Week 3

  1. 1.
    Default summary for Finspread. Now get a default summary for your finspread documents. Key constituents of the default summary are: 1. Transaction count 2. Name of company 3. Time period 4. Total amount of different COA categories and sub categories
    To enable this for your existing P & L and Balance sheet document types: Reset document type from settings . :)
  2. 2.
    Auto repopulate COA categories based on the edits made to COA subcategory 2. Making edits to the Finspread documents got easier. Now change the COA subcategory 2 for any line item and the COA subcategory 1 and COA category will be auto repopulated based on the COA model you are using for your extractions.
  3. 3.
    Persistent "filter" and "search" while switching between different tabs on "My documents" We have changed the way the search and filter works on your "My documents" page. Both filter and search query will now remain persistent upon switching between different tabs on this page.
  4. 4.
    New filters for activity logs Filter your logs better using the new activity filters available on your activity page.
  5. 5.
    Improved bulk download workflow Bulk download requests which are not instantly downloaded on your computer are now logged on the activity page. Once the file is ready to download, you receive an email along with the download link, also you get the downloadable link to this file in your activity logs.

Week 1

  1. 1.
    Summary for Finspread Document Finspread extraction is now even cooler, you can now decide what you want see as the summary of the document under review. To do so all you have to do is, write a simple JS code which can be inserted inside the custom code block of your document settings.
  2. 2.
    COA Mapping Validation Added a validation check while choosing the COA categorisation from document settings to map your own COA categories imported in database table. The validation check added is to verify if the file selected is correct to ensure error free mapping of the COA categories.
  3. 3.
    Vertical Split View: We have added a new vertical split view for better user experience to your spreadsheet review screens. With this view you can visualise both the cleaned sheet and the original document on your screen, similar to the horizontal view but in a vertical fashion. To make use of it, you can use the view icon at the top right corner of the screen to set the view.
  4. 4.
    Industry based COA Mapping For PNL and Balance Sheet we now have the COA model pre trained for SaaS and CRE industry, available for your use without any hassle. To make use of this new categorisation model all you have to do is enable COA custom model and choose the model for SaaS or CRE from the dropdown menu.
  5. 5.
    Rent roll default custom code Now you can choose to have a custom summary for your rent rolls. Choose anything you would like to see on the summary section of the review screen for better decision making. You can make use of this option with a simple JS code which can be inserted inside the custom code block of your document settings.


Week 49

  1. 1.
    COA: Model and Training We have a good news! You can now train your own model for the COA classification(for Profit & Loss and Balance Sheet documents) from inside your account, no engineering support would be required now. Checkout this link here for more info on how you can use this feature for your financial documents.
  2. 2.
    COA Mapping and settings support in user account.
    Now you can use your own COA classification and manage everything from your account itself(previously available with engineering support only). Which means, if the default classification provided by "Docsumo" doesn't fit into your process, it's nothing to worry about, you can change this anytime, without anyones help! :) Import your own COA classification sheet in your account and just a few more clicks and you are good to go! Checkout this link here for more info on COA mapping.
  3. 3.
    Real-time low confidence highlight removal on spreadsheet review screen. Now you do not have to click "rerun" to see the low confidence highlight removed from the cleaned sheet. Any changes/interaction made to the highlighted cell(Low confidence/Yellow highlight) will automatically result in removing the highlight for that cell.
  4. 4.
    Profit & Loss and Balance sheet APIs are now available in free our version. Profit & Loss and balance sheet API's which were previously only available in our paid/pro version, can be now used in freemium accounts as well.

