Sales lead recommendation system for business-to-business companies

Abstract

The present invention relates to a sales lead recommendation system for B2B companies, characterized by comprising: a data collection unit that collects corporate data including at least one of images, text, video, and sound source; a data processing unit that extracts necessary data from said corporate data and tokenizes said extracted necessary data; a scoring unit that calculates a matching suitability score between companies using the data tokenized by said data processing unit; a potential customer recommendation unit that recommends potential customers using the score calculated by said scoring unit; and an evaluation unit that scores the results of the matching process between customers recommended by said potential customer recommendation unit and for whom the matching process has proceeded.

Description

Technology Field

[0001] The present invention relates to a sales lead recommendation system for enterprises, and more specifically, to a sales lead recommendation system for B2B enterprises that improves the potential for customer acquisition by utilizing various data, such as customer data and social media data, to analyze and recommend sales leads (potential customers) who are highly likely to be interested in the company's products or services. Background Technology

[0003] Traditional B2B companies primarily utilized past experience or manually collected customer data to identify sales leads, and this approach relied on the personal judgment and experience of sales representatives, resulting in limited data-driven decision-making.

[0004] In addition, as data analysis was performed manually, it was difficult to quickly identify the latest market changes and the purchasing intentions of sales leads due to information asymmetry or discretion among each person in charge. Consequently, the efficiency and accuracy of sales activities gradually declined, eventually leading to inefficiency in time and cost.

[0005] To address this, many companies have adopted next-generation technologies such as digital marketing, CRM, and big data analytics. By utilizing digital marketing tools to collect various customer behavioral data—including website visit history and email open rates—and systematically managing customer histories based on CRM, it has become possible to formulate data-driven sales strategies. However, since many of these tools rely on individual data sources and present inherent difficulties in integrating and analyzing them, there have still been limitations in accurately identifying sales leads.

[0006] In particular, the current situation makes it difficult to consider the efficiency of corporate sales targeting to be high, as it fails to reflect in real-time the rapidly changing directional trends of data—such as social media, industry-specific data, and trend data—driven by advancements in the Internet and smartphones.

[0007] To overcome these limitations, active research has recently been conducted on advanced sales lead recommendation systems that combine big data analysis and AI-based machine learning technologies. It has become crucial to analyze sales lead activities and market trends from various sources, such as news, social media, and industry reports, using machine learning and natural language processing (NLP) technologies, and to automatically identify and predict potential customers highly likely to be interested in specific products or services based on this analysis.

[0008] However, artificial intelligence models and machine learning technologies for identifying and predicting potential customers commonly used models released in the existing market.

[0009] Therefore, research is required on a sales lead recommendation system for B2B companies that maximizes the efficiency of sales activities through an automated recommendation system and higher accuracy than other companies by applying proprietary algorithms and formulas based on various data to AI-based recommendations. Prior art literature

[0011] Korean Registered Patent No. 10-2576725 The problem to be solved

[0012] The present invention aims to recommend customers with a higher probability of success and a more objective outcome by collecting and processing corporate data, quantifying it, and reflecting it in a matching score.

[0013] In addition, the purpose is to improve sales efficiency by enhancing recommendation accuracy through the matching process based on recommended customers, by collecting additional evaluation scores from customers who have met with them, and reflecting these scores in subsequent customer recommendation scores for updates. means of solving the problem

[0015] A sales lead recommendation system for B2B companies according to one embodiment of the present invention may include a data collection unit that collects corporate data including at least one of images, text, video, and sound source; a data processing unit that extracts necessary data from the corporate data and tokenizes the extracted necessary data; a scoring unit that calculates a matching suitability score between companies using the data tokenized by the data processing unit; a potential customer recommendation unit that recommends potential customers using the score calculated by the scoring unit; and an evaluation unit that scores the matching process results between customers recommended by the potential customer recommendation unit and for whom a matching process has been conducted.

[0016] Additionally, the data processing unit may include a text tokenization unit that converts the corporate data into text and tokenizes the converted text data into units that can be analyzed; a topic analysis unit that clusters the text tokens generated by the text tokenization unit by applying a pre-configured artificial intelligence model to the text tokens generated by the text tokenization unit; a similarity analysis unit that vectorizes the clustered text tokens and calculates similarity between vectors; and an indexing unit that generates and stores an index corresponding to the similarity between vectors analyzed by the similarity analysis unit.

[0017] In addition, the scoring unit uses the vector similarity analyzed by the similarity analysis unit and the previously collected email transmission and reception history data to determine the matching suitability (MF score ) can be calculated according to [Mathematical Formula 1] below.

