, a positively evaluated R&D organisation such as STACC.

At this stage, STACC can support you with planning the research and development activities, evaluating suitable technological approaches, and, if needed, building a data-driven prototype or test model.

Full details about the grant’s terms and eligible activities are available on the Enterprise Estonia website.

What is the development grant?

If the idea has already been validated and the direction is clear, but your team lacks the technical skills or resources to move forward, the development grant may be the right choice. This grant supports the next phase of technology development – for example, building a working prototype, testing components, or conducting more advanced product testing.

It is important that the development idea is already sufficiently mature when applying for the development grant. While innovation grant projects generally remain at technology readiness level* (TRL) 2–4, from defining the idea to creating an initial prototype, the development grant enables progression to TRL 4–6, where the technology is tested in a relevant environment and a functional prototype is developed.

* Technology readiness level (TRL) – a 1–9 scale used to assess how mature a technological solution is: from idea and feasibility evaluation (TRL 1–3) to prototype testing (TRL 4–6) and finally to real-world implementation (TRL 7–9).

For example, you have built a basic prototype but need a partner who can help improve its reliability, scalability, or accuracy. Or you’re planning to develop a data-driven solution that needs to be validated before going to market. Perhaps you already have a working solution, but it requires certification, technical refinement, or the integration of a new component – areas where your team needs additional expertise.

STACC can be a valuable development partner in this phase, who helps to:

• define technical goals and tasks so they are well-justified and clearly understood by grant evaluators;
• draft a realistic and technically sound development plan that meets both your company’s goals and the grant’s expectations;
• assess whether your existing data and technical architecture are suitable for further development;
• develop and validate prototypes, including machine learning models and other data-based components;
• test alternative technological approaches to find the best way to solve your business problem.

It is important to note that STACC does not begin development work before the grant is approved. Our role is to help plan what could be developed, not to deliver results during the application phase.

To apply for the development grant, your company should have:

• a clearly defined business problem whose solution would offer a competitive advantage;
• an idea that falls within one of the RDIE focus areas;
• a reliable service provider as a partner – such as STACC, a positively evaluated R&D institution with experience in executing projects funded by the development grant.

If your project involves data-driven methods, AI components, or analytical capabilities, it’s best to contact STACC already during the planning stage. This allows us to help you refine your idea, define the necessary steps, and prepare you for a successful application.

Details about eligible activities and requirements can be found on the Enterprise Estonia website.

What is the programme for applied research?

The programme for applied research is intended for development projects that aim to create new technologies, methods or solutions and require substantial research and development work to bring them to life. This is not just the next step after prototyping – it’s meant for situations where existing knowledge is not enough and a new, research-based approach is needed.

Applied research focuses on solving specific practical problems through scientific methods. While basic research aims to generate new knowledge in a general sense, applied research seeks to expand or refine existing knowledge in order to develop new methodologies or technological solutions for achieving a defined objective. This may involve combining known approaches or exploring entirely new ones when existing methods fall short.

Applied research projects often also involve experimental development – for example, building and testing a new data model and assessing its impact.

The programme is a good fit for companies that:

• have a large dataset and/or a complex technical challenge that requires testing and developing new scientific or methodological approaches;
• are ready to commit to long-term development (projects typically last 1–3 years);
• are interested in creating a technological edge and have a clear vision of the business impact they want to achieve.

STACC is the most valuable partner at this stage due to our experience in leading research-based development and our ability to take projects to a technological breakthrough that surpasses existing solutions.

The applied research projects are successful when the company can clearly articulate the problem and goal that data and technology should help solve. STACC can help define this as a technical research question, evaluate the scientific novelty, develop the methodology, and design the development process.

To apply for the programme, companies must first go through a mandatory pre-consultation. The purpose of this is to help them put together a strong and realistic project plan. This is where STACC’s strategic role becomes especially important – we help think through key elements before the consultation, such as:

• whether suitable data is available to solve the problem;
• which methods could be used;
• how to measure impact and results:
• what would be an optimal project timeline and division of tasks.

In the case of the applied research programme, the grant amount can reach up to 2 million euros, but it is also possible to apply with a smaller budget – for example, as a small project with funding between 100,000 and 250,000 euros. In both cases, it is essential that the company clearly understands the scope of the project and is ready to prepare it thoroughly.

Overview of support measures

* The funding rate under the programme for applied research depends on the size of the company, the type of project (applied research, experimental development, or feasibility study), and the form of cooperation. Detailed funding rates are available on the Enterprise Estonia website.

STACC’s role in successful project planning

To ensure a successful project, cooperation with STACC should start at the right time and under the right conditions.

