Lead scoring and intelligent routing: prioritise your leads with low-code workflows

Why manual scoring fails (even if your team is good)

Almost every B2B team starts the same way: leads come in through the web form, somebody reviews them in a spreadsheet or the CRM inbox and distributes them "using judgement". It works with 20 leads a month. It stops working with 200. And at 500 it becomes the main bottleneck of your pipeline.

The symptoms are always the same:

• The criteria live in the rep's head. Everyone prioritises differently: one looks at the job title, another at the industry, another goes by "the feeling". When that person is on holiday, the criteria leave with them.
• Hot leads wait for days. A visitor who just used your pricing calculator sits in the same queue as someone who downloaded a PDF three weeks ago. And the first minutes matter: the probability of reaching a lead drops sharply as the hours pass after their request.
• Distribution is opaque and uneven. Without explicit rules, the good leads "assign themselves" to whoever opens the inbox first, and nobody can audit why an opportunity ended up where it did.
• Marketing and sales argue about lead quality without a common language: there is no shared, measurable definition of what an MQL is.

The solution is not hiring more SDRs or buying yet another prospecting tool. It is turning prioritisation into an explicit process: scoring rules anyone can read, a threshold agreed between marketing and sales, and routing that assigns every lead in seconds with a response SLA. Exactly the kind of process a low-code BPM solves without writing code.

The three scoring models (and when to use each)

A good scoring model answers two questions: does this lead fit my ideal customer? and are they showing buying intent right now? Each model answers one part:

The recommended sequence for most B2B teams:

1. Start with explicit rules (profile + behaviour): they are transparent, sales understands them and they calibrate within weeks.
2. Accumulate clean history: every routed lead with its outcome (accepted, rejected, won, lost) is future training data.
3. Add the predictive layer once you have hundreds of outcomes: AI complements the rules rather than replacing them, and the final decision always remains traceable.

How to design your scoring matrix (with an example)

The scoring matrix is the contract between marketing and sales: a table of criteria and points anyone can read and challenge. Build it in a joint 90-minute session: sales brings the signals that distinguish a good lead; marketing brings the behaviours it can measure.

A starting example out of 100 points for a B2B SaaS:

With this matrix, define three action bands:

• 60 points or more → MQL: routed to sales immediately with a first-contact SLA.
• 30–59 points → nurturing: enters a nurture sequence until their behaviour moves them up a band.
• Under 30 points → marketing pool: low-frequency educational content; no sales time consumed.

Three golden rules: always include negative points (they remove noise better than any positive criterion), add time decay (an inactive lead cannot keep its score forever) and review the matrix monthly against the leads sales accepted or rejected.

Intelligent routing: every lead to the right person, now

Scoring without routing is stopping halfway: the value appears when the right lead reaches the right rep within minutes. There are four basic assignment methods:

In practice, mature teams combine methods in a cascade: territory first, then specialisation within the territory, and load-balanced round-robin as the tie-breaker.

And the piece almost everyone forgets: SLAs and automatic escalations. Assigning does not guarantee acting. Define in the workflow itself:

• First-contact SLA per band: for example, 15 minutes for a hot MQL during business hours, 4 hours for the rest.
• Automatic reminder to the rep when half the SLA has elapsed with no logged activity.
• Reassignment or escalation to the manager if the SLA expires: the lead returns to the pool or goes up to the manager, and the reason is recorded.

Implementation with low-code workflows, step by step

Everything above is built as a single visual workflow in Dokuflex. The same platform CaixaBank uses to manage 4.2 million contracts a year, Hospital Sant Pau to save 12,000 hours annually and Mutua Terrassa to orchestrate more than 250 processes also serves something as focused as scoring and assigning leads — which is why the process scales when your volume grows. The five steps:

1. 1 · Capture. A web form from the workflow itself (or your current form connected via integrations with the CRM, email or webhooks) creates a process instance per lead, with all its data normalised from the first second.
2. 2 · Scoring rules. The scoring matrix becomes a rules node in the visual editor: conditions like "if the job title contains 'director' → +15". Changing a criterion or a threshold means editing the rule, not opening an IT ticket.
3. 3 · Enrichment and AI. An AI step classifies the free text from the form ("what problem are you trying to solve?"), infers industry and intent and adjusts the score. For borderline cases, a human review task decides before moving on: supervised AI, not a black box.
4. 4 · Automatic assignment. A decision gateway applies the cascading routing (territory → specialisation → workload) and assigns the first-contact task to the chosen rep, with the SLA running from that instant.
5. 5 · Notification and follow-up. The rep gets the alert (email, mobile or their task inbox) with full context: score, the signals that produced it and the suggested next action. If the SLA expires with no activity, the escalation fires by itself. Every step is traced so you can audit and improve the model.

Want an even faster start? The Dokuflex template gallery includes ready-to-adapt sales processes, and you can estimate the hours your team would save with the ROI calculator.

Looking at 2027: continuous scoring recalculated by AI agents

Rule-based scoring is calculated when the lead comes in and when it triggers an event. The natural next step — and where the market is heading towards 2027 — is continuous scoring: AI agents that live inside the workflow and recalculate the score of the entire pipeline the moment any signal changes.

• Permanent re-ranking: the lead sitting at number 40 in today's queue jumps to number 3 if their company announces a funding round or revisits the pricing page.
• Cooling detection: the agent alerts the rep when an active opportunity stops showing signals, before it silently dies.
• Suggested calibration: the AI proposes adjustments to the matrix ("comparison downloads predict conversion better than you are scoring them") and revenue ops decides whether to apply them.

The condition for getting there is the same we defend in every AI automation: governed agents inside the process, with every recalculation traced and a person deciding at the sensitive points. If you build your scoring on a governed workflow today, adding the agent layer tomorrow is an evolution, not a migration.

The metrics that will tell you it works

Activate the workflow and watch three metrics from week one. All three come straight out of the process itself, because every step is recorded:

• Speed-to-lead: time from lead arrival to the first real contact. It is the metric routing with SLAs attacks head on: from days to minutes. Measure it per scoring band and per rep.
• Contact rate: percentage of routed leads you actually manage to speak to. If speed-to-lead drops and contact rate does not rise, review the contact channels or assignment hours.
• MQL→SQL conversion: percentage of marketing-qualified leads sales accepts as a real opportunity. It is the matrix's thermometer: if sales rejects many MQLs, the threshold is too low or weak criteria carry too many points.

Complement them with two health signals: the percentage of leads reassigned because an SLA expired (if it grows, you have a capacity problem, not a scoring problem) and the workload distribution across reps (a very uneven split usually means badly sized territory rules). Review the whole set once a month in the marketing-sales meeting: the matrix is a living organism.

Frequently asked questions