[0018] [Mathematical Formula 1]

[0019]

[0020] (Here, MC n The number of email checks, MS n·all is the total number of emails sent, RT a is the actual email reply time, RT s is the standard email reply time, W a·n is the number of positive words, W n·n The number of negative words, W n·all Eun is the total word count, US is the urgency score, V s is vector similarity, TA c is the cumulative transaction amount with the company, TA total ) refers to the total accumulated transaction amount

[0021] In addition, the evaluation unit can calculate the matching evaluation score according to the following [Equation 2] by collecting the meeting time, whether price inquiry was made, whether a contract was established, satisfaction score, and whether the next meeting will be held from the customer who performed the meeting in response to the matching process.

[0022] [Mathematical Formula 2]

[0023]

[0024] (Here, MT a is the actual meeting time, EMT av is the average meeting estimated time, PI v is the price inquiry output value, W PI is the price inquiry weight, CC is the contract formation status value, EV is the expected contract amount, W EV is the expected contract amount weight, CP is the expected contract period, W CP is the expected contract period weight, SS sum is the sum of the satisfaction scores of the companies being met, SS diff is the difference in satisfaction scores of the meeting target companies, SS max is the maximum satisfaction score, MS is whether a next meeting is scheduled, M t·a is the actual remaining days until the next meeting, M t·av represents the average number of days remaining until the next meeting)

[0025] In addition, the evaluation unit transmits the matching evaluation score to the potential customer recommendation unit, and the potential customer recommendation unit can update the potential customer recommendation ranking by summing the matching suitability calculated by the scoring unit and the matching evaluation score. Effects of the invention

[0027] According to the present invention, by collecting and processing corporate data, quantifying it, and reflecting it in a matching score, it is possible to recommend customers with a more objective and higher probability of success.

[0028] In addition, by collecting evaluation scores from customers who have met according to the matching process based on recommended customers, and reflecting these scores in subsequent customer recommendation scores for updates, recommendation accuracy can be improved, thereby enhancing sales efficiency. Brief explanation of the drawing

[0030] FIG. 1 is a block diagram illustrating a sales lead recommendation system for B2B companies according to an embodiment of the present invention. FIG. 2 is a diagram illustrating an intermediate block diagram of a data processing unit within a sales lead recommendation system for B2B companies according to an embodiment of the present invention. Specific details for implementing the invention

[0031] Specific details regarding the problem to be solved, the means for solving the problem, and the effects of the invention as described above are included in the embodiments and drawings to be described below. The advantages and features of the present invention, and the methods for achieving them, will become clear by referring to the embodiments described below in detail together with the accompanying drawings.

[0032] The scope of the present invention is not limited to the embodiments described below, and various modifications can be made by those skilled in the art within the scope of the technical essence of the present invention.

[0033] Hereinafter, the sales lead recommendation system for B2B companies according to the present invention will be described in detail with reference to the attached FIGS. 1 and 2.

[0034] First, FIG. 1 is a block diagram illustrating a sales lead recommendation system for B2B companies according to an embodiment of the present invention, and FIG. 2 is an intermediate block diagram illustrating a data processing unit within a sales lead recommendation system for B2B companies according to an embodiment of the present invention.

[0035] Referring to FIG. 1, a sales lead recommendation system for B2B companies according to one embodiment of the present invention may include a data collection unit (110), a data processing unit (120), a scoring unit (130), a potential customer recommendation unit (140), and an evaluation unit (150).

[0036] The above data collection unit (110) can collect corporate data including at least one of images, text, video and sound source.

[0037] Here, the above corporate data refers to data collected using at least one of the following methods: government and public institution databases, third-party data providers, industry associations, chambers of commerce, corporate websites, social media, public web crawling, blockchain transaction data, surveys, customer interviews, and transaction agreements, and images, text, video, and audio data included in the collected data can be collected as corporate data.

[0038] In this case, the aforementioned government and public institution databases refer to the Electronic Disclosure System (DART), the National Statistical Portal, trade statistics sites, the Korean Intellectual Property Office, etc., and third-party data providers refer to professional data collection and provision companies such as Dun & Bradstreet, PitchBook, Refinitiv, Statista, Nielsen, CB Insight, Crunchbase, etc., and industry associations and chambers of commerce may refer to corporate data such as KOTRA, the Korea International Trade Association, the Korea Electronics and Information Communication Industry Promotion Agency, and the Korea Iron and Steel Association.

[0039] In addition, the data collection unit (110) can collect data such as news, press releases, major contracts, and business announcements published through the company’s own homepage or social media via the company website, collect data on the company’s activities and customer reactions from social media such as LinkedIn, Twitter, Facebook, and Instagram, and can further collect major events, transaction information, news, press releases, industry reactions, customer feedback, and public data of related industries using public web crawling (Google, Bing, blogs, forums, AI-based web crawling, etc.).

