R&D Expenditure Accounting Model for Tech Startups: A Field Study

At present, many new tech companies are at the forefront of technological change and can change the entire market by using new technologies and rapid changes. As the companies grow and gain market share, they will also spend more money on Research and Development. As a result, there is an increase in the investment of these funds in research and development.

But in the process of financial expansion, there is widespread confusion about how to record R&D expenses in finance. The department must follow the rules of the present accounting standards for external reports and at the same time meet the fine-tuned requirements of internal operation management regarding the real-time performance of the project. Therefore, the financial personnel in the early stage often have difficulty in dividing the research and development stages reasonably, applying the capitalisation conditions correctly under severe uncertainty, and tracking the expenditures of the projects that have not yet been completed. Improper handling of expenses will not only lead to penalties from regulatory authorities but also affect the clarity of internal cost statistics, which may result in incorrect strategic plans and insufficient funds.

Due to the inherent differences between the old, general rules for the accounting of R&D expenditure and the characteristics of a rapidly changing early-stage technology company, it will produce the above-mentioned operational friction. The old accounting systems and regulations were mainly developed for the needs of large, stable companies; they were formed based on the stability of the life cycle of new products and the continuity of development, and thus they had a relatively low threshold for technological feasibility. On the other hand, the tech start-up environment has high-speed iteration, continuous deployment, high failure rate and other characteristics; traditional milestones are not practical. The traditional accounting rules are too rigid and do not fit the flexible, overlapping development work of agile startups. Therefore, if we force these changing operating conditions into a fixed accounting model, it will cause many problems in measurement and other areas, and the financial statements will not accurately reflect the actual economic situation of the company or provide useful information for managers.

Such a large gap between the standardisation of the accounting system and the actual situation in the early stages of technology development will lead to problems in daily financial management. Therefore, it can be said that the reason for this field study is to address the issue of financial management in this particular environment, which is the urgent need for a reasonable accounting system in the current complicated practical situation. The goal of this paper is to study the reasons for the operational problems of the accounting system in these technology start-ups and build a model for the accounting of R&D expenditures in such circumstances, based on regulatory requirements and the actual needs of enterprises.

There are three levels of goals for the first level of this research, namely solving the practical problems of R&D expenditure accounting in technology startups. The first is to collect first-hand data from newly developing technology start-up companies to find out what problems they are facing regarding R&D expenditure accounting. At this time, enterprises often have problems such as unclear divisions between the research stage and development stage, and problems in collecting and distributing indirect expenses. Based on the on-site field investigation of the project, this paper hopes to obtain real-life problems in practice that have not been recorded in standard accounts. The second is to set up a targeted and simple-to-use model for R&D expenditure accounting that can meet the operating characteristics of technology startups. This model should not be an ideal one; instead, it should convert complex professional standards into simple operational procedures and match the characteristics of early-stage enterprises with agile and iterative development. The third is to test whether the model works in the real situation of a firm. The proposed model can be placed in actual business situations to confirm if it works in practice and does not add too much administrative burden to the company.

The main purpose of this paper is to make up for the shortcomings of rigid professional norms in early times and adjust them to new circumstances. From the practical point of view, it is very helpful for the financial departments of high-tech start-ups to learn how to organise the data in their research and development. With a clear way to carry it out and operate, the staff can move from unplanned record-keeping to orderly compliance and reduce the risk of violation of regulations. In addition, a standard accounting model will be used to record the input and output of R&D costs for these companies, so that it can support the creation of follow-up financing and incentive plans. Accurate capitalisation and expense distribution are the first requirements for a good financial report, and they are the basis for the external valuation of a company in an equity financing round and the establishment of a fair equity incentive system for the company's core technical staff. From the wider industry view, this study will enrich the general rules of the industry with specific examples of how to improve the industry-level R&D accounting guidelines for new small and medium-sized enterprises. Most regulatory systems now exist to manage large companies, but they lack sufficient flexibility for new businesses. Based on the field-validated solutions, this study will offer some examples to enrich the policy discourse and help formulate better rules for the industry that better suit the operating speed and resource conditions of technology startups.

