Quantis Next Best Actions — AI-Powered Sales Recommendation System
A salesperson starts the day with a list of several dozen clients. They don’t know where to begin. They check purchase history, review notes, ask colleagues — and waste time before even picking up the phone. Meanwhile, somewhere on that list is a client who has quietly stopped ordering. And another who is ready to buy something new — but no one has offered it to them.
Next Best Actions is a module of the Quantis system that solves this problem. Every day, automatically, and individually for each salesperson — it generates a personalized action queue. Who needs contact. What to offer. Why right now.
Who is Next Best Actions For
Next Best Actions was designed for two roles — and each sees the same system through a different lens.
For the Salesperson
A salesperson’s job is not to analyze data. Their job is to talk to customers. NBA shortens the time between waking up and the first valuable conversation. Instead of browsing purchase history and guessing who to call today — they open the dashboard and see a ready-made list with justification for every item.
For the Owner and Sales Manager
NBA is a management tool. It answers questions that previously required meetings, reports, or pure instinct: which client is at risk of leaving, where the untapped opportunities are, and whether salespeople are working on the right priorities.
The Problem Next Best Actions Solves
Sales opportunities disappear in silence
A client stops ordering. They don’t quit dramatically — they simply start buying less. More slowly. Less frequently. And by the time anyone notices, it’s too late for a simple conversation. You have to save the relationship instead of developing it.
At the same time, there are clients in the database who buy only a fraction of what they could. Similar companies in the same industry regularly buy a certain category. This client doesn’t. No one had time to check.
Prioritization based on intuition, not data
In most sales teams, the order of actions comes from who called last, who is considered an “important client” in the salesperson’s mind, or who has an upcoming payment deadline. This is not a bad starting point. But it is not optimization.
A system that processes purchase history, sales dynamics, and the behavior of similar clients sees things that a person managing several dozen active accounts simply cannot check in time.
Knowledge exists but is inaccessible
Data about which client is at risk of churn, what similar companies are buying, and which products fit a given customer — all of this exists in the database. Quantis calculates it. NBA turns these calculations into concrete, actionable tasks for a person who doesn’t have time for analysis.
How Next Best Actions Works — System Logic
NBA is not a simple rule like “if a client hasn’t bought in 30 days, call them.” It is a three-layer analytical engine that combines three independent perspectives into one ranked list.
Layer One — Detecting Churn Risk
The system continuously tracks every client’s behavior in each product group. When a client starts buying less than their historical pattern — the system registers a risk signal. It doesn’t wait until purchases drop to zero. It reacts to early deviations.
This is the most important signal in the system. Protecting existing revenue always takes priority over building new revenue.
Layer Two — Cross-Sell Opportunity Identification
The system compares each client with a group of similar companies in the database. It looks for product categories that those companies buy regularly — but this client doesn’t. These are portfolio gaps: areas where cooperation could be broader but isn’t yet.
Opportunities are identified at the product group level — not individual SKU. This allows the salesperson to talk about needs, not specific catalog items.
Layer Three — Specific Product Recommendation
For each identified group opportunity, the system selects specific products. It filters out ones the client already buys. From the rest, it chooses those with the highest fit score — taking into account product popularity among similar clients and the strength of the group opportunity.
The salesperson receives not only “offer something from category X” — but “specifically offer this product, because similar clients buy it regularly and this client doesn’t know it yet.”
Building the Action Queue
All three layers are combined into one action table — sorted by priority. Protecting existing clients always comes first. Then expanding cooperation. Finally, specific product proposals.
Each client receives only as many recommendations as needed to keep the list actionable — not overwhelming.
Next Best Actions Dashboard — What the Salesperson Sees
The NBA dashboard has two views. One tells you what to offer. The other tells you who to call first. Together they replace an hour of morning data review.
Product Recommendation List
For each client — suggested product, fit score, target category, market popularity, and recommendation confidence level. The salesperson sees not only what to propose, but also how strong the signal behind the suggestion is.
