Odoo Predictive Lead Scoring: How It Works, Setup and Limits
How Odoo predictive lead scoring computes win probability with naive Bayes, which variables it uses, setup steps, data needed, limits and Odoo 20 AI.
Most sales teams in Odoo see a probability figure on every opportunity, but few know where it comes from or whether to trust it. Our AI and Odoo integration services start by checking whether that number is telling your team anything useful.
Related service: Odoo implementation services — CRM-to-cash set up end to end, from lead to invoice.
This guide is for sales managers, CRM owners and Odoo admins on Odoo 19 or Odoo 20 who want to prioritize leads with data. For the wider CRM setup, pipelines and stages, start with our complete Odoo CRM guide.
What is predictive lead scoring in Odoo CRM?
Predictive lead scoring in Odoo CRM is a built-in machine-learning model that estimates how likely each open lead is to be won. Odoo documents it as always active in CRM, so every database already uses it.
As deals close, Odoo records which leads were won or lost and compares each new lead with that history. The Odoo documentation on predictive lead scoring confirms that more closed opportunities give more accurate probabilities.
The score shows as the Probability field on the lead or opportunity form. It feeds prorated revenue, which is expected revenue multiplied by probability, so it also shapes your pipeline forecast.
This post covers only scoring. If you want the full path from a won lead to a quote, order and invoice, see our Odoo lead to invoice workflow guide.
How does Odoo calculate lead probability?
Odoo calculates lead probability with a naive Bayes classifier that compares a lead's values against frequency tables of past won and lost leads. Each sales team gets its own tables, so one team's results do not distort another's.
The CRM source code describes the method in plain steps:
- Odoo works out the team's overall win rate, P(won), and loss rate, P(lost), from closed leads.
- For each variable on the lead, it looks up how often that value appeared on won leads and on lost leads.
- It multiplies P(won) by every "value given won" rate to get a win score, and does the same for a lost score.
- Probability equals the win score divided by the win score plus the lost score.
- The result is rounded and kept between 0.01% and 99.99%; a lead in a won stage shows 100%.
To avoid the "zero frequency" problem, where a value never seen on a won lead would wipe out the whole score, Odoo adds 0.1 to each count. Its own code notes that with little data the result is "not really realistic".
A worked example (illustrative numbers)
The numbers below are illustrative, not from a real database. They show why one strong variable can lift a score sharply.
| Input | Won leads | Lost leads |
|---|---|---|
| Team total | 50 of 200 (25%) | 150 of 200 (75%) |
| Source = Website | 30 of 50 (60%) | 45 of 150 (30%) |
| Reached stage "Proposition" | 50 of 50 (100%) | 30 of 150 (20%) |
| Score | 0.25 x 0.60 x 1.00 = 0.150 | 0.75 x 0.30 x 0.20 = 0.045 |
| Probability | 0.150 / 0.195 = about 77% |
Stage matters because of how Odoo counts it. A won lead increments every stage, while a lost lead increments only the stage it was lost in and the stages before it. So leads that reach later stages score higher.
Which variables does Odoo lead scoring use?
Odoo lead scoring always uses stage and sales team, and lets you switch seven optional variables on or off. Odoo's default settings turn all seven on.
| Variable | What it measures | When it helps |
|---|---|---|
| Stage | Current pipeline stage (always on) | Every pipeline |
| Team | Assigned sales team (always on) | Teams with different win rates |
| State and Country | Where the lead comes from | Regional sales or export markets |
| Phone Quality and Email Quality | Whether a usable phone and email exist | Inbound web leads of mixed quality |
| Source | Search, social, referral and so on | Teams that track UTM sources |
| Language | Language set on the lead | Multilingual markets |
| Tags | Tags on the lead | Only tags used on 50+ closed leads |
The tag rule surprises many admins. The source code ignores a tag's frequencies while fewer than 50 won or lost leads carry it, so a new tag has no effect on probability for a long time.
You cannot pick your own custom fields from the settings screen. The list comes from a fixed set of scoring fields shipped with CRM, so signals like company size or deal value need custom work.
How do you set up predictive lead scoring in Odoo?
You set up predictive lead scoring in Odoo from CRM settings in about five minutes, because the model is already running. The real work is choosing variables that match how your team sells.
- Go to CRM, then Configuration, then Settings.
- Under Predictive Lead Scoring, click Update Probabilities.
- Remove variables you do not trust, such as Language if most leads have the default value.
- Set "Consider leads created as of the" to a date after your pipeline stages and data entry became consistent.
- Click Update to save and recompute.
- Open a few opportunities and click the AI icon next to Probability to see what drives each score.
That AI icon shows the three variables pushing the probability up most and the three pulling it down most. It is the fastest sanity check on the model.
One detail to know: at install, Odoo sets the scoring start date to eight days before the install date. If you migrated years of history, move that date back or the old won and lost leads will not count.
How much data does Odoo lead scoring need?
Odoo publishes no minimum, but scoring only starts once a sales team has at least one won and one lost lead, and it gets reliable only with far more. Treat the early weeks as a warm-up.
The hard rules from the code are simple. A team with no won or no lost leads gets no computed probability. Tags need 50 closed leads each. Everything else improves gradually as counts grow.
Our own rule of thumb, which is practice guidance and not an Odoo figure, is a few hundred closed leads per sales team before you let probability drive assignment. Below that, use it as a hint, not a rule.
Honesty matters more than volume. Mark dead leads lost, with a reason, or the model sees too few losses and inflates every score. Our guide to Odoo dashboards and alerts shows how to flag stale leads before they skew results.