The tech start-up in the field study is at an early stage, has high technological uncertainty, has limited resources and does not rely on tangible assets to conduct business; it has been newly established. R&D expenditure is the economic resources that are directly invested in research and development, and it consists of the costs of personnel labour, material consumption and other expenses related to new product design, etc. All these are in a series of new prototype production and market testing to meet one's own needs through small changes; there is no fixed time limit for traditional R&D. Therefore, R&D expenditure in this environment is not just a simple division of research and development but also a dynamic way of accounting; after the iterative output is generated, it will be regarded as an intangible asset and amortised over time rather than be expensed immediately.

There are three basic theories that can be applied in this research. According to accounting relevance theory, financial information should be predictive and confirmatory to aid various stakeholders in their decision-making. The traditional way of paying the cost cannot accurately reflect the benefits of innovative investment for tech start-ups operating in an unstable environment. Based on this theory, the design of the on-site research plan will first investigate whether the current reporting mode can meet the requirements of internal management decision-making, and then establish a benchmark for diagnosing the reasons for insufficient funding of iterative outputs in accounting.

Second, according to the activity-based costing (ABC) theory, accurate cost allocation should be based on which activities the resources are used for, and not rely on broad averages based on volume. In the field of iterative innovation R&D, cross-functional personnel are often switched from one project to another, and arbitrary cost pooling will lead to significant errors. ABC theory will guide the next step of customised model construction; that is to say, different cost drivers and resource pools in rapid prototyping cycles must be found to allocate R&D expenses based on the actual use of technology rather than arbitrary departmental apportionment.

Thirdly, real-time financial matching theory extends the idea of periodic matching and requires that the expense recognition time should be the same as when the economic benefits are created. Due to iterative innovation's continuous growth and not one-time market introduction, the rigid accounting period is not suitable for the financial position of the company. This theory will serve as the basis for accounting deficiency diagnosis, that is, it will show the time lag between immediate R&D expenses and the delayed revenue generation, and also guide the construction of a customised model, requiring agile recognition triggers to be added in line with real-time milestones during the development life cycle to control cost deferral. In short, these theories are now beginning to serve as a theoretical basis for changing the R&D accounting of start-up companies from a problem to be dealt with in accordance with the law to a strategic management tool.

The sample selection for the field study is required to be reasonably close to the operation of the area under study and academic standards. The targeted tech startups should not be older than five years, which reflects the characteristics of early-stage companies that are relatively young and have limited resources. In addition, these companies have not yet had stable positive operating cash flow; therefore, they urgently need good management of R&D spending to maintain their financial resources for a relatively short period and meet the requirements of external investors. The chosen companies will be those engaged in the independent development of technology for specific parts such as software, hardware prototypes, etc., and new materials. These sectors have different accounting difficulties in R&D: some may be quick to change, while others need more time to test the material physically. Additionally, to record the real organisational changes of startups, each new company has a small formal finance team with no more than three full-time accountants, and the organisational structure is relatively simple to implement the new accounting system.

The above plan is to set up eight research stations in the urban innovation park that naturally attracts various tech startups. The Innovation Park is a collection place for a large number of enterprises and has a good working environment for startups with various functions. The samples are for both the seed and Series A stages of technological ventures. Both of these stages are crucial; at the seed stage, companies are experiencing initial difficulties with capital for research and development due to lack of resources, and at the Series A stage, institutional investors are requiring more details on R&D expenses in their financial statements to enhance the representativeness of the collected data.