Week 47

  1. 1.
    New: COA mapping and custom training in Finspread models. Happy to announce that as of today, we have added COA(Chart of accounts) mapping support to our list of features for Finspread.
    With COA mapping, you can now have your own “COA values” being intelligently auto filled by our system. All you have to do is map the COA values with our provided default values and upload the same to database tables and voila! Even if you have any COA values which cannot be mapped with the default values, no problem you will be able to use them on the review screen. You can also train a custom COA model(with engineering support for now), post which you can have your own categorisations being auto filled by the system once the training is complete
  2. 2.
    Improvement: Default COA The default COA categorisation is now even more robust and detailed with 3 levels of categorisation, trained to deliver best of the class predictions. Also, the model is now trained with 121 COA categories for Profit and Loss statements and 69 for Balance sheets. Click here to checkout the updated list of default COA categories:
  3. 3.
    Bug Fix: Document orientation correction issue Though the bug wasn’t dominant but we did notice a few cases where the file orientation was not changing in case wrong orientation with the original document. We have fixed this issue to make sure the automated orientation correction works flawlessly in all the cases.
  4. 4.
    Improvement: Extraction highlight and data referencing colour changes The colours being used originally for extraction highlight and data referencing between the original and cleaned sheet has been changed to bit lighter side to improve the readability of the document and at the same time clear any distractions or noise created due to the highlighting.
  5. 5.
    Improvement: 2 new calculated fields in rentroll The rentrolls now have 2 new calculated fields which will be intelligently calculated based on the original data fed to the system. The fields added are: - Resident Status - Unit Type
  6. 6.
    Improvement: Split document capabilities on review screen You can now split PDF documents into 2 or even more parts from the review screen. Along with the split feature, you get the capability to classify the document or delete it if it's not required and even rename the splitted parts.

Week 45

  1. 1.
    Auto Classification of spreadsheet type documents: Document auto classification engine now supports both PDF and spreadsheet type documents. Now you can train and automatically classify your spreadsheet type documents with our Auto Classification API & services.
  2. 2.
    Classification review: If auto classification engine shows low confidence around any document, the document is put for human/manual review. While reviewing these documents you can split and classify them into various types. Step 1: Click "Review"
    Step 2: Classify the document using the dropdown on the top left corner. If you wish to split(Only PDF) the document first and then classify it you can click on the scissor icon from the tool bar to perform the spilt and classify action on the next screen.
    Step 3: Here you can split the documents, delete a part of the document and even classify them separately into different document types.
  3. 3.
    Rename File/Folder Now you can rename a document or folder after uploading them to "Docsumo" directly from the "My Documents" page or even while splitting the document.
  4. 4.
    Data referencing between original file and cleaned sheet Validation/sanity check is now even more easier, any cell item in the cleaned sheet is now linked with the original sheet data with a simple highlighting on the original file. With this functionality you can now easily map a value between both the files How? While using the split view mode, click on a cell in the cleaned sheet and you get to see the same value in the original file with a colour highlighting. Checkout this video for a detailed demo on Finspread document spreading and data referencing:
  5. 5.
    Highlighting for Extracted data in original file Data that is extracted successfully from a document is now highlighted(Light blue shade) in the original document as well, which will make it easy for you to validate if everything is extracted properly.
  6. 6.
    Low confidence and Error highlighting for COA categories COA values which are predicted using machine learning models can be incorrect in a few cases and to figure out what needs an edit, we have brought this cell highlight feature where you will get to see 2 different colour highlighting on the cleaned sheet. 1. Yellow: Shows low confidence around the predicted value. 2. Red: Shows an error or failure case for the calculated value.
  7. 7.
    Easy edit for COA values Editing/changing the COA value in case a change is required is now super easy. For this you are required to click on the cell with the COA value, use the dropdown menu to select the value that is needed in place of the current value and you are done with it. No more typing hassles or errors!
  8. 8.
    Hide unnecessary columns in Finspread. Fields like row type and categories from the original file are now hidden in the cleaned sheet. With this change, you will only see what is absolutely needed for analysis of any document. But if you need any of these fields to enabled we can do that for your account!