[0040] In addition, the data collection unit (110) can collect data by collecting transaction data through transaction history stored on a blockchain network in specific industries such as logistics or finance, collecting feedback data and transaction trend data from corporate officials or customers using a survey platform (SurveyMonkey, Qualtirics, etc.), and directly receiving transaction data through a B2B transaction partnership agreement that enters into a data sharing agreement under specific conditions.

[0041] Through the above method, the data collection unit (110) collects corporate data including company profiles, business cards, meeting data, web crawling data, CRM data, transaction data, etc., and can collect detailed corporate data such as company name, business group, company size, number of employees, sales information, company location, year of establishment, legal status, transaction history, purchase frequency, purchase pattern, payment history, creditworthiness information, website visit time, visit frequency, page viewing history, SNS activity status, lead status on CRM, customer status on CRM, inbound / outbound lead, opportunity status, cold mail response rate, response time, key decision-maker (person in charge) information, personnel changes, employee career history (previous company information), number of emails sent and received, reply time, reply cycle, etc.

[0042] In addition, the data collection unit (110) classifies the subdivided corporate data items into images, text, videos, and audio sources and collects them. Among the data included in the corporate data, images such as product images, logos, marketing images, high-resolution product photos, and brand images can be converted into corporate image data through processes such as image preprocessing, labeling, OCR (Optical Character Recognition), and image embedding generation. In addition, in the case of text, text areas within corporate-related news articles, blog posts, social media posts, reviews, news, SNS, scanned documents, and scanned image files can be converted into data through processes such as OCR-based data collection, text cleaning, and preprocessing, and collected as corporate text data. In addition, in the case of video, frames can be extracted and labeled from videos such as video platforms like YouTube and Vimeo, videos uploaded to websites, product demo videos, and provided tutorial videos, and metadata including the title, description, tags, and upload date of the video can be collected as corporate video data. In addition, regarding audio data, corporate podcasts and advertising music can be collected through APIs of music platforms such as Spotify and SoundCloud, as well as business audio data such as customer consultation recordings and call center audio. Audio-to-text conversion based on ASR (Automatic Speech Recognition) technology is performed, and simultaneously, key features (frequency, volume, sentiment analysis, etc.), categorization, labeling, and audio inbedding generation are carried out using speech signal processing technology to convert the data into corporate audio data, which can then be collected.

[0043] Meanwhile, the data collection unit (110) may include a mail data collection module, a mail data processing analysis module, a customer behavior analysis module, a personal information protection module, and a data visualization module to collect mail-related data such as mail sending and receiving history, cold mail response rate, and response time.

[0044] Here, the email data collection module collects email data such as the time of sending, receipt, viewing, and reply time, which are records of email transmission and reception, by linking with the company's email server, and can verify the recipient's response by inserting a unique tracking code that tracks whether the email was checked, whether a link in the body was clicked, and the viewing time within the cold email (advertising email) sent to the customer.

[0045] The above-mentioned email data processing analysis module analyzes whether customers open and reply to cold emails to calculate the response rate, and evaluates customer interest by calculating the average time customers reply after sending the email; furthermore, by recording the email sending time simultaneously with a timestamp, it can further analyze and store the time periods when quick responses occurred.

[0046] The above user behavior analysis module estimates customer interest based on the response rate and response time of customers who received cold emails, and prioritizes the selection of customers with high interest who show a high response rate and quick response within a certain period, and performs the selection of customers with high interest by applying the cold email response rates collected from the prioritized customers to a pre-configured machine learning model.

[0047] Here, the aforementioned machine learning model may refer to a binary classification model such as logistic regression, random forest, or support vector machine (SVM) that classifies a customer as responding when they receive a cold mail and click a link, or as not responding when they do not check the mail; a regression model that quantifies interest based on response rates using linear regression, random forest regression, or XGBoost regression models when predicting the response rate of the recipient as a continuous score; and a learning to rank model that sorts the quantified interest into ranks.

[0048] The above personal information module can encrypt and store data sensitive to personal information, such as email sending and receiving history and response times, and control data access to ensure compliance with legal requirements by restricting access rights to sensitive information.

[0049] The above data visualization module provides a dashboard format that visualizes the email response status of companies or customers for easy viewing, derives a list of highly interested customers, and can simultaneously express key indicators such as response rates and response times for each customer in graphs and charts.

[0050] Through this, the data collection unit (110) can perform a systematic classification process of data related to the company according to the types of data collected from various channels.

[0051] The data processing unit (120) can extract necessary data from the corporate data and tokenize the extracted necessary data.

[0052] Here, the above-mentioned necessary data may refer to core data representing the characteristics of a company or customer among the image, text, video, and audio data extracted from the company data collected by the data collection unit (110).