We add the greatest value when a company has:

• a clearly defined business problem or development need that requires a technological solution;
• a basic understanding of what needs to be achieved and in which areas cooperation is needed;
• willingness to dedicate time to planning and discussions to reach a well-thought-out solution.

It is also important that there is someone in the company with a technical overview or the ability to engage in meaningful collaboration – this accelerates the process and increases chances of success.

Let’s talk about your idea

If you have a data-driven development idea that could qualify for one of the grants above, feel free to contact us. We help you assess which phase your project belongs to, whether it qualifies for support, and how STACC could add real value as your development partner.

The post Which Enterprise Estonia grant is right for your idea? appeared first on STACC.

Many products and services fall short not because they’re unneeded, but because the user experience isn’t smooth or engaging enough. Analyzing user behavior helps reveal where a product excels – and where it needs improvement.

A great example of this approach is a collaboration on the Multiplication Game. STACC analyzed user data to provide the development team with insights for applying targeted enhancements. This project shows that even modest data volumes can uncover patterns that significantly improve user experience and product value – not just in education, but across industries.

The Multiplication Game – a fun way to learn multiplication

Math can be challenging for kids – especially memorizing multiplication tables, which often feels repetitive and boring. The Multiplication Game (Korrutusmäng) transforms this chore into an engaging and playful experience. First, the game’s designers tested a paper prototype with over 2,000 learners. The mobile app now allows kids to independently practice multiplication.

The game includes seven levels, each designed with a specific learning goal:

Level 1: Meet ten characters representing numbers.

Level 2: Listen to fun stories linking the math operations.

Level 3: Reinforce numbers with added letter cues.

Level 4: Teach how to trace numbers on operation cards.

Level 5: Deepen skills in “code language” reading.

Level 6: Understand multiplication through simple problems.

Level 7: Integrate all concepts and master multiplication proficiency.

Challenge: why didn’t kids finish the game?

The developers wondered why many kids didn’t complete the game. To answer that, STACC analyzed user activity logs to identify specific points where learners got stuck or dropped off. Using clustering techniques, we identified groups of users with similar learning patterns and uncovered clear bottlenecks:

• Certain levels were more difficult than others, causing 30% of users to drop out by level 2.
• Children with reading difficulties needed more support to continue.
• More successful learners often revisited earlier levels – for example, reviewing content from level 1.

Applying results in further development

The insights from our analysis guided the team to enhance the game’s learning support and engagement. Next steps include:

• Streamlining and improving the flow of specific levels to make them smoother and more engaging.
• Redesigning exercises to support kids with reading challenges.
• Adding personalized review exercises focused on concepts where each child previously struggled, reinforcing long-term retention.

From data to better solutions

Data analysis is a powerful tool for anyone looking to enhance the user-friendliness of their products or services. In education, it helps understand learners’ needs and improve learning tools – but the same approach works just as well in marketing or software development.

By recognizing where a product truly delivers – and where it doesn’t – you can make data-informed decisions and build more user-centered solutions.

Want to understand what your users really experience? Let us help turn your data into a clear plan of action. STACC can show you how to translate user behavior insights into improved experiences and greater value.

The post How User Data Creates Better Products appeared first on STACC.

As the share of renewable energy continues to grow, the need for large-scale energy storage solutions becomes increasingly critical. Ensuring energy security and maximizing profitability requires careful, data-driven planning of investments in energy production, storage, and consumption. Long-term decisions demand robust calculations – presented in a way that enables all stakeholders to make swift and informed choices.

Providing clients with clear insights into system profitability

Energiapartner, a seasoned designer and installer of solar parks and battery storage systems, aimed to be a strategic ally for clients navigating the energy transition. To enhance their sales process, Energiapartner partnered with STACC to develop a tool that enables clients to quickly and clearly understand the profitability of proposed energy solutions – factoring in equipment, financing, and market participation.

“Investing in storage solutions is a significant business decision. We wanted to provide our clients with prompt and comprehensible information about the profitability of development projects. Together with STACC, we developed a tool for our sales team that accounts for the specific characteristics of a client’s project – both in terms of equipment and financing details – and helps determine the optimal solution for the client.”
— Mikk Saar, Member of the Management Board at Energiapartner

A powerful tool for the sales team

The tool developed for Energiapartner builds upon STACC’s existing energy system profitability calculator. It calculates the investment payback period based on the client’s energy consumption and the parameters of various energy system components (such as solar parks, EV chargers, and batteries). The logic was further customized to include elements critical for Energiapartner – such as loan-based financing, participation in multiple electricity markets, and the separation of fixed and variable costs.

