DEA Analysis of Cloud Bookkeeping Cost Optimization for Small E-Commerce Sellers

Recently, the small e-commerce seller sector has grown rapidly, and there are more and more individual and micro-sized merchants on various mainstream third-party e-commerce platforms. However, although the company has expanded in the commercial sector, many problems have arisen in operation. Their own internal financial management is usually weak and has no formalised financial accounting system. At the same time, traditional offline bookkeeping services are too expensive to be used by small enterprises. Therefore, these sellers often encounter compliance risks such as tax declaration and expense accounting, and manual or relatively simple financial tracking cannot meet the strict requirements of the regulatory system and industry rules.

To address the above problems, cloud bookkeeping services have emerged as a third-party financial solution and are rapidly expanding in the market. Cloud bookkeeping uses the digital infrastructure of the cloud to provide a flexible, cost-effective, and alternative solution for small e-commerce sellers, theoretically reducing compliance and efficiency problems. But at present, there is still an urgent problem. Although cloud bookkeeping reduces the cost to some extent compared with the old way, there is no targeted and quantitative evaluation tool to know whether the cost that small e-commerce sellers spend on cloud services will be in direct proportion to the operational benefits. There is no reasonable way to determine how much to spend on finance versus how much to make, so the sellers cannot adjust their costs reasonably and improve the use of their resources. They still do not know whether they have overinvested in unnecessary functions or underinvested in necessary compliance support.

There is a lack of research on how to evaluate the improvement of service efficiency; thus, we will use Data Envelopment Analysis (DEA) to conduct this research. Non-parametric methods are used to calculate the relative efficiency of decision-making units (DMUs) with many inputs and many outputs; this is known as data envelopment analysis (DEA). By means of the above analysis, one can determine whether the current cost structure of cloud bookkeeping is optimised for operating income and help small e-commerce businesses formulate standardised, objective plans to identify problems precisely in their financial management, improve their financial situation, etc.

The goals of this paper are divided into three levels. First, this paper will construct a targeted evaluation system based on the operational characteristics of small e-commerce sellers to measure the input-output efficiency of cloud bookkeeping costs. Based on the subscription fee, the labour involved in migrating data, the training time, etc., as well as the output indicators of financial automation gains and compliance accuracy, this system can transform the abstract theory of cost management into a specific operation process. Second, based on the above framework, this paper will use the DEA empirical calculation method to determine the level of cloud bookkeeping cost utilisation efficiency of the selected small e-commerce sellers. To analyse the reasons for cost inefficiency, one can determine whether it is due to excessive resource use or a lack of financial output. Third, this paper will put forward some operable and data-based suggestions for small e-commerce sellers to use cloud bookkeeping services. The above strategies will offer clear pathways for the redistribution of resources, and technical accounting solutions will be in line with the actual business scale.

In terms of theory and practice, the importance of this study will be shown. Theoretically, it is to expand the application scenarios of cloud accounting efficiency evaluation research for small e-commerce sellers. Most of the traditional accounting efficiency models aim at medium-sized and large enterprises, and they do not consider the particular constraints and large amount of low-profit transactions of micro e-commerce. By adjusting the DEA model to the operation characteristics of micro-main-body enterprises, it can make up for the lack of relevant research on the financial digitisation of micro-main-body enterprises in the literature. In practice, it can provide a basis for small e-commerce sellers to decide how much to invest in cloud bookkeeping costs and increase their financial management returns with limited resources. Based on the results of the empirical study, rather than arbitrarily reducing costs, sellers can precisely control the amount of inputs and cut unnecessary expenses. This will make the adoption of cloud bookkeeping services change from a fixed operating expense to an investment with a certain degree of planning and promote the continuous development of small enterprises financially.

To avoid the vague expansion of research objects and findings, one needs to set a clear range and boundaries for the study. First, the "small e-commerce sellers" in this paper are limited to individual industrial and commercial households and micro-enterprises that have independent stores on mainstream consumer e-commerce platforms. The operating income of these companies is less than 5 million yuan per year, and there is no full-time professional financial management staff. Therefore, large-scale cross-border e-commerce enterprises, self-operated brand platforms, and individual sellers engaged in casual part-time transactions are excluded from this definition. Based on the above people, the study will focus on the daily life and financial difficulties of small enterprises with limited resources.