Semi-structured interviews are used to collect data based on the varied kinds of information provided by different interviewees. For startup founders, the interview framework mainly examines how to distribute strategic R&D funds, determine the timing of commercialisation, etc., and whether they consider accounting data as a means of strategic decision-making rather than just compliance requirements. In terms of the head of the R&D team, this survey will be about the technical implementation path of the project and how to record the milestones of the project, the speed of resource consumption, and other aspects of the project; these records will then be presented in the form of common accounting items. For full-time accountants, this will involve the practical implementation of the current R&D accounting policies and the specific problems in operation, such as distinguishing between research and development periods, capitalisation of prototype costs, etc. Through the above three kinds of interview methods, it is possible to study both the actual situation of technical R&D and the limitations of financial reporting; therefore, it will serve as an empirical basis for customising the accounting model.

The field investigation data collection framework will be used to obtain empirical data and provide a basis for the construction of the research and development expenditure accounting model based on real-world conditions. It is mainly a structured, multi-level way to collect quantitative financial traces and qualitative operational data, avoiding secondary processing errors and guaranteeing the authenticity of first-hand data. The operation process of the above framework will use four interrelated core modules.

First, a basic operation information collection template will be prepared and then used for collecting the operation data of each sample startup. The module will standardize the extraction of key operation parameters, such as the size of the R&D team, the proportion of R&D budget in the total operating expenses for the year, and the current progress of R&D projects. A certain amount of data will be used in this quantitative baseline to adjust the later accounting analysis of the actual scale and intensity of the innovation activities of start-up enterprises.

Second, the actual current R&D expenditure accounting voucher review system will conduct a thorough check of all financial data. The operation path is to examine the original accounting vouchers and classify the specific items of outlays that are now recorded as R&D expenditure for each enterprise. At this time, the old way of classifying expenses will be identified; we will also find any irregularities in capitalisation or recognition of expenses; and the current structure of R&D costs will be known.

Third, a framework for collecting the pain point interview record will be developed to collect the problems of daily accounting in an operation-oriented and qualitative manner. Based on a structured conversation with the Finance staff, some of the difficulties, uncertainties in the application criteria and administrative obstacles that have arisen in the operation of R&D accounting can be recorded. The collection of these subjective but professional experiences is essential to ensure that the custom model is tailored to the actual operating environment rather than being overly idealised in theory.

Fourthly, the cross-source data verification rule is the quality control system of the framework. The second module's financial accounting records will be used to match with the R&D activity logs provided by the R&D team in the first module. By comparing the financial outflow with the actual progress of the project and the consumption logs of resources, it is determined that the data collected are accurate, authentic, and have not been processed. This strict reconciliation is very important to maintain the empirical validity of the field study.

The basic concept of R&D expenditure accounting is to systematically recognize, measure and disclose the costs of research and development activities. The main idea is to divide the research phase and the development phase, and then set the capitalisation threshold, so as to objectively reflect the situation of intangible assets. However, field data from tech startups indicates that these general accounting rules do not match the actual conditions of fluidity and uncertainty in the R&D of startups.

First of all, the general accounting standards set out a clear division of the "research phase" and the "development phase", but this is at odds with the repeated processes of tech startups in practice. In practice, early-stage ventures are rarely in a linear growth path; rather, their R&D is characterised by continuous pivoting and rapid changes based on immediate market reactions. Based on the survey data of one software startup, it is known that their core algorithm development repeatedly changed from thinking to operation within one month. The traditional two-phase division does not reflect this kind of continuous change, so it is difficult to allocate expenses according to time and accurately, and there will be distortions in the periodic financial statements.

Second, the general standards have established a strict capitalisation threshold, so the project needs to obtain all necessary technical and commercial approval documents. This way cannot be used by early-stage enterprises facing high uncertainty. A field survey of a hardware startup found that its initial feasibility proof-of-concept did not have the mature, formal project charter required by capitalisation. Therefore, a large amount of expenditure that can be used to create intangible assets is fully expensed, reducing the company's assets and exaggerating the loss in the initial operation period, thereby affecting the value of later financing.

Thirdly, the general expenditure classification standards are not suitable for the changes in the distribution of funds among startups. The general structure is a well-structured and paid research and development department with mass purchase of materials; thus, it does not provide reasonable categories for many expenses. Field observation found that biotech start-ups often hire temporary part-time researchers to perform special micro-assays and buy small batches of rapid prototyping raw materials. Flexible, scattered R&D is important for the development of a start-up company; however, there are no dedicated categories for such flexible, scattered R&D in the general classification rules, and they are therefore coded as general administrative or overhead expenses by accountants.