Week 30

  1. 1.
    Redesigned Apis and Services Page. i. Updated look and feel, and api description.
    ii. Search functionality added to quickly find required api.
    iii. Directly find api of required category on single click.
  2. 2.
    New document types added for different type of Rent Roll processing.
  3. 3.
    Webhook events added in activity log with their status.
  4. 4.
    Knowledge Base, API Documentation options repositioned.
  5. 5.
    Updated navigation through documents. i. Document processing not completed modal updated.
    ii. Skip errored document with message on bottom right.
  6. 6.
    Contact sales from within app to update from free to pro account or request for specific service. (Free account have access to limited services and functionalities.)
  7. 7.
    Other bug fixes and extraction improvements.

Week 27

  1. 1.
    Spreadsheet Improvements. i. Added original file preview option on spreadsheet review screen. Original file can be viewed on split mode, same screen or different tab.
    ii. As of Date field added on rent roll extraction.
    iii. Enabled duplicating document type Rent Roll
    iv. Fixed issue of data auto formatting. Eg 2/2 auto changed to 02-Feb.
  2. 2.
    Onboarding flow redesigned with checklist for each steps.
  3. 3.
    New document type P and L added.
  4. 4.
    Enabled sending data on webhook for document on reviewing status.
  5. 5.
    Document setting updated. Now allows complete OCR on digital documents.
  6. 6.
    Document search updated. Documents can now be searched by 3 ways. i. Direct search by document name. The search engine tries to match search value to name of the documents.
    ii. Search based on value on any field. The search query follows pattern search:value. The search engine searches the provided value on each field of each documents. search is the keyword that specifies the engine to search document based on values.
    iii. Search based on value of a specific field. The search query follows pattern label:value. The search engine finds all the document that has field name label and value value.
  7. 7.
    Others: i. Formulas updated. Formulas like TABLEOPERATOR, CSVOPERATOR can now be used nested with other formulas. Eg: IF(TABLEOPERATOR(<<expr>>), <<cond1>>, <<cond2>>) ii. Data extraction improvements. iii. Other bug fixes.

Week 23

  1. 1.
    User Management updated. Allows document type authorization for admin users.
  2. 2.
    New document type Acord 125 added.
  3. 3.
    Email Notification added when password is reset by admin
  4. 4.
    Updated Signup/Login Pages.
  5. 5.
    Document Auto Split based on QR code added.
  6. 6.
    Formula support added on line item cells.
  7. 7.
    i. Review screen loading time improved. ii. Improvement on text extraction from logo. iii. Other bug fixes.

Week 20

  1. 1.
    Improvements on Spreadsheet Workflow.
    i. Summary Panel added for Rent Roll document type.
    ii. Updates summary on rerun validation.
    iii. Added navigation between documents in spreadsheet review screen.
    iv. Alert added when closing unsaved document or browser tab.
    v. Enabled webhook integration for spreadsheet documents.
  2. 2.
    Updated User Management System i. Shows list of all document types that any user has access to. (Admin will have access to all the document types.)
    ii. Authorize or revoke access to document types on adding or editing existing user.
    iii. Shows list of those users only who have access to document type on auto assign.
  3. 3.
    i. Improvements on Rent Roll data extraction.
    ii. Updates on custom functions.
    iii. Optimization on data extraction and rendering and other bug fixes.

Week 13

  1. 1.
    Excel file upload support added. Review screen redesigned for reviewing excel files.
  2. 2.
    New document type - Rent Roll added.
  3. 3.
    Line items updated.
    1. 1.
      Option added to add row in between existing rows.
    2. 2.
      Search functionality added on grid header.
    3. 3.
      Show/Hide empty column option added.
    4. 4.
      Added functionality to change column width.
    5. 5.
      Added option to wrap and align text.
    6. 6.
      Added border on grid to identify the selected cell.
    7. 7.
      Error icon added to help identify rows with error.
  4. 4.
    Pagination added on database table.
  5. 5.
    Download analytics report functionality added.
  6. 6.
    Controllers moved to top of the screen.
  7. 7.
    Performance improvement on making table grid on review screen.
  8. 8.
    Other fixes:
    1. 1.
      Unicode character issue on parsing digital pdf.
    2. 2.
      UI issue fix on autoclassification modal.
    3. 3.
      Performance improvements with retry mechanisms.