The tool enables the sales team to:

• use up to 60 input parameters, from system-specific metrics to aggregator fees, including the ability to import hourly resolution data for energy consumption and solar production;
• quickly adjust inputs, such as financing terms or market strategies (e.g., day-ahead vs. frequency markets);
• save and reuse configurations, allowing the team to easily generate alternative proposals for the same client.

For client-facing outputs, the tool generates:

• a summary of calculation results with intuitive visualizations that clarify the investment’s value;
• detailed tables and formatted reports that can be downloaded and shared with business partners or financial institutions.

The end result is a user-friendly tool that transforms complex calculations into clear business value – for both the sales team and the client.

Collaborative development and STACC’s role

The solution was delivered as a full-service project: the client defined their needs and requirements, while STACC handled the methodology, technical development, hosting, and ongoing support. Development followed an iterative process, allowing the team to continuously test and fine-tune new features based on feedback.

“Several new capabilities had to be added to the existing profitability calculator based on Energiapartner’s requirements: participation in mFRR, aFRR, and FCR frequency reserve markets; detailed PDF report generation; loan- and grant-based financing models; and additional market and grid parameters. This gives the sales team, on one hand, the ability to simulate real-life scenarios with a high level of accuracy, and on the other, makes it easy to explore different what-if cases. Energiapartner’s domain knowledge also significantly enhanced STACC’s development team’s expertise in the energy sector,” commented STACC data engineer Karl Kevin Ruul.

Reduced manual work, increased client focus

The tool’s greatest impact for the sales team was not just time savings, but also the ability to respond quickly to client needs. Proposals that once required manual recalculations can now be regenerated in minutes – shifting daily work away from repetitive tasks and toward strategic collaboration.

“Sales work is intense and time-sensitive. Many proposals are in progress at once, each with multiple variations, and calculating them manually used to take hours. If a client requested changes, we had to start over. Now, we can use saved inputs to generate a fresh, updated report in no time – ready for discussions with partners and banks. I can now focus more on designing the best solution for the client, rather than spending time on repetitive calculations.”
— Karl Miilits, Sales Engineer at Energiapartner

Interested in a similar solution?

Every project is unique – and deserves a tailored tool. If you’re looking for a data-driven solution that fits your business needs, simplifies complex decisions, and helps communicate value clearly to your clients, get in touch. Let’s explore how STACC can support your journey.

The post Data-Driven Approach Adds Value to the Sales Process appeared first on STACC.

Artificial intelligence projects are becoming an essential part of digital transformation, but ensuring their success requires more than just cutting-edge technology. At STACC, we’ve seen firsthand that successful AI implementations depend on clear goals, strong collaboration, and a structured approach.

The building blocks of a successful AI project

A well-executed AI project starts with a solid foundation. Organizations should consider the following critical factors:

• Clear business objectives: AI should serve a defined purpose, solving a real problem rather than being implemented for the sake of innovation.
• Alignment between business and technical teams: Miscommunication between decision-makers and developers can lead to misaligned expectations and wasted resources.
• Defined responsibilities: Both the client and the development team must understand their roles in the project.
• Budget control: AI projects should be managed with a structured financial plan to avoid overspending.

When to involve AI and data professionals?

In some cases, businesses may need to involve AI and data professionals early in the process, especially when facing unique challenges such as:

• Non-standard solutions requiring custom AI models
• Special data protection requirements
• High licensing costs for commercial AI tools
• Long-term AI and data strategy planning
• A need to move quickly and systematically

To successfully launch an AI project, organizations need a structured approach that balances business objectives, collaboration, and financial planning. The following workflow outlines the key steps in AI development, from defining the problem to delivering a scalable solution.

Client and developer responsibilities

AI project success is not just the responsibility of developers – it requires active engagement from clients as well.

Client responsibilities:

• Foster enthusiasm and readiness for AI adoption within the organization.
• Appoint a product owner with a clear vision and decision-making authority.
• Remove roadblocks that could hinder development, such as data access issues.

Developer responsibilities:

• Understand the client’s business needs and translate them into technical solutions.
• Optimize resource allocation to maximize efficiency.
• Deliver a high-quality system that meets the agreed requirements.

Early stages: Managing uncertainty in AI projects

AI development often involves a level of uncertainty, especially in the early stages. Pilot projects are an effective way to mitigate risk. Instead of aiming for large-scale deployment immediately, organizations should start with small, focused initiatives to validate concepts before scaling up.

AI projects often start with a high level of uncertainty. Using structured methodologies helps organizations navigate these challenges while ensuring continuous improvement. The machine learning canvas below provides a framework for breaking down complex AI initiatives into manageable steps.