Secondly, the "cloud bookkeeping cost" in this paper is limited to three aspects: direct paid service fees charged by cloud bookkeeping platforms, supporting hardware and software configuration costs for operating these cloud systems, and the time cost of internal staff to upload, sort, and organise financial data. All other costs related to operations do not involve these items; instead, they are not included in the direct financial recording expenses. Separating these specific costs helps the assessment model identify the actual costs of implementing and maintaining the cloud accounting system.

Third, the Data Envelopment Analysis (DEA) used in this paper only covers the full fiscal year 2023. Therefore, the results are only applicable to small e-commerce sellers of similar scale in the domestic consumer e-commerce track. These boundaries will not be extended to large e-commerce enterprises or cross-border e-commerce merchants, as their operations are subject to complex cross-border tax regulations and fundamentally different financial compliance environments. Based on the above-stipulated time, space and operation, the study has certain analysis results and provides specific and practical suggestions for improvement for a certain group of people.

We will use various research methods to investigate the problem of cloud bookkeeping for small e-commerce sellers and reduce the cost. First, we have reviewed the relevant works on cloud accounting and cost management in academic papers. This will show the current situation of the field and help us know where there are problems and lay a theoretical foundation for the next empirical analysis. The data envelopment analysis (DEA) will be employed as the primary quantitative method. Based on the characteristics of cloud bookkeeping input and output indicators, DEA can determine the relative efficiency of various decision-making units and help us see whether sellers are now operating at an optimal resource allocation under the existing cost structure. To better understand the results of the DEA, we will use case comparison analysis. This method will compare the operation characteristics and efficiency differences of technically valid and invalid decision-making units and find out the specific operating bottlenecks that cause cost inefficiency. Finally, based on the projection analysis method, the precise adjustment value for the input and output of non-effective samples can be calculated. Therefore, there are clear targets to reduce costs and improve operation, and the gap between theoretical efficiency measurement and practical cost optimisation has been filled.

The general logic of the thesis is that the chapters are closely related and arranged in an orderly manner. The first chapter will introduce the research background, purpose and ways. The second chapter is a literature review and theoretical foundation. Chapter 3 will specify the DEA model and construct an indicator system for e-commerce cloud bookkeeping. The fourth chapter is the empirical analysis and projection of the cost adjustment efficiency score. Chapter 5 is the conclusion of the study and practical suggestions for improvement. The general logic of the whole research is to first identify the problem through a literature review, then build a theory and indicator system, measure efficiency with Data Envelopment Analysis, compare and project to find inefficiencies, and then propose targeted solutions. This arrangement has a smooth change from a qualitative theoretical study to a quantitative empirical study, and finally provides some guidance for the practice of cloud bookkeeping cost optimisation.

Cloud bookkeeping services are used in the case of small-scale e-commerce and refer to a software model that provides financial services for all types of businesses in the cloud in an efficient, secure and convenient way. Based on the existing definitions of cloud accounting in academia, this idea is intended to solve the problems of small e-commerce merchants having to do a large volume of work at low profit margins in a dynamic environment. It is different from the traditional offline agency bookkeeping mode, which is manual in nature and has to be updated regularly, making it slow and prone to human error. In other words, it is not as good as the general enterprise financial software that requires a large amount of investment for new construction of hardware and software in-house and is too rigid for small-scale enterprises. The most basic idea of cloud bookkeeping for small e-commerce is to be light, convenient and able to meet the specific financial data processing needs of such enterprises.

The operation of this service is customised to the e-commerce ecosystem. First, it can be linked with all kinds of e-commerce platform back-end systems to automatically collect the required data for financial management without manual data entry; therefore, one does not need to manually input data and can keep up with the latest financial information at all times. Second, it can automatically identify and classify a large amount of order-transaction data through rule-based algorithms and then combine scattered micro-transactions into a unified account. Third, it will automatically calculate the revenue and handling costs of e-commerce, and thus it can accurately identify the gross profit margin per transaction to help understand the profitability of an enterprise. Fourth, remote on-demand upgrading of tax declaration rules can be performed; when the system is upgraded, it will automatically adjust according to the new rules without users having to update them manually. Finally, based on the small amount of money available for small enterprises, a tiered charging scheme has been established; the service fee will be determined by how much they have to pay in transactions or how many modules they need rather than being a high, fixed licence fee. Based on these different operating characteristics, the DEA input-output indicator system will be built.