The systematic misalignment shows that although the existing standards are strict for mature companies, they actually conceal the real situation of R&D activities and the ability to create assets in tech start-ups. To meet these new forms of business, the accounting model needs to be adjusted so that it can provide clear financial information for the early-stage investors.

The accurate accounting of R&D expenditure can be achieved by using real-time verification and reasonable allocation methods to ensure that the financial records actually reflect the economic reality of innovation activities at that time. However, based on the field data, many tech start-ups are unable to follow these principles due to some real problems, which will affect the reliability of their financial accounting.

At present, there is no mutual information sharing between the R&D team and the finance department in time. The survey results of the field questionnaire show that 85% of the participating start-ups are currently summarising periodically rather than having an integrated project-financial track system. Therefore, there is a lack of timely data for accounting the R&D costs; they are often recorded after a long time. According to the interview transcripts, the delay is more than 30 days. The delay in verification will make it impossible to know the real burn rate of the R&D project, making real-time cost control impossible and thus violating the principle of timely recognition in accounting.

Rational allocation also has many issues. There is no clear division of resources; that is, these resources are often used for several purposes. Office equipment, cloud computing service resources, etc., are often used for both R&D activities and daily work in start-ups. Field data reveals that 72% of the sampled companies do not have a reasonable allocation basis for the proportion, and they often arbitrarily split it. Since there is no metering record of usage or a reliable allocation standard, the finance department cannot reasonably allocate the relevant expenses of the shared system; consequently, both the cost of R&D and that of other operating expenses have been inaccurately stated.

Besides, the difference between general technical adjustment work and formal innovative R&D activities is not obvious for most early-stage startup accountants. Industry accounting standards are supposed to distinguish incremental innovation from ordinary maintenance, but 68% of the surveyed financial personnel reported problems with applying this conceptual division in practice. Technical ambiguity will directly lead to a systematic error in recording. According to the questionnaire, 43% of the startups have long-term low R&D expenditures that should have been included in eligible development and optimisation; at the same time, 29% of the research and development expenses are not properly recorded. These misclassifications are caused by the absence of operational procedures that can connect the technical project's milestones to accounting recognition events.

Based on the above reasons, we can know that without structured information sharing and clear rules for distributing resources and classifying activities, the R&D expenditure accounting model of tech startups will have some shortcomings in both financial integrity and compliance with the rules of the industry.

The multi-dimensional classification framework for R&D expenditure items is tailored to meet the needs of the iterative innovation rhythm of tech startups. The traditional accounting system uses only one classification method, and it cannot reflect the dynamic changes of R&D in startups. The framework will not be limited by the above, and all R&D-related expenses will be divided into three main categories. The first is the attributes of R&D activities; there are three types of expenses in exploratory trial, iterative improvement, and mature pre-launch testing. The first kind of exploratory trial is to develop new ideas and prototypes; the second is to improve the agile development of the system continuously; the third is to complete the stability work before the product is released. The second criteria are to determine the scope of different development stages of the company; that is to say, how many bugs will be repaired in a certain period of time, how many new functions will be added, etc. The second type of expenditure economic nature will be divided into labour cost, material input, technology service outsourcing, long-term asset amortisation and other items. The secondary classification criteria will separate the specific cost drivers, and so on; for example, the salary of the core developer is higher than that of a freelance consultant, and so is the rent for cloud infrastructure. The third one is the connection to certain R&D projects; the second criterion is to differentiate direct project-specific allocation, shared cross-project expenses, and general R&D platform support costs.