[0053] More specifically, the aforementioned necessary data may refer to corporate profile information defining the corporate identity, such as the company's size, location, industry, major products, and services, as well as interaction records, which are information regarding the potential for activating relationships with the company or customer, such as the email and contact information of key representatives of the company or customer, the frequency of interactions via email and means of contact, and response times. In addition, it may further include transaction history data, which includes information on external factors such as market conditions and competitor information to which the company or customer belongs, as well as data on past transaction history, contract terms, and performance.

[0054] Here, the data processing unit (120) is explained in more detail with reference to FIG. 2.

[0055] Referring to FIG. 2, the data processing unit (120) may include a text tokenization unit (121), a topic analysis unit (122), a similarity analysis unit (123), and an indexing unit (124).

[0056] The above text tokenization unit (121) can convert the above corporate data into text and tokenize the converted text data into units that can be analyzed.

[0057] More specifically, the text tokenization unit (121) extracts text corresponding to the data type classified as necessary data among the corporate data collected by the data collection unit (110), which is classified into images, text, video, and audio. In the case of images, text within the image is extracted using OCR technology; in the case of voice data, speech is converted into text using an Automatic Speech Recognition (ASR) model; and in the case of video, the voice is converted into text or subtitle files are extracted and stored as text data. In this way, unnecessary information can be removed by performing preprocessing such as removing special characters and unnecessary spaces within the extracted text, normalizing to standard terms, and removing stop words that are frequently used but have no meaning. Subsequently, the preprocessed text data is separated into sentence units, word units, and word units to perform morphological analysis that retains only meaningful words. If necessary, n-grams, which are groups of n consecutive words, are generated to analyze the association between words, and a word embedding model is applied to vectorize the meaning and relationships of the words. In addition, the meaning and relationship of vectorized words are quantified and utilized for data analysis, and the generated vectors are stored in a database provided within the data collection unit (110), and metadata such as the source, date of creation, and date of update of each text can be further collected and stored.

[0058] The above topic analysis unit (122) can cluster by topic by applying a preset artificial intelligence model to the text tokens generated by the above text tokenization unit (121).

[0059] More specifically, the topic analysis unit (122) can identify semantic similarity between text tokens tokenized by the text tokenization unit (121) and cluster texts that share similar topics into groups. To perform this, the topic analysis unit (122) can convert each text token into a vector form using a word embedding model such as Word2Vec or BERT, place words with similar meanings close to each other in a vector space, and apply a clustering algorithm such as K-means clustering or DBSCAN based on the converted vector data to divide the text vectors into multiple clusters according to similarity, and the clusters can be formed around a specific topic or keyword.

[0060] The above similarity analysis unit (123) can vectorize clustered text tokens and calculate similarity between vectors.

[0061] More specifically, the similarity analysis unit (123) may perform a vectorization process in which text tokens within each cluster analyzed by topic by the topic analysis unit (122) are converted into real-valued vectors using a fixed-dimensional embedding model such as Word2Vec, GloVe, or FastText, and if necessary, context-based embedding models such as BERT or GPT are additionally applied to generate vectors that reflect the fact that words, which are text tokens, may have different meanings depending on the context. Subsequently, similarity is calculated using cosine similarity, which is the directional difference between the vectorized tokens, and it can be determined that the higher the value, the more similar the meaning between the two text tokens. At this time, depending on the case, methods such as calculating Euclidean distance or Manhattan distance, which measure the actual distance between vectors, are further utilized, and such methods can calculate similarity based on physical distance in vector space. Vector similarity, which is the similarity between vectors calculated through the above method, can be used for identifying relationships between clusters, topic-based search, and related token recommendation.

[0062] The indexing unit (124) can generate and store an index corresponding to the similarity between vectors analyzed by the similarity analysis unit (123).

[0063] More specifically, the indexing unit (124) is a process of structuring vectorized text tokens from the subject analysis unit (122) and similarity analysis unit (123) so that they can be efficiently used for searching and comparison. A vector index is constructed using a vector search library or database such as FAISS, Annoy, or HNSWlib. The vector index can be stored along with the subject containing the vectors, the text token, and related metadata (original document ID, creation date, etc.) by applying an Approximate Nearest Neighbor (ANN) algorithm to quickly find adjacent vectors in a high-dimensional space. At this time, the metadata is utilized for understanding the meaning of the vector and filtering during searching, and can induce fast searching and consistent result return based on similarity between vectors. In addition, the indexing unit (124) can provide results close to real-time search and maximize data utilization in analysis, search, and recommendation systems by quickly finding the vector most similar to the vector of the input text through the index when searching for text related to a specific topic in similarity-based search, recommendation, and topic-based text analysis.