Financial considerations: Fixed price vs. time and material

One of the key challenges in AI development is choosing the right pricing model. Fixed-price contracts can force vendors to cut corners, potentially compromising quality. Instead, a time-and-material approach, with prioritization and phased deliveries, allows for greater flexibility and ensures a focus on quality rather than just cost.

The inevitable role of AI in the future

AI is not just a passing trend – it is a fundamental shift that businesses need to embrace. The key takeaway for organizations is to start thinking about AI now, even if they are not ready for full implementation. By taking small steps, aligning teams, and fostering a culture of innovation, businesses can position themselves for long-term success.

At STACC, we help organizations navigate the complexities of AI adoption, ensuring that projects are built on strong foundations with clear objectives and measurable outcomes. If you’re considering leveraging AI in your business, let’s talk about how we can make it work for you.

The post How to Ensure the Success of an AI Project appeared first on STACC.

The electricity market has developed rapidly in recent years, creating new opportunities but also bringing complex challenges. Market participants must manage vast amounts of data daily – from real-time prices and consumption forecasts to production volumes. In such conditions, manual data processing is both time-consuming and prone to errors, directly impacting a company’s efficiency and profitability.

Automated solutions help alleviate these challenges by reducing the risk of human error and making data-driven decisions faster and more precise. To capitalize on these opportunities, STACC and eTerminal partnered to optimize energy trading processes, supporting the company’s rapidly growing customer portfolio.

Siim Meeliste, Head of Electricity Trading at eTerminal, emphasized that efficient and accurate data processing solutions provide a significant competitive advantage in the electricity market. “Before starting our collaboration with STACC, we conducted a thorough review of existing solutions. However, we decided to build from scratch with a strong partner who had prior experience in energy sector data processing and could also contribute insights regarding co-financing from support programs.”

A solution for automated electricity trading

eTerminal needed a solution that would allow them to:

• Accurately predict electricity consumption and production within their portfolio,
• Use weather data and historical electricity prices for forecasting,
• Automate the creation and submission of purchase orders to Nord Pool,
• Send daily balance reports to Elering based on forecast results and executed transactions on Nord Pool,
• Gain an overview of and intervene in automated processes when necessary.

In collaboration with STACC, a solution was developed consisting of several key components:

1. Data flow integration: To accelerate data processing, information from various sources was standardized, eliminating reliance on manual work.
2. Consumption and production forecasts: Machine learning algorithms were used to develop initial consumption and production forecasting models, which can be refined over time.
3. Automation of trading and balance planning: Instead of a manual process, the system now generates and submits Nord Pool purchase orders and Elering balance plans autonomously.

The daily process begins with analyzing new input data, resulting in a forecast for the next day’s electricity production and consumption. Based on this, the necessary electricity volumes for purchase from Nord Pool’s day-ahead market are calculated. Once the auction results are available, a balance plan is generated and submitted to Elering.

The primary goal of the system is to free up traders’ time – rather than handling each step manually, traders now take on a supervisory role and intervene in critical situations. Manual intervention may be necessary in case of technical failures or unexpected changes in market conditions that the model has not yet learned to handle. The automated system significantly reduces the need for manual work, improving both the accuracy and efficiency of processes.

The solution was developed in cooperation with Positium, whose representative, Marko Peterson, stated: “The uniqueness of this project lay in the fact that previous software solutions could not be directly reused. The biggest challenge was to create a minimum viable product (MVP) where balancing quality and development speed required continuous attention. We had many discussions on how to strike the right balance between rapid development and delivering a high-quality final product. Additionally, testing new approaches and experimenting with different methods played a crucial role in this project, making it a challenge for all participants while laying a strong foundation for future developments.”

Future prospects and the transition to 15-minute data intervals

Following the successful deployment of the initial solution, collaboration with eTerminal continues to support the transition to the 15-minute data exchange standard and a 15-minute interval day-ahead market. The new 15-minute standard will be implemented in data exchange as of March 1, 2025.

Under the current standard, the electricity market operates with 24 data intervals per day:

1. 00:00–00:59
2. 01:00–01:59
…
24. 23:00–23:59

With the new system, there will be 96 data intervals per day:

1. 00:00–00:14
2. 00:15–00:29
…
96. 23:45–23:59

This transition will make data storage and transmission four times more granular. While this increases data volumes, it also presents a significant opportunity to improve electricity grid balancing. Elering will implement the new standard in early 2025, with Nord Pool planning to follow in the near future.

For electricity traders, this change will allow the use of more detailed measurement data to enhance forecasting accuracy, thereby reducing both trading costs and end-user electricity prices. However, the transition also comes with additional workloads and challenges. All automated systems communicating with Elering or Nord Pool must be adapted, and electricity consumption and production forecasting models must be updated – steps in which STACC is supporting eTerminal.