The core idea of the frame is to help startup Finance staff quickly sort R&D expenses, not needing complicated formal pre-approval documents or strict project codes that lag behind the actual progress of development. If there is an iterative optimisation task, the staff can classify it immediately according to the actual reason of the expenditure and the nature of the economy. For example, according to the field research samples, a startup had a payment for third-party API integration testing. Since there was no formal project number at the beginning, the finance team classified it as follows: in terms of R&D activities, it was an iterative optimisation (secondary; feature integration); in terms of the nature of the economy, it was technology service outsourcing (secondary; API service fee); and in terms of project relevance, it was a direct special allocation to a particular project. Based on various aspects, rather than approval, we will now be more flexible and parallel in our categorisation. Therefore, it maintains the flexibility needed for the start-up and meets the requirement of accounting for fine-grained costs, so that enterprises can know about changes in costs at all stages of rapid development and provide accurate data for the subsequent capitalisation or deduction of R&D expenses.

The method of step-by-step capitalization and expense recognition based on the field operation path has fundamentally altered the way in which tech startups account for R&D expenses. The old accounting standards required a clear division into two stages of research and development, which did not match the flexible nature of the initial small company's invention. This mechanism does not adopt a dualistic model and establishes three stages of development that correspond better with the true state of tech startup R&D activities, that is to say, after the technical feasibility verification of the first stage of the R&D objectives; after creating a functional prototype that can operate normally; and after finishing the pre-release test of the developed technology product. The trajectory-based method enables the accounting recognition to be based on the actual situation rather than some arbitrary deadline.

At each new stage, the R&D expenditures that meet the last recognition condition can be partially capitalised according to the actual technical progress achieved at that time. Expenses incurred during the initial unproven exploratory period before the first node are recorded as current expenses. It can thus safeguard the company's income statement from the risks of speculation due to early-stage exploration, and it can also build asset value gradually as technology becomes more certain.

An example of this model is taken from a full R&D activity path of an independent company that has developed an AI-driven analytics engine. The start-up will invest $50,000 in algorithm research and data acquisition at the beginning. Since it is before the first node, this entire amount is considered an expense in the current period. After verifying the technical feasibility of the core algorithm, we have arrived at the first node. Subsequently,$80,000 was spent on model training and construction design. The second node is reached when the start-up has completed a stable and functioning prototype; then, $80,000 will be partially capitalised based on the level of technical progress and thus recognised as an asset. The first$40,000 will be used for the development of the product. After the third node has been reached after a full pre-launch test, the $40,000 is capitalised; at this point, the product has shown its readiness for sale in the market.

By linking capitalization to actual technical milestones instead of subjective phase names, this way can directly show the R&D expenses for each project. Therefore, it can meet the needs of both the prudence principle in financial reporting and the economic reality of the value creation of a startup. Thus, the current accounting model is both theoretically sound and quite practical for operation in the field.

Fieldwork samples from various cases are selected to verify the model, and this is the basic method to ensure that the built targeted R&D expenditure accounting model is both logically coherent and practically useful. The basic idea of this verification is to test the reliability of the model by comparing the results it produces with the baseline data of different environments and eliminate the contextual bias that arises from various background conditions to prove that it is generally applicable. A total of four representative technology start-ups were chosen for the research; two were at the seed stage and two at the Series A stage, and among them, one is a software development company and the other an intelligent hardware research and development enterprise in each growth period. Therefore, according to different operational and financial rhythms of various sub-sectors and early development periods, it can be segmented.

The newly-built accounting model is applied to reorganise and recalculate the twelve months of historical R&D expenditure data of each of the four samples. Subsequently, the output accounting results generated by the model will be compared with the original accounting data that was processed by the startups' internal finance departments. The comparative analysis mainly included two aspects: first, to see whether the operational efficiency of the startup finance staff has been improved by reducing their workload, and second, to test the technical accuracy of the accounting model by checking if there is a close match between the recognised R&D expenses and the actual R&D progress of the startup.