The logic of cost optimization for small e-commerce sellers to use cloud bookkeeping is not the old way of blindly reducing expenses. In the past, cost reduction was considered to be less expenditure, but for small e-commerce sellers, this simple model weakens the ability to continue operating. The purpose of reducing the cost of cloud bookkeeping is to maximise the marginal benefits of each unit of cost under a certain total cost. Through the reasonable adjustment of the resource allocation structure related to cloud bookkeeping costs, sellers can improve the efficiency of financial management, reduce the risks of violation and provide more accurate operating decision support. This logic no longer regards the minimum total cost of cloud bookkeeping as the goal of cost optimization, but rather to maximise the value generated by each unit of financial investment. For instance, when the basic manual data entry is changed into an advanced cloud-based analytics tool, the cost of basic manual data entry becomes a strategic asset and increases the company's profits directly.

In addition, this logic that follows input-output efficiency is also suitable for the resource shortage characteristics of small e-commerce sellers who do not have many funds. Large enterprises usually use standard cost management methods and set up strict budget controls and economies of scale to reduce absolute costs. However, small sellers do not have the necessary funds to cover the opportunity cost and other compliance penalties caused by underinvestment in financial infrastructure. A small penalty for non-compliance or a cash flow error could cripple their business. Therefore, the cost optimisation of small sellers should not be for expenditure but rather to maintain security and decision-making through cloud bookkeeping. Small sellers will adjust their financial resources in real time by adjusting the scalable modules of cloud bookkeeping; thus, for every unit of cost paid, they will receive a much higher marginal return in reducing risks and obtaining strategic information, which is more relevant to their life and growth than reducing the absolute cost of accounts.

Data Envelopment Analysis (DEA) is a new kind of nonparametric method for judging the efficiency of the decision making unit (DMU) that has many inputs and many outputs. The basis of its operation is the construction of an empirical production frontier based on linear programming and the observed data of the best sample. This frontier serves as a reference for evaluating the relative efficiency of each DMU. The basic idea of the logic is that any DMU on the frontier has achieved the maximum output for a given input and is thus highly efficient; conversely, any DMU below the frontier is relatively inefficient, and the radial distance from the frontier can be used to measure the degree of inefficiency.

DEA is a good tool for evaluating the cost reduction of cloud bookkeeping for small e-commerce sellers, so it does not need to set the weight relation between different input and output indicators in advance. Traditional parameter-based methods often introduce large errors due to subjective weight allocation and may lead to biased evaluation results. By obtaining the weight of the indicator endogenously through mathematical programming, DEA can avoid the problem of subjective weight deviation. Small e-commerce sellers have diverse needs for different cloud bookkeeping cost input indicators; some are in urgent need of funds, some are not so urgent, etc. Therefore, there is no need to have one standardised and subjective weight structure, and the model should also consider the actual situation of each seller in terms of size and strategy.

In addition, the DEA model has the function of breaking down the general efficiency into pure technical efficiency and scale efficiency. Pure technical efficiency is the ability of a seller to use the bookkeeping resources at the current size of operations; scale efficiency is whether the seller operates at the best size for resource allocation. According to the above decomposition, the research can determine the specific reason for cost inefficiency, i.e., whether it is due to poor management or too many resources. Therefore, this Methodological Feature can offer some technical support for the formulation of targeted cost reduction advice and guide practitioners in more accurate resource adjustments than generalised cost reduction measures.