Optimizing electricity trading with data science

STACC’s data scientist Kaspar Valk emphasizes the complexity and importance of the project: “It’s highly motivating to develop a solution where the benefits for the client are clearly measurable and provide substantial daily time savings. Collaboration with eTerminal has been excellent, thanks to Siim Meeliste’s professional approach, strong technical understanding, and goal-oriented mindset. From a data science perspective, forecasting day-ahead production and consumption volumes for eTerminal’s portfolio is both fascinating and challenging. One complexity is the dynamic nature of electricity traders’ customer portfolios – each day, it can grow or shrink, requiring models to adapt continuously. Another challenge lies in the diverse consumption patterns of different electricity users, ranging from households to industrial facilities. The key question is whether to use a single model for the entire portfolio or develop specialized models for different customer segments.”

“The time savings from automating these systems can be measured in entire work shifts.”
— Siim Meeliste, Head of Electricity Trading, eTerminal

If your company is facing complex energy management challenges or growing data volumes, it’s time to consider automation. Contact us to explore how STACC can help optimize workflows, reduce errors, and increase efficiency.

Cover photo: eTerminal

As part of this collaboration, machine learning models developed under subproject 1.14, “Data Analytics for Electric Energy Management,” within project EU48684, were validated.

The post The Rapid Growth of eTerminal Drives The Need for Automation appeared first on STACC.

High-quality data is essential for successfully applying machine learning or AI and building reliable, efficient systems. In retail, where quick decision-making and precision directly impact a company’s profitability, data quality is critical. However, data is rarely perfectly structured, consistent, or error-free. The success of machine learning and AI projects heavily depends on thorough data cleaning.

Data cleaning involves many details. Simple tasks include correcting typos, filling in missing values, handling duplicate rows, and assigning appropriate data types. More complex examples include aggregating data into consistent time intervals, such as summarizing events by the hour. In retail, these tasks come with additional industry-specific challenges, which STACC has extensive experience addressing.

Integrating data from multiple sources

Retail businesses often need to consolidate data from multiple sources, a process that is both complex and time-consuming.

For instance, data may come from different store units or subsidiaries using various POS systems. Pricing and customer data differ significantly between e-commerce, retail, and wholesale. For companies operating internationally, factors such as currencies, regulations, and market conditions further complicate data collection. These diverse sources must be harmonized to ensure effective use within a unified system.

Cleaning product data

Product data cleaning is often critical in retail and presents unique challenges across different data types.

Retail processes rely on a variety of product identifiers, such as barcodes, SKUs, or database IDs. These identifiers may vary across product sizes and colors. When building systems like recommendation engines, it’s essential to decide whether to analyze data at the model level or account for variations across different product versions.

If accurate inventory data is required for business objectives, data processing must align with inventory updates and reconcile data from multiple sources. Older product information structures may not meet modern standards. Additionally, outdated products and “phantom items” like plastic bags or deposit packages need to be removed to provide accurate data for further analysis.

Cleaning customer data

Customer data cleaning often revolves around privacy regulations and unique identification challenges.

It’s crucial to determine which personal data can be processed directly, which needs to be anonymized, and how it should be stored to comply with data protection laws.

Like product data, customer data may encounter identification issues. Different systems often use various identifiers such as email addresses, loyalty card numbers, personal identification codes, or database IDs. A single person may be registered multiple times with different contact details, and it’s common for family members to share a loyalty card. Additionally, legacy customer information may not align with current standards, complicating processing.

Cleaning transaction data

Efficient transaction data processing is vital in retail, as it directly affects analysis outcomes and business decisions.

Depending on the objective, transaction data may need to be transformed to allow calculations at the level of individual purchase lines or entire shopping carts. It’s also essential to account for cross-cart discounts and decide whether to use unit prices or total prices for purchased quantities.

Automation of transaction data processing requires consideration of data availability and latency. For example, transaction data might update in real-time or at midnight, with delays ranging from immediate availability to several days.

Additional data types may also need to be processed. For instance, e-commerce activity data, such as product page visits, might need to be merged with other customer data. Similarly, distinguishing between local and nationwide campaigns can be an essential step in cleaning and processing data.

How can STACC help?

Over the years, STACC has helped numerous retail businesses optimize their data cleaning processes and solve complex, industry-specific challenges. Our solutions not only organize data but also establish a solid foundation for building accurate and reliable AI systems.

If you want to assess whether your company’s data is ready for AI implementation or need guidance on how to get there, reach out to us. We’ll evaluate your data’s condition and create a plan to bring you closer to data-driven innovation.

The post Data Cleaning in Retail: The Foundation of AI Implementation appeared first on STACC.