Sales Action Queue
A ranked list — client, what to do, priority, goal, and source of recommendation. The salesperson starts from the top. They don’t have to decide who to call first — the system has already done it.
Three types of actions on the list:
- Recover Client — client used to buy regularly, now buying less. Signal to protect the relationship before revenue visibly drops.
- Expand Cooperation — client doesn’t buy a category that similar companies buy. Opportunity to broaden the portfolio without acquiring a new customer.
- Propose Product — specific SKU recommendation. Ready starting point for a sales conversation.
Filtering and Context
The dashboard can be filtered by salesperson, region, client, or action type. Each recommendation shows where it came from — the salesperson understands the logic behind the suggestion before making the call.
What Next Best Actions Does Not Do
This is an important boundary of the system — and it’s worth stating clearly.
NBA does not send messages to clients. It does not generate offers without the salesperson’s knowledge. It does not make sales decisions. It does not replace the human relationship with the customer.
The system provides analysis and suggestions. The decision — what to say, how to lead the conversation, whether to call today or tomorrow — belongs to the person who knows the client and understands the context.
NBA eliminates guesswork. It does not eliminate the salesperson.
Integration with the Quantis Ecosystem
Next Best Actions does not work in isolation. It is one of the modules of the central Quantis system — and uses the data that Quantis calculates for the entire company.
Churn risk data comes from ERP transactional history analysis. Cross-sell opportunities are built on ML segmentation of the entire customer base. Product scores are based on demand forecasts generated by XGBoost and Facebook Prophet models.
The salesperson receives a recommendation — but behind it stands the full analytical power of the Quantis system. Not a simple filter. Not a rule based on the date of the last order. It is behavioral analysis of the entire customer portfolio.
→ Read about the Quantis system architecture
→ Quantis Logistics — Order and Inventory Management
Frequently Asked Questions
Does Next Best Actions require a separate implementation?
NBA is a module of the Quantis ecosystem — it runs on data that Quantis already calculates. Implementing the NBA module does not require building analytical infrastructure from scratch if the company is already using the Quantis system.
Does the system take the client’s industry specifics into account?
Yes. Client comparisons are based on behavioral segmentation — the system groups companies by purchasing behavior, not just by PKD code or company size. Recommendations are built on market similarity, not only categorical similarity.
What if the salesperson disagrees with a recommendation?
A recommendation is a suggestion, not an order. The salesperson knows the relationship context that the system doesn’t have — and their judgment is always final. The system does not evaluate whether the salesperson acted on it. It provides information. What they do with it is up to the human.
Does NBA support multiple salespeople at the same time?
Yes. Each salesperson receives their own personalized queue — based on their client portfolio. The manager has a cross-sectional view — they can check individual salespeople’s queues and see where the biggest opportunities or risks are in the team.
How is the effectiveness of recommendations measured?
Quantis tracks the history of recommendations and their outcomes over time. Which types of actions bring results, which products actually lead to sales after recommendation, and which clients were saved from churn. This data allows continuous improvement of the system’s logic with every new cycle.
Does the system work if the company has few clients in the database?
The more data, the more accurate the recommendations — this is true for any ML system. With smaller customer bases, the system switches to more conservative comparisons and a higher confidence threshold. There are fewer recommendations, but they are more precise.
What does cooperation with Origami Effect look like after implementation?
The project does not end when the code is launched. Every solution at Origami Effect is created with the assumption that we only build tools we ourselves would like to use every day — without compromises and without mass templates. Every system is treated with craftsman-like attention to detail, like our own venture that must work as well as possible and genuinely build company value.
Experience gained from working on the investment fund side has shaped the principle that a strong business needs constant access to the best, clean data. That is why after implementation, Origami Effect remains a long-term technical partner. The system is continuously monitored, optimized, and developed together with the company. Quantis is not a project that gets closed and put on a shelf — it is a living tool that grows and adapts to the changing physics of the market.