- [Tools](/blog/topics/tools) # What Is a Tax Determination Engine and How Does It Work? [![Jennifer Dunn](https://cdn.sanity.io/images/jldqsjg6/production/f4e992938370233aee8323439f74067030161c1b-1486x1059.png?q=80&fit=max&auto=format&w=3840)](/author/jennifer-dunn) [Jennifer Dunn](/author/jennifer-dunn) 2 October 2026 ![Tax Determination Engine](https://cdn.sanity.io/images/jldqsjg6/production/8a45ef10e2dc197cc6219855b03590881e815e41-2280x1521.png?q=80&fit=max&auto=format&w=3840) - What Is a Tax Determination? - What Is a Tax Determination Engine? - How Do Tax Determination Engines Work? - How Determinations Are Generated Around the World - Why TRAM Is a Different Class of Tax Engine - Which Tax Determination Engine Is Right for You? AI-Native vs. Manual Explore with ChatGPT IconClaude IconPerplexity Icon Every time you sell a product or service, you must determine two things: - Is the transaction taxable? - If so, at what tax rate? A tax determination engine is the system that answers both questions automatically, across thousands of products and jurisdictions, in the time it takes to process a checkout or send an invoice. This article explains what a tax determination is, how a determination engine works step by step, and why the way determinations get generated in the first place is the real difference between tax engines on the market today. Key Takeaways - A tax determination is a ruling on whether a specific product is taxable in a given jurisdiction, and at what rate. - A tax determination engine applies these rulings to real transactions in real time, using product tax codes and customer location. - Tax determination is more difficult when selling online, because you generally must determine if products are taxable at your buyer’s delivery location. - Determinations are traditionally built by human tax research teams reviewing statute law jurisdiction by jurisdiction, a slow and expensive process. - [Sphere's TRAM](/blog/building-tram) is an AI-native research system that generates and updates determinations automatically, with tax experts reviewing the output. - AI-native determination engines offer broader coverage, finer accuracy, and faster response to law changes than manually maintained systems. ## What Is a Tax Determination? A tax determination, or tax calculation, is a ruling on how a specific product, including everything that makes it what it is, gets treated under the tax law of a specific jurisdiction. It answers one narrow question: is this exact product taxable here, and if so, how much tax applies? The same product rarely gets the same answer everywhere. A bottled beverage might be fully taxable in one state, exempt as a grocery item in another, and taxed at a reduced rate in a third. A downloadable software license might count as tangible personal property in one country and as a nontaxable service in the next. The product does not change. The jurisdiction's tax rules do. If you have [sales tax nexus](/blog/nexus-tax-software) in all three states, you have to ensure you’re collecting the right tax on every transaction. If you overcollect, you risk a poor customer experience. If you undercollect, you risk tax liability, which you will owe out of your profits to the relevant taxing authority. Your tax determination software is vital. This is what makes indirect tax so hard to manage at scale. A company selling a handful of products in one state or country can track these rules in a spreadsheet. A company selling hundreds of products across dozens of countries cannot. That is where a tax determination engine comes in. ## What Is a Tax Determination Engine? A tax determination engine is a system that applies tax determinations to transactions automatically, across any product and any jurisdiction, at the moment a sale happens. It takes over the work a human would otherwise have to do transaction by transaction. For every transaction, the engine answers two questions. Is this taxable in the buyer's jurisdiction? If it is, what rate applies? The engine has to get both answers right instantly, because most transactions today happen at checkout, when your customer is ready to hit “buy now.” A fast, accurate tax determination engine is vital for keeping your business running and your customers satisfied that you are charging them the exact rate. ## How Do Tax Determination Engines Work? A tax determination engine works in three steps: assigning tax codes to products, building a determination matrix for those codes, and applying that matrix to live transactions. ### Step 1: Product Tax Codes Every product in a business's catalog gets assigned a tax code, and that code is supposed to capture everything about the product that matters for tax purposes. This includes things like how the product is delivered, whether it is sold to a business or a consumer, and what category it falls into. The quality of this step decides the accuracy of everything downstream. A tax code that is too broad, lumping several different product types under one label, will produce wrong determinations no matter how good the rest of the engine is. This is one of the most common failure points in tax compliance. According to Thomson Reuters, manually mapping large product catalogs to the right tax codes is consistently the slowest and most error-prone part of the process, since different reviewers often classify similar products differently, and a single miscoded product can generate incorrect tax on every invoice tied to it until someone catches the mistake. ### Step 2: The Determination Matrix Once products have tax codes, the engine needs a determination matrix. Think of this as a large table. Each row is a tax code. Each column is a tax jurisdiction. Every cell in the table tells the engine whether that product type is taxable in that jurisdiction, and at what rate. Building this matrix is not a one-time task. Tax law changes constantly. States adjust thresholds, add exemptions, or reclassify entire product categories. A jurisdiction that did not tax digital products two years ago may tax them today. The matrix has to be kept current, or the engine will apply outdated rules to current transactions. ### Step 3: The Transaction When a sale happens, the engine needs two pieces of information for real-time tax calculations: which product tax code applies, and where the customer is located. With those two inputs, it looks up the correct cell in the determination matrix and applies it. If the product is taxable in that jurisdiction, the engine calculates the tax owed at the correct rate and returns it, typically within the same second the transaction is processed. This