An audit finds the defects in your lifecycle stages and lead scoring. A rebuild replaces them: new definitions, one set of rules for who can change a stage, a scoring model tested against your own closed deals, and a handoff that someone is accountable for. If you have not diagnosed the current state yet, start with the lifecycle audit guide. This guide covers what to build, in what order, and how to judge a partner who offers to build it.
The short version: Rebuild in five parts: definitions, governance, scoring, handoff, reporting. Write definitions before configuring anything. Allow one writer per stage transition. Build separate fit and engagement scores from closed-won data, backtest them, and run the new model in shadow mode before it routes a single lead. Pick a partner by how they ask about these, not by the tool screenshots they show.
When to rebuild instead of repair
Repair when one layer is broken and the rest holds. Rebuild when several of these are true.
- Sales and marketing give different answers to "what is an MQL?"
- Many workflows, imports, and integrations write to lifecycle stage.
- The scoring model has never been compared with closed-won data.
- Reported conversion between stages does not match what reps experience.
- Custom stages were added to patch edge cases, and nobody remembers why.
- You are about to change tiers, migrate, or restructure teams.
Part 1: Definitions
Write a one-page definition for every stage before opening HubSpot. For each stage, state the entry criteria, the exit criteria, the owning team, and the system that sets it.
| Stage | Entry criteria | Owner | Set by |
|---|---|---|---|
| Lead | Known contact with enough data to market to | Marketing | Form, import, or integration |
| MQL | Meets fit and engagement thresholds | Marketing | Scoring workflow |
| SQL | Sales has accepted and begun working it | Sales | Rep action or acceptance workflow |
| Opportunity | Qualified need, budget, and timeline, with a deal created | Sales | Deal creation |
| Customer | Closed-won deal | Sales | Deal stage workflow |
Keep it close to HubSpot's default stages. Add a custom stage only when you can state the report it enables and the person who will maintain it. Separate two ideas that teams constantly blur: lifecycle stage is where a record sits in the funnel, and lead status is what sales is doing with it right now. Mixing them is how a funnel report ends up measuring rep activity.
Part 2: Governance
Governance is a short list of rules that keep the definitions true after the project ends.
- One writer per transition. Exactly one workflow or integration moves a record into each stage. Other automations may read the stage but not write it.
- Forward by default. HubSpot generally moves lifecycle stage forward only, and moving backward requires clearing the value first. Decide in advance how recycled and disqualified records are handled instead of discovering it later.
- A named owner and a change log. One person approves changes to definitions, scoring, and routing, and records each change with the date and reason.
- No silent imports. Imports and integrations that set lifecycle stage need approval, because they bypass the rules above.
Part 3: Scoring logic
HubSpot's scoring tool supports engagement scores, fit scores, and combined scores, and lets you add or subtract points per rule, apply score limits, and use decay on engagement criteria. It requires Marketing Hub or Sales Hub Professional or Enterprise, so confirm your tier first.
Build it from your data, not from a template.
- Pull the evidence. Take closed-won and closed-lost deals from the last four to six quarters and the contacts attached to them.
- Find what separates them. For each candidate attribute, such as title, company size, industry, pricing page visits, and demo requests, compare how often it appears on won deals versus lost ones. Keep the attributes with the largest difference.
- Score fit and engagement separately. Fit describes who they are. Engagement describes what they do. A perfect-fit account that has gone quiet and an excited student with no budget should not look the same.
- Keep it small. Five to eight criteria per score is usually enough. More criteria mean more maintenance and no better prediction.
- Use negative points and decay. Subtract for disqualifying signals such as competitor domains or personal email addresses. Apply decay so old activity stops counting.
- Set thresholds from conversion data. Set the MQL line where the historical conversion rate justifies sales time, not at a round number.
- Backtest. Apply the model to historical leads and check that higher bands converted at higher rates. If they did not, the model is not ready.
Part 4: Handoff design
Handoff is where most of the revenue is lost, so design it as carefully as the score.
- Routing. Define who receives each band: round-robin, territory, or named account. Include a fallback for absent or departed owners.
- Response time. Set a service-level target, such as first touch within one business day for the top band, and measure it.
- Accept or reject, with a reason. Sales must accept the lead or send it back with a required reason code. Free-text rejections cannot be reported on.
- A recycle path. Rejected and stalled leads return to nurture with a defined re-entry rule, not to a graveyard.
- Alerts and tasks. Create the task and notification at the moment of assignment so no one has to watch a list.
Part 5: Reporting
Capture a baseline before you change anything, so you can prove improvement. You need these views: conversion rate and time in stage between each pair of adjacent stages, MQL to SQL conversion by source, win rate by score band, and response time by owner. If a stage shows conversion above 95 percent or below 2 percent, it is automatic or imaginary.
Rolling it out safely
Run the new model in shadow mode for two to four weeks. Score every new lead with both models, keep routing on the old one, and compare outcomes. Cut over when the new bands predict better, then keep the old properties read-only for a quarter so you can compare. Plan a short period of close monitoring after launch. Never switch routing and scoring in the same week.
How to evaluate a partner for this work
- Ask for their lifecycle definition template and a sanitized example. The quality of the artifact tells you how they think.
- Ask how they choose scoring weights. "Industry best practice" is a red flag. "We compare your won and lost deals" is the answer you want.
- Ask whether they backtest, and what they do when the model fails the test.
- Ask what they do with existing workflows that write to lifecycle stage.
- Ask who owns the documentation at the end.
- Ask what they would not automate.
Partners worth hiring will ask you questions you cannot answer yet, such as how sales defines acceptance. That is the work. A broader checklist is in our guide to choosing a HubSpot partner for complex CRM architecture.
Three ideas worth borrowing from outside RevOps
State machines, from software engineering. Define every allowed transition between stages and every condition that triggers it. Anything not on the diagram is not allowed.
Champion and challenger, from credit risk. Banks run a new scoring model alongside the current one on live traffic before replacing it. Shadow mode is the same discipline.
Service-level clocks, from support desks. Every handoff gets a timer that starts at assignment and escalates when it expires. Without a clock, urgency is a feeling.
Frequently asked questions
- How do you rebuild HubSpot lifecycle stages?
- Write a definition for each stage with entry and exit criteria and an owner, assign exactly one workflow or integration to set each transition, then reconfigure HubSpot to match and migrate existing records into the new stages with a documented mapping.
- How should lead scoring work in HubSpot?
- Use separate fit and engagement scores built from the attributes that differ between your won and lost deals, add negative points and decay, set the MQL threshold from conversion data, and backtest before it routes leads.
- What lifecycle stages should a B2B SaaS company use?
- Start with HubSpot's defaults: subscriber, lead, MQL, SQL, opportunity, customer, and others as needed. Remove stages you do not report on, and add a custom one only when you can name the report it supports.
- How long does a lifecycle and lead scoring rebuild take?
- Typically four to eight weeks, including definitions, configuration, a shadow-mode period, and cutover. Stakeholder agreement on definitions is usually the slowest part.
- What is the difference between lifecycle stage and lead status in HubSpot?
- Lifecycle stage describes where a record is in the funnel. Lead status describes what the sales team is currently doing with it. Keeping them separate keeps funnel reporting clean.
- What should you ask a HubSpot partner before a lifecycle rebuild?
- How they derive scoring weights, whether they backtest, how they handle workflows that already write to lifecycle stage, who owns the documentation, and what they would choose not to automate.