Overseas research on cloud accounting service efficiency assessment has developed from an initial exploration of the value creation path of using cloud computing to improve financial management services to the present day. Some scholars believe that the main purpose of cloud bookkeeping is to lower fixed costs for enterprises and increase the flexibility of their funds. The basic idea is that the cloud architecture no longer requires on-site infrastructure maintenance and can be run in the form of a scalable, subscription-based accounting system. The later empirical study is aimed at measuring the input and output efficiency of cloud accounting services for small enterprises in various industries. The inputs in these studies are usually operating expenses, the cost of data migration, and the training time of the staff; the outputs are the speed of processing, the accuracy of financial reports, and the speed of access to real-time data. According to the empirical data, it was found that small-scale e-commerce enterprises with the most seamless integration between digital payment gateways and cloud accounting systems have relatively high marginal efficiency gains. However, foreign literature also points out that there are other reasons why small enterprises cannot obtain the expected returns from cloud accounting investment. Data security risks, the difficulty of integrating with older systems, and a lack of professional technical accountants have been identified as major problems. These constraints will affect the normal operation of cloud services and may raise implicit costs for users. Overseas empirical studies generally use specific efficiency evaluation tool models to quantitatively assess the above problems. Data Envelopment Analysis (DEA) and Structural Equation Modelling (SEM) are two typical methods that do not need to define production functions for multiple-input and multiple-output problems. DEA has been used in many places abroad as a general, non-parametric benchmarking tool to compare the relative efficiency of different portfolios of cloud accounting services. The advanced experience obtained from this overseas study has provided a theoretical and methodological basis for the present research. Therefore, it is necessary to build a unified evaluation system with DEA that can account for industry-specific constraints and integration difficulties and carry out a strict and practical cost optimisation study of small-scale e-commerce sellers' cloud bookkeeping.

As the digital economy has developed rapidly in China, there have been more studies on the operational cost management of small and micro e-commerce enterprises in China, which are mainly based on their particular financial characteristics. Scholars generally consider the operational costs at this time to be financial management expenses, tax compliance expenses, and information system investment costs. For the financial management cost issue, researchers have pointed out that small e-commerce sellers often have high capital occupation and inefficient fund allocation due to the lack of standardized accounting procedures. There are many tax compliance costs, so a number of papers have pointed out that there is a lack of standardisation in the small and micro-enterprises' business management and accounting systems, resulting in relatively high tax compliance costs for them. For the cost of information system investment, domestic studies have been carried out to compare the funds and operating expenses of small sellers using localised bookkeeping software with those of cloud-based solutions; however, due to the high initial cost of hardware and maintenance fees, the financial situation of small sellers is currently in a relatively difficult state.

Although some of the reasons for the high cost of domestic services have been pointed out, many problems remain in the field of domestic services at present; for example, there is no fine division of costs in the current stage, serious information asymmetry in financial processes, and general purpose accounting software that cannot meet the needs of agile, high-frequency transactions in e-commerce. To address the problems above, most of the solutions in the existing domestic research have been cloud-based accounting architectures, integration of automatic data collection technologies, and regularised internal financial controls. However, there is a serious deficiency in the domestic academic environment for research on this topic; that is, most of the studies on the problems of cost management and concept development of strategies are qualitative and general in nature. There is no quantitative empirical research on the specific segmentation scenarios of cloud bookkeeping. There have been few studies on the use of strict mathematical programming models (Data Envelopment Analysis) to conduct quantitative evaluations of the cost optimisation efficiency of small e-commerce sellers in the process of cloud bookkeeping, and there is a lack of data-driven operational verification for the theoretical qualitative frameworks.

According to the theoretical basis and literature review provided above, the main problems that need to be solved in the relevant research are presented here. First, most research on cloud accounting efficiency evaluation is aimed at large and medium-sized enterprises. These organisations have a large structure and plenty of funds, so the evaluation index system cannot be applied to small e-commerce sellers. Small e-commerce sellers have a high number of small-value transactions, changing inventory turnover, and lower financial literacy. Thus, according to their own operating characteristics, an evaluation index system has not been built in the existing literature, and there is a lack of theoretical support for assessing the efficiency of cloud bookkeeping for small e-commerce sellers. Second, although there are many studies on small e-commerce cost management, they have not introduced strict non-parametric quantitative methods such as Data Envelopment Analysis (DEA) to empirically calculate the cost-input-output efficiency of cloud bookkeeping in practice. DEA can be used for multiple inputs and outputs at the same time without specifying a particular production function, so theoretically, it is suitable here. The empirical application of DEA to measure the relative efficiency of cloud bookkeeping investments in small e-commerce sellers is not well-researched. Third, the existing recommendations for cloud bookkeeping cost optimisation are generally general measures of the industry, such as the widespread proposal of standard software packages and the automation of basic data entry. They have not been optimised according to the actual efficiency measurement results of small e-commerce seller samples. Without the diagnosis of specific inefficiencies through empirical DEA projections, such as technical inefficiency or scale inefficiency, the proposed solutions will be too general and lack precise guidance. Based on the above three deficiencies, namely building a targeted indicator system, using DEA for empirical measurement, and obtaining a targeted optimisation path based on efficiency results, this thesis will construct a standardised quantitative and practical framework for cloud bookkeeping cost optimisation that can be applied to small e-commerce sellers.