is the part of the process that involves ERP systems, e-commerce platforms, and CRM tools directly. The engine needs to plug into wherever the transaction actually happens, whether that is a shopping cart, an invoicing system, or a procurement workflow, in order to apply tax logic in real time. ## How Determinations Are Generated Around the World The most complex part of running a tax determination engine is generating and maintaining the underlying determinations in the first place. There are two ways to do this. ### The Traditional Approach Most tax engines on the market today rely on large tax content teams. These teams manually review statute law, administrative guidance, and court precedent for every jurisdiction they cover, then translate their findings into rows and columns in a determination matrix. This approach has its limits. It is slow, because legal research takes time and every new jurisdiction or product category adds more work. It is expensive, because it requires specialized tax researchers. And it is reactive by nature. Law changes get caught after they happen, when someone on the research team notices a new statute or ruling and updates the matrix. In the gap between a law changing and a human catching it, businesses can end up applying the wrong tax without knowing it. ### How Sphere Does It: TRAM Sphere takes a different approach with its [Tax Review and Assessment Model](/blog/building-tram), or TRAM. TRAM is a tax-focused AI system built to ingest and monitor tax law directly from primary sources, then generate and update determination matrices without waiting for a human researcher to work through each jurisdiction one at a time. Tax experts still play a role in this process. They review TRAM's output to confirm accuracy and to feed corrections back into the system, which improves the accuracy of TRAM's first-pass determinations over time. The difference is that instead of a team of researchers building the matrix from scratch, TRAM builds a first draft instantly and experts refine it, rather than starting from a blank page every time a jurisdiction updates its rules. ## Why TRAM Is a Different Class of Tax Engine An AI-native research system changes what a tax determination engine can realistically cover and how quickly it can respond to change. TRAM's advantages fall into three categories. ### Scalable Coverage A manual research team can only cover so much ground. Every new region, every new product category, and every new tax type, whether that is sales and use tax, value added tax (VAT), goods & services tax (GST), [input tax](/blog/input-tax), withholding tax, or [tariffs](/blog/tariff-vs-tax), adds work that has to be done by a person. TRAM scales across all three of these dimensions, regions, product categories, and tax types, without requiring the business to add headcount every time it expands. ### Superior Granularity Manual tax research runs into a practical wall. Reviewing every possible product variation in every jurisdiction would take years, so human teams simplify. They group similar products together and apply broad rules, because going product by product is not feasible at scale. TRAM does not face that constraint. It can process granular product attributes instantly, which means it can produce more precise determinations instead of relying on broad categories that may not fit every product exactly. ### Automated Monitoring Traditional tax teams check for law changes after the fact, often during a periodic review cycle. TRAM is connected to major tax law sources directly and monitors them on an ongoing basis, which means it can catch a rate change or a new exemption closer to when it happens rather than weeks or months later. This matters because every day a determination matrix runs on outdated law is a day a business risks under-collecting or [over-collecting tax](/blog/is-it-illegal-not-to-refund-sales-tax), both of which create real financial and audit risk. ## Which Tax Determination Engine Is Right for You? AI-Native vs. Manual ### Traditional, Research-Based Engines Most established players, including Avalara, Vertex, Anrok, Numeral, and Fonoa, are built around the traditional model. Human tax content teams research and maintain determinations jurisdiction by jurisdiction, and the software layer applies whatever those teams have documented. This approach has been the industry standard for years, and it works. Its limits are the ones inherent to any manually maintained system: coverage expands only as fast as the research team can keep up, and law changes get reflected only after someone catches them. ### Sphere — AI-Native Determination via TRAM ![AI-Native Determination via TRAM](https://cdn.sanity.io/images/jldqsjg6/production/d29c578636158a99347c799b734211e1c2fda43e-3966x2100.png?q=80&fit=max&auto=format&w=3840) Sphere is built differently, from the ground up. Rather than layering software on top of a manual research process, Sphere's core engine is TRAM, an AI system designed specifically to research, generate, and update tax determinations. Tax experts remain in the loop to verify accuracy, but the foundation of the system is AI-native rather than research-team-native. For businesses selling across many product categories and many countries, whether that is physical goods, digital products, or services, this structure is built to keep pace with a tax landscape that changes constantly and unevenly across the world. ![Blog post CTA background](/images/blog-post-cta-bg.avif) Want to see how Sphere's tax determination engine handles your product catalog? [Schedule a demo with Sphere today](/contact) ## You might also enjoy [ ![Building TRAM | The World’s First Tax Review and Assessment Model](https://cdn.sanity.io/images/jldqsjg6/production/a0238f10ffcdbfa5f59a4fcb73138751ebb57270-1269x846.avif?q=80&fit=max&auto=format&w=3840) AI ### Building TRAM | The World’s First Tax Review and Assessment Model Nov 23 2025 ](/blog/building-tram)[ ![AI Tax Preparation for Global Compliance](https://cdn.sanity.io/images/jldqsjg6/production/7a98f55110472519a37da95d518e883b6fb6886a-1224x816.avif?q=80&fit=max&auto=format&w=3840) AI ### AI Tax Preparation for Global Compliance Mar 6 2026 ](/blog/ai-tax-preparation)[ ![AI-Powered ERP Solutions for SaaS Companies](https://cdn.sanity.io/images/jldqsjg6/production/9c29b1ecce8fc4320b9be1c9578eea1308d15c67-1224x816.avif?q=80&fit=max&auto=format&w=3840) AI ### AI-Powered ERP Solutions for SaaS Companies Mar 6 2026 ](/blog/ai-powered-erp) ## It shouldn't be your job to understand 190 tax systems.\*It shouldn't be your job to understand 190 tax systems.\* [Get a demo](/contact)