Week 51

  1. 1.
    TableML added: allows user to train ML model on Table data (line items).
  2. 2.
    Row number added on line items.
  3. 3.
    More options added for Expiry Time For Temporary Token and Remove Data After settings.
  4. 4.
    New formulas added.
    1. 1.
      CONVERTTOBOOL: convert any string to Boolean value.
    2. 2.
      CSVOPERATOR: use Database Table as custom code.
  5. 5.
    Bug fixes and Others:
    1. 1.
      Multiple table support on formula TABLEOPERATOR.
    2. 2.
      CSV download issue on multi-table document fixed.
    3. 3.
      UI fixes and other security updates.

Week 43

  1. 1.
    View mode added for skipped documents.
  2. 3.
    Extraction improvement for existing document types (debit settlement, 1120s, bill of lading).
  3. 4.
    1. 1.
      filter and transformation duplication issue fixed.
    2. 2.
      line item glitches issue fixed.
    3. 3.
      document loading time reduced and other bug fixes.

Week 39

  1. 1.
    New document types added.
2. Line Item section reorganized. Line Item (Table) can now be added inside section along with key value pairs as opposed to single line item in single section.
4. Pagination added on edit screen.
5. Settings page reorganized. Integrations page added. Webhook for different event can now be selected from frontend. Supports document status change along with user activity.
6. Extraction improvement for existing document types (acord28, debit settlement, bill of lading).
7. Others: - field input box removed from edit screen - bounding box not overlapping on extracted value on multipage document issue fixed - confidence and review_required mismatch issue fix on webhook and json download fixed - number when ended with dot not recognized as number issue fixed and other bug fixes

Week 31

  1. 1.
    New document type added.
2. Added advance search for numbers.
3. Webhook event for doc status change made customizable. Webhook event can now be triggered for every status change as opposed to approved documents only.
4. Password protected file upload support added via api (
5. Banner added on edit and review screen for on screen tutorial.
6. Line items redesigned. Grid and buttons redesigned. Icon added to mark header row. Help texts updated.
7. Value fields removed from edit screen
8. Account region added in settings page.
9. Error messages updated for erred documents.
10. Error messages updated on date type fields. Also support for month first format updated.
11. Data extraction improvements on acord25, flood certifications and debit settlement document types.
12. Index column removed on csv file download
13. Skip and Start Review button added on client review tool.
14. Others: - metadata updated on webhook data - support page, icons, field input box updated and other ui changes - duplicate email check made more strict and other bug fixes.

Week 24

  1. 1.
    New document types added.
2. Support for multiline values added. (Press Shift + Enter to move to a new line)
3. Pagination added on model and training page.
4. Each tables downloaded in different sheets for some document types (Financial Statement and Table Vision).
5. MAKELIST formula added to convert string values to list.
6. Username saved when doc approved using temp token, and sent on webhooks.
7. Bug fixes and others: - processing time for bank statement tables improved. - timeout error issues fixed - model training and extraction issues fixed and more.

Week 19

  1. 1.
    Model priority introduced, which enables different fields to use a different models for better extraction.
  2. 2.
    Enabled few shot learning based on exact page number match on documents.
  3. 3.
    Credit Banner added.
  4. 4.
    Region selection added on signup.
  5. 5.
    Comparison of Test and Train data for single model view (previously it only showed test data).
  6. 6.
    Field Input Box repositioned for values whose position not found on document.
  7. 7.
    Primary side bar re-organized. Schedule User Training added.
  8. 8.
    DELTADATE formula added to validation.
  9. 9.
    Bug fixes - Remaining time calculation on Models & Training and more.

Week 16

1. Model hub added. Centralized section for training different models, viewing their metrics and comparing different models.
2. Reorganized document settings. Model trained on one document type can be used on another document type.
3. Template Matching improved. 4. Combining extracted data from different models improved. 5. Filtering documents based on modified date added for api requests. 6. Added confirm popup on approve with error.
7. Bug fixes: a) Filtering documents by date when time parameter sent. b) NaN value extractions resolved. c) Documentation links updated and more.