To build the index system of cloud bookkeeping costs for small e-commerce sellers, we will use indicators that are representative, easy to obtain and independent. Based on the operating characteristics of these sellers, such as a large number of transactions, low profit margin and lack of specialised financial personnel, we have constructed three dimensions of input.

The first input indicator is the annual direct service fee investment of the cloud bookkeeping platform. The explicit financial expenditure paid to the service provider for the purpose of software licence, data storage and function modules. Its selection basis is its representativeness as the main reason for cloud computing adoption, and its accessibility is that this fee can be seen directly on the subscription invoice.

The second input indicator is the annual supporting resource investment, which includes necessary equipment and auxiliary accounting tools to support the use of the cloud bookkeeping system. E-commerce sellers need to buy compatible point-of-sale peripherals, barcode scanners and API integration tools. The reason for choosing this indicator is that it contains the hidden hardware and middleware costs needed to run the system, and it is not directly linked to the platform's service fees.

The third input indicator is the labour time cost that small e-commerce sellers spend annually to organise transaction vouchers and cooperate with the cloud bookkeeping data entry. This operating cost monetises the time spent by the seller or its staff extracting sales data from the e-commerce back-end system and organising the freight bill before data entry. It is included because human-machine interaction is one of the important inputs in a semi-automatic cloud environment, which will directly affect the operation's efficiency.

It is also important to determine what other expenses are not in this category. The indirect opportunity cost is not included because it is a subjective indicator and there is no objective data available. Similarly, the sunk costs of previous accounting systems will not be part of the cloud bookkeeping's current optimisation inputs. Finally, the macro-environmental cost of general internet service is ignored because it is not directly related to the bookkeeping activity and thus violates the principle of independence. It will exclude the DEA model that is currently available.

A good output index system needs to be built to show the value that small e-commerce sellers have created from the cost investment of cloud bookkeeping and cost optimisation. The essence of this system is to convert the direct financial and time inputs into tangible operating results; a strict matching rule for the resources consumed and the value produced must be in place. It is divided into three multi-dimensional indicators. The first is the annual reduction rate of financial non-compliant penalty losses due to tax declaration errors, account mismatches, etc. It is based on the percentage decrease in actual penalty expenses in the current fiscal year relative to the baseline period before the adoption of cloud bookkeeping. Logically, this is equal to the cost input indicators of subscription fees and software maintenance fees, and the automated compliance algorithm and real-time reconciliation function of the cloud platform can directly reduce the risk of manual regulatory violations. The second output indicator is the average working hours saved per month for the seller's hand-financial and accounting work. Based on the difference between the average number of hours manually spent on bookkeeping in the past and the current supervision of the cloud system. The above outputs are all related to the cost of personnel training and system construction; one-time learning and construction costs for the cloud system will gradually be offset by the continuous reduction in manual work in daily life, thus improving overall time efficiency. The third is the accuracy improvement rate of operational profit data that supports daily operation, commodity stocking, and pricing decisions. Based on the variance between the cloud-generated profit forecast and the actual profit, compare it with the variance of manual accounting. The logical correspondence is that the first is the data processing and update service fee among the cost items, and the second is the ability to continuously provide financial data and advanced analysis so that precise, real-time profit tracking can be performed. The above three indicators reflect the risk reduction, time savings, and decision-making improvements brought by the cost input of cloud bookkeeping and provide a basis for the subsequent DEA efficiency evaluation.