Policymaking requires access to high-quality, relevant information. But how can critical data be identified amidst an ever-growing volume of content? This challenge prompted the Government Office of Estonia to commission STACC and TEXTA to develop a solution. The result: a prototype of a semantic search tool that enables civil servants to find information faster, easier, and with greater accuracy.

Decision-making requires a clear data overview

Each year, thousands of Estonian-language studies, reports, laws, and other documents are created. The vast volume and scattered nature of these valuable resources make finding relevant information time-consuming and increase the risk of overlooking critical insights during the policymaking process. Given that policies impact society as a whole, improving decision-making workflows has far-reaching benefits.

To address this challenge, the Government Office commissioned a prototype for a semantic text search application. The idea is straightforward: an advanced language model analyzes user queries, retrieves relevant sources, and compiles clear responses. For example, during the evaluation of a draft law, the tool can quickly provide an overview of pertinent public studies. The initial target group includes all those involved in preparing policy decisions.

GPT-4o was the optimal choice

Which language model best meets user needs? We evaluated eight different models to identify the optimal balance of cost, speed, and quality. Based on user scenarios, the assessment included summary generation and the ability of large language models (LLMs) to answer factual questions. Using Estonian Public Broadcasting articles as a test dataset, we established evaluation criteria and ranked the models. The results pointed to GPT-4o as the recommended choice for the application.

We conducted an in-depth analysis of the client’s use cases, assessing feasibility and determining user needs and workflows. The most suitable solution proved to be retrieval-augmented generation (RAG), where the system retrieves relevant text segments from documents based on user queries and passes them to the language model to generate responses. This approach mitigates common LLM issues, such as “hallucinations,” where the model might provide inaccurate information.

Making search smarter

Developing the prototype involved extensive data processing to ensure the search tool is both precise and efficient. Using references provided by the client, we compiled the necessary datasets from the internet and standardized them into a uniform format. Before integrating the data into the database, we segmented and vectorized the collected files. This preprocessing step was crucial, as the input length for language models is limited and directly affects operational costs.

Next, we developed the semantic text search prototype along with its software and user interface. Users can specify time periods and sources when submitting queries. The tool provides two types of responses: one based on the model’s internal knowledge and another derived from the analyzed sources. For transparency, the tool displays citations of the sources used in generating the answers.

A fully operational prototype

This semantic search application stands out due to its extensive Estonian-language knowledge base, long-term testing period, and broad potential for future use. The prototype is currently being used by a test group, and we are collecting usage statistics and feedback to further develop and refine the application.

Future project phases aim to map public sector text data comprehensively and create a robust semantic text search application tailored to the Estonian language and built on local language technology tools. The final application would broadly support public sector officials, facilitating the automation of preparatory tasks for decision-making across government agencies and local authorities.

“The prototype was developed through a procurement process in collaboration with TEXTA and STACC, who demonstrated a clear understanding of our needs, incorporated them into the design, and remained focused on achieving the stated goals. The result is a well-functioning prototype, carefully aligned with defined limitations. We plan to develop it further in the near future to create a solution that enhances policymaking efficiency. TEXTA and STACC were undeniably competent partners, and the simplicity of the user interface received very positive feedback from the prototype testers.”
— Erik Ernits, project manager of the Government Office of Estonia

The activities and projects of the Government Office Innovation Fund are funded by the European Union Cohesion Policy 2021-2027 measure “Enhancing Public Sector Innovation Capacity”.

Sources:
[1] Retrieval-Augmented Generation (RAG), pvml.com
Cover photo: Ametniku tööruum, ERA.5637.0.501762, National Archives Photo Database

The post Smart Search Supports Policymakers’ Work appeared first on STACC.

How can a retailer identify the products that keep customers coming back? In collaboration with Coop Eesti, STACC developed a data-driven solution that offers customers personalized product discounts based on their preferences. Leveraging machine learning, we designed an innovative model that personalizes the shopping experience and strengthens customer loyalty. This solution is unique in Estonia, as the personalized discounts are not displayed on price tags but tailored specifically for each customer.

Opportunities for data-driven solutions at Coop Eesti

Coop Eesti, a leader in the Estonian retail market, has over 120 years of history. Holding nearly 24% of the market share, the company boasts a net revenue of €863 million and operates 320 stores across the country. Over 600,000 loyalty cardholders – roughly equivalent to the number of working people in Estonia – frequent Coop Eesti stores regularly, alongside hundreds of thousands of other customers without loyalty cards. This massive pool of data offers unparalleled opportunities for creating data-driven solutions and delivering a more personalized shopping experience – a vision that Coop has successfully realized with STACC’s help.

Personalized pricing program: what is it and how does it work?