Week 13

1. Classified API and services into various categories based on industries.
2. Added accord 28 to frontend. 3. Added databases table to maintain database in Docsumo itself. More detail: Database Table and Dropdown Mapped
4. Added new datatype as dropdown Mapped that can used with database table to auto populate the fields. More detail: Database Table and Dropdown Mapped
5. New algorithm to match document in few-short learning. Improvement results in both multipage and single-page documents. 6. Rewrite the API documentation as per new API and Services. 7. Bugs: a. Setting not sync between users when new user is created. b. Value in processed document could be change from input fields. c. Fixed no of straight through processing number bug on analytics screen and more.

Week 10

1. Major UI change in review screen to show the status of key and line item.
2. Added alert email when document status changed.
3. Updated document count based on filters applied.
4. Added functionality to parse US address. New function PARSEADDRESS. 5. Pop up removed when edit field screen is closed. 6. Added time parameter along with the date in document filter.
7. Bug fixes a. Added log for document auto assign. b. Document setting duplication issue fixed.

Week 8

1. Added support for documents deletion for more than 1K document using auto delete.
2. Added log for document deletion from review screen. 3. Updated metadata on webhook and download data to make response uniform across the APIs.
4. Improved text selection from documents during annotation. Added separation based on visual clues like lines and boxes. 5. Bug fixes: a) Document types not being copied to the new users. b) Old document type title being used in the notification email. c) Document setting not being used during auto-classification and more.

Week 5

1. Collapsible left nav bar on review screen for better visibility of line item.
2. Support for multiple headers in webhooks.
3. Added feedback email for files that are not uploaded to docusmo through email. Details- 4. Enhanced analytics: processed category split to STP approved, approved, and approved with error
5. Bug fixes: a. Start review start opening different documents. b. Split document uploading outside the folder. c. Auto-assign document to the user when no user available and more.

Week 2

1. Added Auto Classify module. Now you can choose document type and version on auto
2. New document Type Table vision added to extract table from any document. 3. New flag added to distinguish files approved with and without error. (Only in processed tab)
4. Added appropriate error messages added for corrupt and encrypted files. 5. Added the created_at_iso and modified_at_iso to CSV and JSON export. 6. Bugs fixes:- a) Negative number not extraction from table grid b) All filters not removed when x in clicked on the filter bar and more.


Week 52

1. Added dedicated activities log tab. You can now search log.
2. Added document split using a fixed number or regex pattern.
3. Added Authentication on webhooks
4. Download is now streamlined.
a) If no of the documents selected is less than 1000 documents, files are download from UI. b) If no of the documents are between 1000 and 10000, an email is sent with the link once the download is ready. c) If no of the documents are more than 10000, please use the filter and select document samples that are less than 10000. You can make batches and download one after the others.
You get a notification and message on the left bottom corner with a different message for each case.
5. Deleting the user is now streamlined with a proper assignment of the document once the user is deleted. 6. Bugs Fixes:- a) Notification to the user for the old documents is now on the document type label. b) Bugfix on notification email link so that you only see files for that user, review status, and sorted by oldest first. c) Fix issue with Advance search not showing all documents. We still have an issue, if your account has more than 30K documents. We are optimizing the query and will fix it in the next release.

Week 49

1. Added log on document label.
Currently we add log on each status change and document assignment.
2. Added feature to add admin and other members to notification emails in all accounts. 3. Financial statement type now support multiple tables on the same page. 4. Allow the user to set an expiry time for temp token.
5. Member no longer have access to expiry token time, remove data after, straight-through process, and webhooks setting. 6. Add support from the new data type i.e optical mark recognition.
7. Bug fixes: a) Added sorting using modify date. b) and many more.