The selection of suitable decision-making units for Data Envelopment Analysis (DEA) is one of the reasons why the data collected can be used in practical applications; therefore, we should be very thorough in our choice criteria for homogeneity and representativeness of the sample data. Selected small e-commerce sellers need to meet three basic conditions at the same time. First, the operating scale should be in line with the definition of small e-commerce sellers in the research, and there is no full-time professional financial personnel. It can be seen that all the sampled enterprises are constrained by similar amounts of resources and have to use cloud bookkeeping services regularly for accounting; consequently, they have a uniform basis for calculating costs. Second, the seller should have been continuously using the same cloud bookkeeping service for at least one year. Based on the above, we can obtain complete and accurate annual data for the costs of cloud bookkeeping and their corresponding effects. Twelve months of continuous operations can better eliminate the influence of short-term seasonality or new system adaptability and other factors; thus, they are suitable samples for studying long-term cost efficiency. Third, the core business of the sellers is physical commodity sales on mainstream domestic consumer e-commerce platforms, and the structure of the transaction data is consistent. The choice of business type is designed to exclude samples with vastly different operating models, such as virtual service providers and cross-border traders, etc., so that the method of DEA does not change; that is to say, the assumption of technology homogeneity for all decision-making units holds. Based on the three aforementioned conditions, a total of 42 qualified small e-commerce seller samples will be used as the basis for the subsequent empirical analysis. The data sources for this kind of sample mainly come from field questionnaire surveys of the target sellers to obtain their operating details and anonymous financial data from the cloud bookkeeping service provider that has been authorised, thus ensuring the objectivity and accuracy of the data. A dual-channel system is in place to ensure the authenticity of the data and the reliability of the empirical results.

After collecting some data from 42 small e-commerce sellers that are sample decision-making units, the data was sorted and standardized strictly to guarantee the accuracy of the subsequent empirical calculation. The first stage was to systematically remove samples that lacked data on the key indicators. In the cost analysis of cloud bookkeeping, incomplete input variables, such as unreported software subscription fees, or missing output variables, like unrecorded transaction volumes, will have a severe impact on the efficiency limits of the Data Envelopment Analysis model. Next, the input and output data of each remaining sample were checked for logical consistency. This step was to verify the basic accounting logic of the data; namely, it ensured that the total cloud bookkeeping cost was the sum of the different types of expenses, and that the number of processed orders, an output index, should logically match the reported labour and time inputs.

After the data was screened for completeness and logical consistency, it was processed in a standardised way according to the operational requirements of the DEA model. One of the key technical reasons for pre-processing was to eliminate the dimensional differences among various indicators. As the input and output variables in this study use different units, such as monetary values for costs and quantitative counts for transaction volumes, direct input into the DEA model would result in mathematics meaningless due to unit incompatibility. Using standardised normalization, all data points were converted into dimensionless values, and the proportions of the original data set were kept to allow objective comparisons of efficiency among all the decision-making units.

To ensure the quality of the data, the Kruskal-Wallis test was employed to test the validity of the preprocessed data. It is necessary to use this non-parametric test to determine if there are any large changes in the selected indicator data, which will distort the efficiency frontier and thus mislead the cost-benefit analysis. After the test, it was found that the differences among the groups were not due to the presence of outliers, and thus, all the data meet the basic requirements for the operation of a DEA model. Through a series of normalisation and verification procedures, a strong basis for empirical evidence has been established, and thus, the following is to carry out an analysis of cost optimisation for small e-commerce sellers under cloud bookkeeping.

Based on the preprocessed data of 42 small e-commerce seller decision-making units (DMUs), the CCR model will calculate the total technical efficiency values for the use of cloud bookkeeping costs in these small e-commerce sellers to reduce input costs. As a basic Data Envelopment Analysis model that assumes constant returns to scale, the CCR model can determine whether a particular DMU has been allocated resources optimally based on the entire empirical frontier. The model will find a scalar efficiency score that represents the overall technical efficiency of cloud bookkeeping cost use by solving the linear programming problem for each seller.

After the calculation, all 42 DMUs will be sorted in descending order of efficiency. If the comprehensive technical efficiency value of a DMU is equal to 1, it indicates that the seller is at the efficiency frontier with no waste in inputs or shortages in outputs. A score below 1 indicates that DEA is not effective; thus, the seller might be able to produce the same quantity of output at a lower cost by reducing cloud bookkeeping costs proportionally. The number and proportion of DMUs that meet the DEA conditions can be calculated and used to judge the structural efficiency of the sample group.