[0064] Referring again to FIG. 1, the scoring unit (130) can calculate a matching suitability score between companies using the data tokenized by the data processing unit (120).

[0065] More specifically, the scoring unit (130) uses the vector similarity analyzed by the similarity analysis unit (123) and the previously collected mail transmission and reception history data to determine the matching suitability (MF). score ) can be calculated according to [Mathematical Formula 1] below.

[0066] [Mathematical Formula 1]

[0067]

[0068] (Here, MC n The number of email checks, MS n·all is the total number of emails sent, RT a is the actual email reply time, RT s is the standard email reply time, W a·n is the number of positive words, W n·n The number of negative words, W n·all Eun is the total word count, US is the urgency score, V s is vector similarity, TA c is the cumulative transaction amount with the company, TA total ) refers to the total accumulated transaction amount

[0069] At this time, the above matching fitness (MF) score ) may mean a numerical score representing the matching suitability between the subject company or individual business owner and each company or customer collected by the data collection unit (110).

[0070] In addition, the above vector similarity (V s ) is the matching suitability (MF) with the aforementioned multiple companies or customers. score Text tokens collected and processed from the corporate data of the company or customer that is the subject of the output, and the above matching suitability (MF score ) refers to the vector similarity with text tokens collected and processed from each company or customer subject to calculation, and the said vector similarity (V s ) can be calculated in the similarity analysis unit (123) above.

[0071] In addition, the previously collected email transmission and reception history data is the matching suitability (MF). score The number of times a target company or customer has checked an email (MC) that has sent general mail, cold mail, promotional mail, etc., sent at least once in conjunction with the email of the entity company or customer calculating ) n), total number of emails accumulated by sending to each customer (MS n·all It means ) and may mean mail data collected by the data collection unit (110).

[0072] Here, the actual email reply time (RT) mentioned above a ) is the above matching goodness (MF score It refers to the time from the time of sending to the time of reply for each email sent by the company calculating it. The above definition of a reply can be defined as a reply email if the content of the previously sent email is included via the reply function for the delivered email and the email subject includes 're'. Furthermore, the difference between the time the email was sent and the reply time is the actual email reply time (RT). a It is calculated as ), and the maximum time of the above actual email reply time (RTa) is the above standard email reply time (RT s It cannot exceed ) and if it does, the above standard email reply time (RT s It can be calculated as the same value as ).

[0073] For example, the standard email reply time (RT s If ) is 24 hours and the actual email reply time (RTa) is 72 hours, the above standard email reply time (RT s The above matching fitness (MF) as a 24-hour period score It can be reflected in ).

[0074] Meanwhile, regarding the actual email reply time (RTa) mentioned above, if the reply history exceeds one time, the average of the reply times is the matching fitness (MF) mentioned above. score ) can be reflected in the output.

[0075] In addition, the above-mentioned standard email reply time (RT) s) is a preset standard time for email replies, which can be set to 3 to 6 hours for regular emails, 24 hours for cold emails and promotional emails, etc., and the above matching suitability (MF) score It can be freely set in accordance with the industry and nature of the entity producing ).

[0076] In addition, the number of positive words (W a·n ), number of negative words (W n·n ) and urgency score (US) are the above matching goodness of fit (MF score Matching suitability (MF) for the entity company calculating ) score The number of positive, negative, and urgency words included in a single email replied to by a target company or customer can be calculated. Here, the positive, negative, and urgency words are determined by classifying and matching the text included in the email with a previously collected word dictionary using natural language processing techniques, counting words such as 'good', 'excellent', and 'thankful' as positive words, and words such as 'bad', 'problem', and 'disappointment' as negative words, and counting the total number of words (W n·all ) may refer to the total number of words included in the email. In this case, urgent words refer to words such as 'urgent', 'now', 'immediately', 'at once', etc., and the urgent score (US) can be calculated by assigning a score corresponding to a pre-set number of urgent words.

[0077] Here, the aforementioned collected word dictionary is a set of predefined words for natural language processing and text analysis, which provides criteria necessary to extract specific meanings or emotions from email text, message content, or other documents, and may include words classified into various categories such as positive words, negative words, and urgency words.

[0078] In addition, the aforementioned collected word dictionary can update new terms and emotional expressions through continuous updates while excluding stop words that are meaningless in semantic analysis.

[0079] Meanwhile, if there is no reply email, the above matching suitability (MF) score ) parameter including the number of positive words, the number of negative words, and the total number of words ((W an -W n·n ) / W n·all ) and urgency score (US) can be calculated as 0.

[0080] For example, the above urgency score (US) is set to 5 when the number of pre-stored urgency words is 1, 10 when it is 2, 20 when it is 3, etc., and the above matching fitness (MF) score It can be freely changed depending on the nature of the subject company producing ).