From the data in the table, it can be seen that there are some changes in practice. In terms of operational efficiency, the new model will help reduce the workload of the early-stage enterprise's R&D accounting. Standardize the collection and classification of scattered R&D expenses such as cloud computing costs for software enterprises and prototype material purchases for hardware enterprises; thus, there is no longer a need for manual sorting and subjective judgment, and the monthly closing process is simplified. The recalculation result is better than the original data in terms of matching the R&D amount with the actual R&D progress; that is, the accounting accuracy is relatively high. The original accounting frequently has timing mismatches, and thus it is either too early to capitalise exploratory expenses or too late to recognise the milestone-triggered expenditures. The new model has been built based on the stage-gate progress standard, so the expenditure recognition time is the same as the time when the technical achievement has been completed.

Finally, based on the cross-case comparison, it can be known that the targeted model can adapt well. Although there are some differences in the nature of software development and intelligent hardware R&D, and the resource constraints of seed investment and Series A are different, the model can be adjusted to meet the needs of various industries and early-stage companies. Therefore, the model has good utility in providing a uniform, reliable and effective financial management system for new high-tech enterprises.

Through the field research, it was determined that the R&D expenditure accounting of small technology companies has several problems. Based on firsthand observation, there are often unclear boundaries of capitalisation, irregular changes in R&D activities that affect standardised accounting, and a lack of dedicated financial personnel, which have caused significant distortion in R&D cost information and raised compliance risks ^[3]. To solve these problems in the operation of the company, a multi-dimensional classification and step-by-step recognition R&D expenditure accounting model has been constructed in this paper. The main advantage of this model is that it can be adjusted according to the actual operation of small tech companies. A multi-dimensional classification is used to categorise the expenditures of the project's different stages, various resource types, and different funding sources, and thus capture the dynamic and iterative nature of startup R&D. At the same time, the step-by-step recognition path can be used to distinguish capitalisation from expense in a step-by-step way, reducing the harshness of traditional binary thresholds. In this way, the accounting results will be based on the economic substance of R&D activities rather than just strict formal compliance. The model is suitable for early-stage tech startups that are relatively new, have high R&D costs, need to quickly change their strategic direction, and are limited in financial management ^[6]. Based on the above research, some practical suggestions are given for organisations to implement the model. First of all, the startup needs to build a relatively simple project tracking system to ensure that the initial data for classification is accurate. In addition, specific stage-gate recognition criteria need to be set for the technological life cycle of the startup, and regular cross-departmental audits should be conducted to align the accounting process with the actual situation of R&D progress and ensure the continuous operation of the model. Although it has certain practical significance, this research has some shortcomings. The sample of the field study is relatively limited, mainly concentrated in the software and information technology sectors, so it cannot be applied to hardware-driven or biotech startups with long R&D cycles. Also, the dynamic relationship between the accounting model and external tax incentive policies has not been comprehensively examined. Therefore, more industries should be added to expand the empirical sample and refine the model's parameters in the future. Further in-depth research can also carry out quantitative optimisation of accounting treatment to maximise legitimate tax benefits, develop a standardised digital integration scheme to embed this accounting model directly into the enterprise resource planning system, etc., to improve the operational efficiency and theoretical soundness of the company.

[1]王宁. 论企业研发项目管理与研发支出核算[J]. 中国民商, 2020, 000(012): 90.

[2]夏友全. 新型研发机构研发支出核算体系探究[J]. 新会计, 2025(6).

[3]罗隽. 浅谈研发流程和研发支出归集规范化[J]. 财经界, 2020(6): 2.

[4]赵雪霞. 企业研发项目管理与研发支出核算研究[J]. 中文科技期刊数据库(全文版)经济管理, 2022(10): 3.

[5]刘肖兰. 研发支出与项目成本界限模糊的成因及对策探讨[J]. 中国农业会计, 2026, 36(6): 18-20.

[6]李春燕. 智能化管理会计在高新技术企业研发成本控制中的应用[J]. 中国电子商务, 2026, 27(2): 73-76.

[7]Sabri Boubaker, Jing Du, Rui Xu. Employee stock ownership plans and the capitalization of R&D expenditures[J]. Technological Forecasting and Social Change, 2026, 233: 124872.