Twelve0s: Developing an ML-Powered Financial & Sales Forecasting Software - Jappware

Twelve0s: Developing an ML-Powered Financial & Sales Forecasting Software

Jappware built a custom sales forecasting platform for Twelve0s, a US-based fintech startup. Integrated with Salesforce and Microsoft Dynamics, the platform used a machine learning engine to analyze historical CRM data and predict how deals would move through the pipeline up to six months ahead.

webSaaSDevOpsPaymentsJavaPythonReactMongoDBGCPSping FrameworkJupyterRESTSOAP
LocationUSA
IndustrySaaS, Fintech
CooperationDedicated development team
DurationJanuary 2018 – November 2020
Team Size6 engineers

About the client

Twelve0s was a San Francisco-based fintech startup that aimed to redefine the way companies forecast sales. The idea behind the product came from the founder's firsthand experience at an enterprise CRM company. While working there, he noticed that most teams still relied on spreadsheets, manual calculations, and gut feeling to estimate future sales performance.

At the same time, companies had large amounts of valuable data stored inside their CRM systems that could support more accurate sales forecasting, but very little of it was being used for that purpose. 

Twelve0s was envisioned as sales prediction software that could automatically identify patterns in CRM data and deliver accurate insights into how deals were likely to progress.

Sales & Financial Forecasting Software Development | Case Study - Jappware

A word from the client

Patrick Kellenberger

Founder & CEOTwelve0s

“The team is responsive, efficient, and understanding; they give solid recommendations and never miss a deadline. Overall, they’ve met our goals and created something special.”

Challenges

The concept behind Twelve0s was clear and compelling. Bringing it to life, however, required overcoming a distinct set of challenges.

Lack of an engineering team

Sales forecasting sits at the intersection of data science and business logic. To build the platform, the client needed a team with both deep engineering expertise and a genuine willingness to learn how sales pipelines work: how deals move through stages, what factors influence outcomes, and how buyer behavior varies across markets.

Spreadsheet-based PoC

The client's initial proof of concept was a manually maintained spreadsheet. Translating its logic into fully functional sales prediction software meant starting from scratch: defining the architecture, formalizing the forecasting logic, and engineering a system that could handle real CRM data across multiple clients at scale.

Complex CRM integrations

To pull in deal data, the platform needed to integrate with major CRM systems. But these systems were highly configurable, allowing users to create custom fields and formula-based values. Building an integration layer that could handle that level of variability required deep CRM knowledge and a careful technical approach.

Sales forecasting automation

The core of the platform was a forecasting engine. To implement it, the team needed to develop machine learning algorithms that could learn from each customer's unique CRM data, accounting for variables like deal size and sales cycle length, and produce predictions accurate enough to be useful for sales reps and business leaders.

Cooperation

With a proof of concept in hand, the client needed a financial software development partner who could turn it into a production-ready SaaS platform. Having previously worked with Jappware on another project, he trusted the team to build the product. 

Here’s how this partnership was structured:

Dedicated team model

Jappware assembled a dedicated team specifically for the project, with three backend developers, two frontend developers, and a delivery manager who also served as solution architect.

Roles and responsibilities

The client brought up product ideas, while Jappware evaluated them for technical feasibility, made architecture decisions, and engineered the solution end-to-end.

Agile delivery

The team worked in two-week sprints. Once features were discussed and approved, they were added to the backlog and developed iteratively.

Day-to-day collaboration

The team held regular weekly meetings with the client to review progress, discuss new ideas, and align on priorities.

Cooperation - Jappware

This collaborative approach allowed us to keep delivery steady and ensure the platform aligned closely with the client's vision.

Solution

As a result of this engagement, we built a custom forecasting software platform that turned complex CRM data into reliable sales and revenue predictions.

ML-powered forecasting engine

A machine learning engine analyzed the client's historical CRM data to build a custom forecasting model. The model predicted how current deals would progress through the pipeline, explaining why each deal was likely to progress or stall. It also continuously updated itself, syncing with the CRM every hour to stay accurate as the pipeline evolved.

CRM integrations

The platform integrated with Salesforce and Microsoft Dynamics via API, allowing customers to connect their CRM in minutes. With a two-way sync, any changes made in the platform were instantly reflected in the CRM. The integration layer was also built to be fault-tolerant, meaning disruptions never resulted in data loss or duplicate processing.

Reporting & analytics

Beyond deal predictions, the sales forecast software platform included a range of analytical reports covering stage transitions, sales rep effectiveness, regional performance, and more. The goal was to help sales teams identify where deals were being won, where they were stalling, and which factors influenced outcomes.

Interactive UI

Rather than presenting static charts and reports, the UI was highly interactive. Users could drag and drop deals, edit deal values, and move deals between pipeline stages directly in the platform. Any changes made in the UI synced instantly to the CRM and triggered a recalculation of the affected sales predictions.

Data storage & security

Each client's data was stored in a separate database environment and encrypted at rest and in transit, which prevented cross-account data exposure and unauthorized access. The platform also ran on a multi-node MongoDB cluster, meaning data was continuously replicated across multiple nodes and couldn’t be lost due to a single point of failure.

Monolithic architecture

Twelve0s was an early-stage startup that needed to move quickly. A monolithic approach allowed us to reduce infrastructure complexity, keep costs manageable, and ship faster. That said, the codebase was structured in layers, keeping the door open for a modular separation down the line if the product continued to scale.

Project tech stack

React - Jappware
Java - Jappware
Python - Jappware
MongoDB - Jappware
JS - Jappware
Figma - Jappware
GCP - Jappware
Wealth Asset Management Platform
Wealth Asset Management Platform

Impact

01Full product delivery

We delivered a fully functional sales and revenue forecasting software platform, which successfully onboarded its first paying customers.

02Early ML-driven sales forecasting

Built in 2020, the platform was an early real-world application of predictive ML to sales forecasting — years before AI-powered tools became mainstream.

03Complex MongoDB implementation

The team implemented a highly advanced MongoDB architecture — expertise so specialized it later became the subject of industry conference talks.

04Long-term partnership

The client came to Jappware from a previous collaboration and returned to work with us on another project after Twelve0s — a testament to the trust built throughout the engagement.

Twelve0s ultimately did not reach the commercial scale needed to sustain the business. But the value delivered throughout the engagement went well beyond the product itself.

If you are looking for an experienced full-cycle software development company to bring your idea to life, feel free to reach out to us.