[0081] In addition, if the above-mentioned replied email exceeds one, the average of the number of positive words, negative words, total words, and urgency words is the above matching fit (MF score It can be reflected in and calculated.

[0082] Cumulative transaction amount (TA) with the aforementioned company c ) is the matching suitability (MF) among the transaction data collected by the data collection unit (110). score It refers to the total cumulative transaction amount between the subject company and the target company that calculates ), and the above total cumulative transaction amount (TA total ) is the above matching goodness (MF score It refers to the total sum of amounts transacted with companies or customers by the subject company calculating ) using the above system (100), and the above matching suitability (MF) score Cumulative transaction amount (TA) at the time of calculating ) c ) and total cumulative transaction amount (TA total It can reflect ).

[0083] That is, the above matching fitness (MF)score ) can calculate and quantify the company or customer with the highest degree of suitability with the subject company or customer.

[0084] For example, the above matching fit (MF score ) refers to, when there are subject companies A and B, customer C, etc., in the case of the above companies A and B, the total number of emails sent from company A to company B (MS n·all ) is 50 times, and Company B checks the number of actual emails out of 50 emails (MC n ) 5 times, set standard email reply time (RT s ) is set to 5 hours, and the number of positive words in one reply email (W a·n ) is 10, number of negative words (W n·n ) 3, total number of words (W n·all When ) is 100 and the urgency score (US) is 0.8 points per urgency word, 1 urgency word is included and 0.8, the vector similarity (V) between the corporate data collected from the above-mentioned Company A and the corporate data of the above-mentioned Company B s ) is 0.9, goodness of fit (MF score The cumulative transaction amount (TA) between Company A and Company B based on the time when ) is calculated c ) is 100,000 won, and Company A's total cumulative transaction amount with all customers (TA total If ) is 1,000,000 won, the matching suitability (MF) between Company A and Company B score ) can be calculated as 2.37(5 / 50+(1-3 / 5)+(10-3) / 100+0.8+0.9+100000 / 1000000).

[0085] Through this process, the scoring unit (130) uses corporate data, email transmission and reception history data, and transaction data between the subject company or customer and the company or customer to be matched to determine the matching suitability (MF score ) can be calculated and quantified.

[0086] In addition, the scoring unit (130) has a matching suitability (MF score In response to the output, customer email responsiveness and interaction data can be systematically analyzed to quantify customer value, emotional state, and commercial opportunities.

[0087] The above potential customer recommendation unit (140) can recommend potential customers using the score calculated by the above scoring unit (130).

[0088] Here, the potential customer recommendation unit (140) is the matching suitability (MF) calculated by the scoring unit (130). score Priority is given to recommending companies or customers with high ), but the above matching suitability (MF) score When calculating ), more weight can be given to the setting values ​​of parameters that the subject company prioritizes.

[0089] For example, the above matching fit (MF score If the entity calculating ) expresses an intention to assign high scores to the number of positive and negative words included in the email content, the word count parameter ((W) in the above [Mathematical Formula 1] an -W n·n ) / W n·all Assign weight 3 to )(3*(W an -W n·n ) / W n·all ) can.

[0090] Meanwhile, the above potential customer recommendation unit (140) responds to a request from a matching entity or customer, and the matching suitability (MF) score You can recommend target companies or customers that fall within a specific percentage range, rather than in a high order such as 80-90% based on percentages.

[0091] The evaluation unit (150) can score the matching process results between customers recommended by the potential customer recommendation unit (140) and for whom the matching process has been conducted.

[0092] Here, the matching process performs matching between the matching entity company or customer and the company or customer recommended by the potential customer recommendation unit (140) based on a high score calculated by the scoring unit (130), mutually provides email addresses, company phone numbers, representative phone numbers, etc., and can generate a meeting schedule by receiving actual meeting schedule plans from each company or customer. Additionally, the matching process can receive satisfaction scores from the meeting parties company or customer after the meeting.

[0093] Meanwhile, the evaluation unit (150) can calculate the matching evaluation score according to the following [Equation 2] by collecting meeting time, price inquiry status, contract establishment status, satisfaction score, and next meeting status from the company or customer that performed the meeting in response to the matching process.

[0094] [Mathematical Formula 2]

[0095]

[0096] (Here, MT a is the actual meeting time, EMT av is the average meeting estimated time, PI v is the price inquiry output value, W PI is the price inquiry weight, CC is the contract formation status value, EV is the expected contract amount, W EV is the expected contract amount weight, CP is the expected contract period, W CP is the expected contract period weight, SS sum is the sum of the satisfaction scores of the companies being met, SS diff is the difference in satisfaction scores of the meeting target companies, SS max is the maximum satisfaction score, MS is whether a next meeting is scheduled, M t·a is the actual remaining days until the next meeting, M t·av represents the average number of days remaining until the next meeting)

[0097] Here, the above actual meeting time (MT) a) refers to the actual time spent meeting between a company or customer matched through the matching process, and may refer to the average of the meeting times entered by each company or customer into the matching process after the meeting.