Week 46

1. Allow user to make shareable link while uploading or from review screen. Detail:
2. Added rerun validation button to rerun the calculated field along with all filters on key value pair and tables.
3. Added support for custom code and validation in table. Now you can write complex code like :-
__{{integer}} is line number.
This can use to check running balance check in bank statements.
4. More custom function CLASSIFYDATA and FINDANDREPLACE. 5. All the popups are redesign and UI is uniform across the app. 6. Added feature to add admin and other member to notification emails. (Only on newly created document type right now, will be added to old account at end of week)
7. When fields are edited for document types, document that is under reviewing is take as reference so that we don't have this issue of document status changes if first document is processed. 8. Bugs fixes: a. SEARCHINLIST and IF can have values with spaces. b. Edit screen not working for document types with large number of documents.

Week 44

1. Add document setting with a ton of new features like auto-assign document, custom machine learning model, custom processing code. Detail:
General Tab
Pre Processing
Import & Export
Post Processing
2. Added support for tiff and tif files. 3. Enable Finance documents for all new accounts. 4. Add two additional options for rule base extraction.
5. Add more post processing functions.
6. Advance Search: We have added support for advance search you can use <label name>:<value>||<label name 2>:<value2>
The label name should be exact match. The label name is case sensitive. Value use regex, it matches any text contains that value. E.g 123 than all values with 123 are filtered like 12345, 4512347.

Week 41

1. Added Error Message on the frontend for KYC documents.
2. Beta Version of the Financial statements is live.
3. Added more on rule base extraction for better data extraction on forms and fix documents format.
4. allow multilevel nearby keywords on rule base extraction
5. Workflow for document upload from API.
6. Bug fixes
  1. 1.
    Landscape pdf files are sent thought Image ocr.
  2. 2.
    The document approved even when the field is empty but the field is mandatory and has validation.

Week 39

1. Added analytics screen for all document types
Accuracy and no change per document are only calculated for new documents added after 2020-09-24 5:00:00(+5:45).
2. Added drop-down and calculated fields on data types.
a. Dropdown Setting
b. Calculated field Setting
3. Change approved to strict check ( no approve if any error) and added button for Approve with an error.
The document is approved with a strict check if the value is empty and is optional. If any other error, the document will not be approved, you can bypass it using a double tick.
4. UI issue fix
5. New functions on support Custom Functions
  1. 1.
6. Bugs Fixes
  1. 1.
    Upload error on few file types
  2. 2.
    Performance improvement and more

Week 37

1. Added validation on labels. IMPORTANT:
1. The label name should be inside ' not " i.e 'Total Due' not "Total Due" 2. Currently, the error message is only re-calculated if you visit the label. E.g on the above video when we add 0.13 to total tax, Total Due was till showing an error icon but once you move to total due and click the tab, an error icon is removed as validation matches. We are working on making all fields update in realtime, i.e as soon as any fields are updated all field effected by that change is recalculated and error message are added or removed. Should be live in 2 weeks as this required major change in the codebase. 3. If you need to use greater or less invalidation, please use >== and <==. For equal, you can use ==. The documentation will be ready next week.
2. Added filter by status and uploaded by.
3. Allow downloads on skipped documents. 4. Added status page
5. Added support for rule base extraction ( need to be activated from backend for now). 6. Added support for the NLP model on the waterfall model 7. Added custom post-processing on API and webhooks. 8. Bug Fixes. a. Allow moving to the next page on the search. b. Only show 500 files in the review page so that the review screen can be open from any tab. & more ..

Week 35

1. Added type selections while adding new labels.
2. Added filter and transformations for the above three types.
3. Add the extraction tab for data extraction using the rule.
4. Added custom processing on each label.
5. Combine documentation and moved our own domain. Docsumo API Documentation
6. Allow retry on failed documents.
7. Allow users to retry and add new keys to all documents in the review.
This will remove any annotation or fix done on reviewing documents. This API just retry documents and doesn't keep track of annotated fields. Only retries document under review. If the processed document is reviewed again it will have old keys.
Happy Automation!