The analysis of the distribution of efficiency values as a whole is for the actual situation of cost management in cloud bookkeeping for small e-commerce sellers. If many DMUs have a score below 1, it indicates that there are inefficiencies in the cost of all the sample and there is room for improvement. The dispersion of the efficiency scores shows that the operational capabilities vary; some sellers have been able to reduce their costs, but many still have problems with excessive technical redundancy in bookkeeping. Finally, the CCR model calculation has provided a solid basis to study the general situation of cloud bookkeeping cost use and shows that most small e-commerce sellers need to reduce unnecessary technical inputs to meet industry standards.

Based on the initial CCR model to calculate the overall technical efficiency, the BCC model is then used to decompose it into pure technical efficiency and scale efficiency. Pure technical efficiency separates the small e-commerce sellers from the rest of the large companies; it does not take into account their size. It is a direct indicator of whether the seller is using the cloud bookkeeping inputs (such as subscription fees, data processing labour) in a way that is most beneficial to the enterprise. The scale efficiency shows that the current input scale of cloud bookkeeping costs has reached the best matching level for output. Scale efficiency of less than 1 indicates that the seller is in the range of increasing or decreasing returns to scale, that is, its input scale is too small or too large with respect to the optimal operating boundary. The product of these two sub-indicators is the comprehensive technical efficiency, and it can be used to diagnose the reason for cost inefficiency.

To verify the validity of the model, the number of samples that have achieved pure technical efficiency and scale efficiency is calculated separately. This shows whether the main reason for poor cost control is poor management or an excessively small operating scale. Next, all the non-effective decision-making units will be divided into three categories according to the reasons for their inefficiency: the first type is inefficiency caused by pure technical efficiency, that is, poor management with a good scale; the second type is inefficiency caused by scale efficiency, good management but a small size; the third type is inefficiency caused by both reasons, that is, poor resource allocation and small size.

A differentiation-based group analysis will then be carried out for the characteristics of these sample types. For units that are only constrained by technical inefficiency, the priority will be to improve the internal processes of their operations and increase the usage rate of cloud bookkeeping features; For units with scale inefficiency, the strategy is to adjust the total input volume to achieve economies of scale. This breakdown gives small e-commerce sellers targeted and practical advice to reduce cloud bookkeeping costs.

Projection analysis is an important part of Data Envelopment Analysis (DEA), and it is used to address the problem that some decision-making units are not in an optimal state according to DEA. Based on the DEA model principle, this analysis will take the relatively good production frontier formed by the effective samples as the standard. The main goal is to determine some numerical control indexes for each non-effective sample to meet the efficiency requirements of the cloud bookkeeping cost. The slack variables of each input index and the expected shortage value of the corresponding output index are used in the projection analysis to convert the abstract efficiency score into an operational plan.

The operation of the procedure is to project the non-effective units onto the production frontier. The slack variables directly show how much can be reasonably reduced for different cloud bookkeeping cost inputs, such as software subscription fees, data storage costs and manual reconciliation labour costs, under the condition that the output remains unchanged. For the Output-oriented Projection, the expected shortage value is the specific improvement space of each output index, such as the volume of processed transactions or the accuracy of financial reports, etc., at the same time.

In practical application, projection analysis can be used to optimise the cloud bookkeeping costs of small e-commerce sellers. Small sellers often have limited resources and have not yet found all the problems in their operations. The projection results can be used to determine the exact amount of redundant costs to be cut or the exact amount of deficient outputs to be increased. Therefore, it is no longer necessary to rely on assumptions for financial management, and the broad intention to reduce costs has been translated into specific operations. Thus, the projection analysis provides data support for the next cost optimization path. It can help small e-commerce sellers allocate their scarce financial resources more reasonably, adjust their cloud bookkeeping expenses, and finally achieve the best production frontier of their industry.