Coop’s personalized pricing program is an innovative customer-centric solution that allows shoppers to benefit from discounts tailored to their individual preferences and buying habits. This approach goes beyond traditional promotional offers, where all customers receive the same discounts regardless of their shopping behavior or needs. Coop’s pilot project in Estonia demonstrates how personalization can make the shopping experience more flexible and engaging.

The system, powered by artificial intelligence, analyzes customers’ past purchasing behavior to select six products that align with their preferences. These can be confirmed by the customer as permanent personalized offers. If a product becomes less relevant, customers can easily update their selection in the client portal or app. Additionally, the algorithm suggests three more products from Coop’s range that the customer might enjoy trying, ensuring even the most discerning shoppers find discounts on their favorite items.

To keep offers relevant, personalized discounts are updated monthly, reflecting changes in customer habits. The customer’s confirmed favorite products remain unchanged, with discounts applied as long as the product is in stock or until the customer decides to make a switch. Similarly, the list of alternative recommendations is refreshed regularly to ensure flexibility and choice. This dynamic process keeps the customer at the heart of the shopping experience.

Here’s an example of personalized offers that have been confirmed as permanent discounts for a Coop loyalty cardholder who works at STACC:

It’s a colleague who enjoys quick and easy meal preparation and is pleased that the selected products make their typical grocery purchases more affordable.

Innovation that bridges customers and technology

While other retailers also offer personalized recommendations among promotional products, Coop takes personalization to the next level by making customers active participants in shaping their shopping experience.

What makes this solution unique?

• Individualized discounts: Personalized offers are available exclusively to loyalty cardholders who meet the conditions for receiving them, ensuring discounts are tailored to their preferences.
• Machine learning and customer collaboration: The machine learning model analyzes customer data to generate new recommendations every month. However, automation isn’t the only focus – customers can actively shape their offers by confirming or adjusting recommendations via the Coop app or portal. This creates a dynamic, personalized interaction between the model and the customer.
• Continuous and evolving process: Once customers confirm their selected products, discounts remain valid until the product is unavailable or the customer chooses to replace it. If no selection is made, the system suggests new products each month, offering fresh and relevant alternatives.

The principles behind the machine learning model

Coop’s personalized pricing model is based on customers’ purchase history. It segments this data to identify both stable purchasing habits and recent behavioral changes. Seasonal trends are also factored in, ensuring, for instance, that winter months bring discounts on items like blood sausages and mulled wine. Additional customer details, such as age and gender, further refine recommendations, making them even more accurate and relevant.

The model doesn’t rely solely on an individual’s data but also considers purchasing patterns of similar customers. For example, if two customers frequently buy the same products, the system can recommend items that one has enjoyed but the other hasn’t yet discovered. This opens the door to new favorites that customers may not have tried otherwise. The model continually improves itself, learning from past recommendations to deliver more precise and relevant offers. If a previous suggestion failed to resonate, the system ensures it doesn’t repeat the same mistake.

Rather than a big “black box” where data is input, and results are output, this solution is a complex combination of several machine learning models working together at different stages. The philosophy behind its creation ensures that each step compensates for any gaps in the previous ones, delivering optimal results for every customer. Feedback – analyzed through purchasing behavior – shows that the recommendations resonate with customers while also generating economic benefits for Coop.

Impact on the customer experience

Personalized pricing reflects a broader retail trend of prioritizing individual customer needs and preferences. This approach helps retailers increase customer loyalty by adding value and showing they truly understand their shoppers. It also allows customers to optimize their spending on items they care about most, providing reassurance about future discounts – something traditional promotions often lack.

Coop’s example demonstrates that personalized pricing isn’t just a future trend; it’s a current reality that makes shopping smarter and more customer-friendly. The solution developed in collaboration with STACC highlights how data-driven innovation can create a truly personalized shopping experience.

If you’re interested in implementing data-driven solutions for your business, contact us to discuss how we can help you build innovative solutions that give your company a competitive edge.

The post Data-Driven Shopping Experience: How Coop Eesti Increases Customer Loyalty? appeared first on STACC.

Weather insights are crucial for solar energy forecasting

But how do we predict something as dynamic as solar energy? It’s a mix of understanding the weather, knowing our solar panels, and recognizing the patterns in energy production. At STACC, we tap into The Global Forecast System (GFS) for weather insights. We consider factors like air temperature, cloud cover, and radiation flux. Technical details matter, too – like a solar park’s capacity, location, and panel angles. However, historical data can fill the gaps, creating a reliable time-traveling dataset.