[0098] In addition, the above average meeting estimated time (EMT av ) may refer to the average of the estimated meeting times entered by the company or customer matched through the matching process before the actual meeting.

[0099] The above price inquiry output value (PI) v ) is calculated as 0.5 if no price discussion took place between the company or customer who conducted an actual meeting through the above matching process, 1.2 if a general discussion involving verbal or approximate amounts was conducted, and 3 if a specific discussion, such as the sending of a quotation or provision in writing, was conducted, and the above price inquiry weighting (W PI ) can generally be set to 2. Here, the output value of the price inquiry (PI) v ) requires the company or customer conducting the meeting to input one of the following: not discussed, general discussion, or specific discussion; however, it is recognized only if the entered items are identical, and if they are not identical, it may be calculated as 0.5 (not discussed). In addition, the above price inquiry output value (PI) v ) and price inquiry weighting (W PI The value can be freely changed and applied depending on the preferences of the matching recommendation entity or the customer.

[0100] The above Contract Establishment Status (CC) is calculated as 1 if the contract establishment status matches as 'established' after the meeting between the two matched companies or customers through the matching process, and as 0 if it does not form or the responses do not match. The above Expected Contract Amount (EV) may refer to the difference between the target amount of the company or customer acting as the matching entity and the actual contract amount concluded with the matched company or customer. Additionally, the above Expected Contract Period (W CP) refers to the difference between the target contract period of the matching entity (company or customer) and the actual contract period concluded with the matched company or customer, and the above expected contract amount weighting (W EV ) and expected contract period weight (W CP ) is set to an initial value of 2, and the value can be freely changed according to the preferences of the subject company or customer.

[0101] The sum of satisfaction scores (SSsum) of the aforementioned meeting target company refers to the sum of satisfaction scores assigned by the matching target company or customer after conducting the meeting with the subject company or customer matched through the matching process, and may be assigned between 0 and 10 points. In addition, the maximum satisfaction score (SS max ) may mean the higher value among the satisfaction scores given between the company or customer that conducted the meeting.

[0102] The above-mentioned next meeting scheduled status (MS) is calculated as 1 if the next meeting is scheduled after the meeting is performed, and 0 if it is not scheduled, but may be calculated as 0 if the meeting scheduled status collected through the matching process does not match.

[0103] The remaining days until the next meeting (Mt·a) above refers to the remaining days from the meeting date to the nearest next meeting date when the next meeting is scheduled, and the average remaining days until the next meeting (Mt·av) above may refer to the remaining days obtained by applying a preset average value of the next meeting cycle corresponding to the corporate data of the subject company or customer.

[0104] Through this, the evaluation unit (150) uses parameters collected from the subject company or customer who performed the meeting in response to the matching process and the company or customer to be matched to obtain a matching evaluation score (ME score ) can be produced.

[0105] In addition, the evaluation unit (150) has the matching evaluation score (MEscore ) is transmitted to the potential customer recommendation unit (140), and the potential customer recommendation unit (140) transmits the matching suitability (MF) calculated by the scoring unit (130). score ) and the above matching evaluation score (ME score You can update the potential customer recommendation ranking by summing )

[0106] That is, the matching evaluation score (ME) calculated in the evaluation unit (150) above score ) the pre-calculated goodness of fit (MF score The scoring unit (130) is updated by summing the results with a preset ratio, and the potential customer recommendation unit (140) can perform potential customer recommendations again by reflecting this.

[0107] Here, the above preset ratio is set to 5:5 as an initial value, but can be freely changed to 1:9 or 9:1 depending on the matching requirements of the company or customer that is the matching entity, but the matching evaluation score (ME) in the above evaluation unit (150) score If there is a history of calculation, even a small percentage can be reflected and updated.

[0108] In addition, for recommended target companies or customers that have no meeting history and are matched through the matching process, the matching suitability (MF) calculated in the scoring unit (130) score Only ) is reflected, and whether to recommend can be determined through the above potential customer recommendation unit (140).