According to the results of the empirical Data Envelopment Analysis (DEA) and projection analysis, grouped cloud bookkeeping cost optimisation paths that are specific to various types of non-effective small e-commerce sellers can be provided. The reason for the inefficiency of the samples in the first batch is that there are problems with the operation of the cloud resources and the volume of inputs is too small. The optimal path is to improve the efficiency of the internal data sorting of cloud bookkeeping and make the most of the free value-added functions provided by cloud bookkeeping platforms. In this way, sellers can greatly improve the output of financial decision-making support while reducing monetary expenses. The samples are not inefficient due to a lack of matching scale efficiency; the main reason is that the input scale is not appropriate. According to the specific circumstances of the enterprise, the grade of the cloud bookkeeping service package will be adjusted reasonably. Sellers who have purchased premium packages but are not using the features provided should downgrade; otherwise, they will have too much idle cost input. Basic packages cannot meet the transaction needs and therefore need to be upgraded to avoid insufficient service function input and achieve scale appropriateness. There are also a small number of samples that suffer from the dual inefficiencies of pure technology and scale. A two-stage optimisation strategy has been put forward; the first stage is to adjust the internal resource allocation structure and solve the problem of pure technical inefficiency by ensuring that all the input resources are used effectively. Once technical efficiency is achieved, the second stage is to adjust the input scale to reach scale efficiency and avoid resource waste in the transition. To make these theoretical paths more practical, a general self-inspection framework for cloud bookkeeping cost efficiency has been put forward. It can meet the needs of most small e-commerce sellers and help them independently determine what kind of inefficiency they have, systematically assess how well they are using their cloud resources now, and implement the relevant improvement measures in accordance with the industry's accounting standards and platform operating rules; thus, standardised and sustainable cost control can be achieved.

The relative efficiency of decision-making units can be evaluated by the non-parameterized method of data envelopment analysis (DEA), and its use in optimizing cloud bookkeeping costs for small e-commerce merchants has some results. In essence, the purpose of this paper is to investigate the reasons for the reduction in costs and see whether it can help to increase the amount of data that can be processed and improve the accuracy of reports. The basic idea is that the overhead of cloud bookkeeping is transformed from a fixed cost to a variable cost; thus, small e-commerce merchants can adjust their administrative department to match the operation scale. Using the DEA method to standardise the operational procedure for identifying inefficiencies in costs, it is possible to compare the performance of a single seller with that of other sellers in the same industry who are under similar regulations and operating conditions. This process separates technical inefficiency and finds specific reasons for inefficiency, such as redundant software subscription tiers or underutilised automated reconciliation functions. According to the results, sellers need to change from traditional bookkeeping to modular cloud-based bookkeeping ^[4]. Based on the above slacks, the vendor will remove costs that do not increase the financial compliance or operational transparency proportionally. The practical significance of the conclusion lies in the fact that the decrease in marginal cost will have an impact on the survival of small e-commerce companies in the market. With the DEA-optimised configuration, the cost structure of sellers will be more streamlined, they will be more flexible and will strictly follow the rules of the accounting industry without requiring specialised internal staff. Finally, it can be concluded that DEA and cloud bookkeeping have been applied to financial resources optimisation, providing a consistent, repeatable reference; thus, small e-commerce enterprises will have the strength to remain competitive in the market and maintain operations in a rapidly changing business environment.

[1]付晓, 刘强, 姚晓刚. 基于"云记账"中小企业代理记账模式内部控制研究[J]. 财讯, 2019(33): 2.

[2]苏巧. 基于财务云共享中心的代理记账模式创新研究[J]. 消费电子, 2021.

[3]凌贤姝, 谢梓钦. 财务共享云环境下小微企业财务记账系统的应用研究[J]. 电脑与电信, 2018(10): 3.

[4]韦棣晶, 金林胜, 农鸿杨. 广西自由贸易试验区小微企业云代理记账研究与实践[J]. 中小企业管理与科技, 2021, 000(017): 148-149.

[5]Otilia Dimitriu, Marian Matei. A New Paradigm for Accounting through Cloud Computing[J]. Procedia Economics and Finance, 2014.

[6]Abdalwali Lutfi. Understanding the Intention to Adopt Cloud-based Accounting Information System in Jordanian SMEs[J]. The International Journal of Digital Accounting Research, 2022.

[7]Jana Singerová. Accounting in Cloud[J]. European Financial and Accounting Journal, 2018.