Harnessing the power of Random Forest for solar energy forecasting

Our secret sauce? A practical and efficient model that thinks fast. We chose a decision tree-based approach called a random forest. It’s quick and adept at understanding daily and yearly cycles. We transformed time into a cycling representation to catch the rhythm of the day and year. And since we had data from various solar parks, we made our model versatile by training it on percentages of maximum capacity rather than raw energy values.

K-Nearest Neighbor as plan B if something goes wrong

But what if weather forecasts decide to play hide-and-seek? No worries, we have a Plan B. We switch seamlessly to our backup model, like having a reliable friend. Using historical data and the k-nearest neighbours (KNN) algorithm, we figure out what similar days in the past were like. It’s our safety net for unpredictable situations.

Solar energy prediction accuracy

Now, onto the all-important question – how good is our crystal ball? We measure it with mean percentage absolute error (MAPE). But here’s the catch: night hours bring many zeros, making calculations tricky. So, we use the weighted mean percentage absolute error (wMAPE) as our primary guide. Tested rigorously, our model consistently scores an impressive 22-25% on a separate dataset. For comparison reasons, Elering, a prominent player, has a wMAPE of approximately 30-31% when predicting solar energy production.

In a nutshell, our model’s accuracy and speed, coupled with a reliable backup, mark a significant step towards a cleaner energy future. By blending weather know-how with machine learning, STACC not only forecasts solar energy accurately but also has a safety net for those times when forecasts decide to play hard to get. It’s a journey towards a brighter, greener horizon for us all. 

If the weather also makes your business challenging to predict, then it’s the right time to contact us!

The post Solar Energy Predictions: Illuminating Tomorrow’s Energy Landscape appeared first on STACC.

Enefit, in collaboration with STACC OÜ, has developed an innovative mFRR frequency market forecasting model aimed at ensuring the stability of the electricity network as the number of electric vehicles and chargers continues to increase.

Optimising consumption and potential risks

Many electric car owners prefer to charge their cars at a time when the price of electricity on the exchange is most affordable. However, scheduling charging to align with low exchange prices could, in the long term, impact the balance of the electricity network, as energy consumption is shifted to a different period than initially planned.

Kätlin Värno, Development Manager of Enefit and the leader of the project, explained that electricity sellers make frequency market offers to network operators to ensure the stability and reliability of the network. ‘If electricity sellers cannot accurately forecast consumption, their imbalance costs increase. Consequently, the electricity price may rise, as companies try to cover the costs resulting from forecasting errors,’ she said.

However, Värno confirmed that such a problem does not currently exist in Estonia. ‘Currently, there are approximately 7,500 electric vehicles on our roads. Although the number of electric vehicles and home chargers is estimated to reach 200,000 over the next ten years, there is no significant risk to the stability of the electricity network at least in the near future. In other words, even if all electric vehicle owners decided to charge their vehicles during the cheapest price hour, it would not critically impact the balance of the network.’

Solutions for managing future consumption patterns

In order to ensure the stability of the electricity network in the future, one solution is to control the consumption of electric car chargers. This means that electric vehicle charging would not occur randomly or be controlled manually by the owner; instead, chargers would be connected to a system capable of optimising consumption and charging schedules, taking into account both exchange prices and participation in frequency markets.

‘This way, charging times and consumption volumes can be shifted, avoiding peak loads and reducing imbalances in the electricity network. At the same time, it is possible to ensure that the car has reached the desired state of charge by the specified time, and this at an even more favourable charging price,’ said Värno.

Such solutions will become increasingly available in the future, as the minimum threshold of 1 MW required for trading on frequency markets is already almost achievable with home chargers. If an electric vehicle owner decides to purchase a charger, Värno believes that it would be reasonable to consider a comprehensive solution where charging sessions are optimised not only for the exchange price, but also taking into account the frequency markets. For Enefit Volt customers, this means even better opportunities to reduce charging costs with the already familiar charging service.

More accurate imbalance forecast and a more stable electricity market

The model created by Enefit and STACC continuously analyses electricity consumption for the next four hours, helping to predict imbalances even more accurately in the future. The more precise the forecast, the better electricity producers will be able to schedule when to offer reserve capacity to the market.

‘The ability to direct production to periods when less electricity is produced than expected helps smooth out price fluctuations and, in the bigger picture, creates a more stable and predictable electricity market,’ said Värno.

‘From the future perspective, and especially with the continued growth in the number of electric vehicles and home chargers, the need for smart and efficient solutions will only increase. The model developed as part of the EVFlex or e-mobility flexibility project, co-financed by the Estonian Business and Innovation Agency’s programme for applied research, is precisely one of these solutions,’ she added in conclusion.

This post was published with permission from Eesti Energia. Original post: Energiatarkuse blog

The post Enefit and STACC develop a forecasting model for balancing frequency markets appeared first on STACC.