[0109] Through the process described above, the sales lead recommendation system for B2B companies collects corporate data through various channels, and specifically processes data collected through corporate (company) introductions and email integration. Through the scoring unit (130), it can calculate the matching suitability of the company or customer with the highest suitability to the company or customer who is the subject of the matching by reflecting items related to email transmission and reception, whether words are positive, negative, or urgent when sending and receiving emails, vector similarity of tokenized text, and the total cumulative transaction amount of the subject company or customer and the cumulative transaction amount of the matching company. Through this, recommendations for companies or customers are performed through the potential customer recommendation unit in order of high score ranking according to the calculated matching suitability. After the recommended company or customer and the subject company or customer schedule a meeting and hold a meeting through the matching process, a matching evaluation score can be calculated by reflecting satisfaction, whether there will be a next meeting, whether there will be a price inquiry, expected contract amount, expected contract period, whether there will be a next meeting, and the number of days remaining until the next meeting. Subsequently, the matching priority is updated by summing the calculated matching evaluation score and the previously calculated matching suitability at a certain ratio, enabling data-driven matching to be performed with higher accuracy.

[0110] According to one embodiment of the present invention, by collecting and processing corporate data, quantifying it, and reflecting it in a matching score, it is possible to recommend customers with a more objective and higher probability of success.

[0111] In addition, by collecting evaluation scores from customers who have met according to the matching process based on recommended customers, and reflecting these scores in subsequent customer recommendation scores for updates, recommendation accuracy can be improved, thereby enhancing sales efficiency.

[0112] As described above, although an embodiment of the present invention has been explained by limited embodiments and drawings, the embodiment of the present invention is not limited to the embodiments described above, and various modifications and variations are possible from this description by those skilled in the art to which the present invention pertains. Accordingly, an embodiment of the present invention should be understood only by the claims described below, and all equivalent or analogous variations thereof shall be considered to be within the scope of the inventive concept. Explanation of the symbols

[0114] 100: Sales Lead Recommendation System for B2B Companies 110: Data Collection Unit 120 : Data Processing Section 121 : Text Tokenization Section 122 : Subject Analysis Department 123 : Similarity Analysis Unit 124 : Indexing section 130 : Scoring Department 140 : Potential Customer Recommendation Department 150 : Evaluation Department

Claims

Claim 1 A data collection unit that collects corporate data including at least one of images, text, video, and audio; a data processing unit that extracts necessary data from the corporate data and tokenizes the extracted necessary data; a scoring unit that calculates a matching suitability score between companies using the data tokenized by the data processing unit; a potential customer recommendation unit that recommends potential customers using the score calculated by the scoring unit; and an evaluation unit that scores the matching process results between customers recommended by the potential customer recommendation unit and for whom a matching process has been conducted; wherein the data processing unit includes: a text tokenization unit that converts the corporate data into text and tokenizes the converted text data into analyzable units; a topic analysis unit that clusters by topic by applying a pre-configured artificial intelligence model to the text tokens generated by the text tokenization unit; a similarity analysis unit that vectorizes the clustered text tokens and calculates similarity between vectors; and an indexing unit that generates and stores an index corresponding to the similarity between vectors analyzed by the similarity analysis unit; and wherein the scoring unit calculates the matching suitability (MF) using the vector similarity analyzed by the similarity analysis unit and previously collected email transmission and reception history data. score A sales lead recommendation system for B2B companies characterized by calculating ) according to the following [Mathematical Formula 1]. [Mathematical Formula 1] (Here, MC n The number of email checks, MS n·all is the total number of emails sent, RT a is the actual email reply time, RT s is the standard email reply time, W a·n is the number of positive words, W n·n The number of negative words, W n·all Eun is the total word count, US is the urgency score, V s is vector similarity, TA c is the cumulative transaction amount with the company, TA total ) refers to the total accumulated transaction amount Claim 2 delete Claim 3 delete Claim 4 A sales lead recommendation system for B2B companies according to claim 1, wherein the evaluation unit collects meeting time, whether a price inquiry was made, whether a contract was concluded, satisfaction score, and whether a next meeting will be conducted from a customer who performed a meeting in response to the matching process, and calculates a matching evaluation score according to the following [Formula 2]. [Formula 2] (Here, MT a is the actual meeting time, EMT av is the average meeting estimated time, PI v is the price inquiry output value, W PI is the price inquiry weight, CC is the contract formation status value, EV is the expected contract amount, W EV is the expected contract amount weight, CP is the expected contract period, W CP is the expected contract period weight, SS sum is the sum of the satisfaction scores of the companies being met, SS diff is the difference in satisfaction scores of the meeting target companies, SS max is the maximum satisfaction score, MS is whether a next meeting is scheduled, M t·a is the actual remaining days until the next meeting, M t·av represents the average number of days remaining until the next meeting) Claim 5 A sales lead recommendation system for B2B companies according to claim 4, wherein the evaluation unit transmits the matching evaluation score to the potential customer recommendation unit, and the potential customer recommendation unit updates the potential customer recommendation ranking by summing the matching suitability calculated by the scoring unit and the matching evaluation score.

Images