How does Odoo keep probabilities up to date?
Odoo updates its scoring tables every time a lead is won, lost or restored, so probabilities move without any manual step. A separate scheduled action can rebuild everything from scratch.
When you mark a lead won or lost, Odoo increments the frequency table straight away. Open leads recompute when their stage, team or a scoring variable changes, which is why probability jumps as a deal moves forward.
The full rebuild is a scheduled action called "Predictive Lead Scoring: Recompute Automated Probabilities". In the Odoo 20 source it runs daily but ships inactive. Turn it on after a big import or after changing the start date.
If a rep types a probability by hand, Odoo stops syncing it with the automated value for that lead. The automated figure keeps updating in the background, and clicking the AI icon switches the lead back to it.
How do you assign leads by probability in Odoo?
You assign leads by probability in Odoo by adding a Probability condition to a sales team's or salesperson's assignment rule. Odoo's own example is a team rule of Probability >= 20.
Turn on Rule-Based Assignment in CRM settings, then choose to run it manually or repeatedly, from every few minutes to weekly. Each team and member gets a domain, and the probability condition sits inside it.
A common pattern routes high-probability leads to senior closers and the rest to a nurture team. "Skip auto assignment" excludes a team or member.
Odoo 20 refines distribution further. Our Odoo 20 CRM release notes show that each team member can now be set to always in rotation, in rotation with a limit, or out of rotation.
For actions beyond assignment, such as alerting a manager when a big deal drops below 30%, use Odoo automation rules with the probability field as the trigger.
Odoo predictive scoring vs manual and custom scoring

Odoo's predictive scoring wins on effort and cost, while manual points scoring and custom models win on control. Most teams should start with the built-in model and add custom logic only where it fails.
| Factor | Odoo predictive lead scoring | Manual or points-based scoring |
|---|---|---|
| Setup effort | Already on; pick variables | Design and maintain rules |
| Logic | Learns from won and lost history | Your team's assumptions |
| Custom fields | Not selectable in settings | Any field via Studio or code |
| Data needed | Closed leads per team | None to start |
| Transparency | Top and bottom 3 factors per lead | Fully visible rules |
| Cost | Included in CRM, no credits | Consulting or development time |
Still choosing a CRM? Our comparisons of Odoo vs Salesforce on cost and CRM fit and Odoo vs HubSpot for marketing and attribution cover the wider picture.
What are the limits of Odoo PLS?
The main limits of Odoo PLS are a fixed variable list, no behavioral signals, and full dependence on clean pipeline data. Knowing them stops you trusting a number that has not earned it.
- Fixed variables: only stage, team and the seven shipped fields; no custom fields, revenue or company size.
- No engagement data: email opens, website visits and meeting counts are not part of the model.
- Naive assumption: variables are treated as independent, so overlapping ones like country and language can double count.
- Team silos: a new team with few closed leads borrows from a shared table and scores poorly at first.
- Tag lag: tags count only after 50 closed leads, so new campaign tags do nothing for months.
- Garbage in: forgotten open leads, skipped stages and reopened deals distort the frequencies.
- Not a 100% switch: Odoo's docs note a lead cannot be manually set to 100% probability.
Users hit these limits often. One Odoo forum thread asks why lead scoring does not calculate a probability for each lead, which usually traces back to missing won or lost history for that team.
Does Odoo 20 AI change predictive lead scoring?
No, Odoo 20 AI does not replace predictive lead scoring; the naive Bayes model still runs inside your database and uses no IAP credits. Odoo AI adds tools around it that do use credits.
The Odoo 20 AI documentation says Odoo AI runs on Odoo's In-App Purchase service, and that installing the AI app on the Standard plan triggers an upsell to the Custom plan. Odoo AI credits are listed at EUR 1 each on the Odoo AI IAP service page.
Where AI helps scoring is upstream. An AI agent in Live Chat can qualify conversations and create leads, and AI fields can fill structured values from free text. Better-filled leads give the scoring model better inputs.
Lead mining is a separate paid service. The Odoo lead mining docs say each generated lead costs one credit, and Lead Generation credits cannot be spent on other IAP services. Our Odoo 20 AI and IAP credits cost breakdown helps you budget them.
How should each industry use Odoo lead scoring?
Lead scoring pays off most where lead volume is high and win rates vary by source, region or channel. The table shows how we usually tune it by sector.
| Industry | Pain or cost driver | Odoo approach | How iVentureTeam helps |
|---|---|---|---|
| SaaS and software | Many trial and demo leads, few buyers | Source, email quality and tags | Clean web forms, route by probability |
| Manufacturing and B2B | Long cycles, few large deals | Stage and country, per-team scoring | Separate teams for export and domestic |
| Distribution | Mixed walk-in and online inquiries | Phone and email quality, source | Rotation rules for inside sales |
| Real estate | High inbound volume, low intent | Source, language, stage | Live Chat qualification before scoring |
| Professional services | Referrals win far more than ads | Source and tags | Tag discipline and lost-reason reviews |
How iVentureTeam helps with Odoo predictive lead scoring

iVentureTeam turns Odoo's built-in scoring from a number nobody trusts into a working prioritization system. We have delivered 150+ ERP projects across 35+ countries with 40+ Odoo specialists, led by a consultant with 13+ years in Odoo.
- Data audit: we find stale open leads, skipped stages and missing sources that skew your probabilities.
- Scoring setup: we pick variables and a start date per team, and rebuild the frequency tables after migrations.
- Assignment design: probability-based rules, rotation limits and alerts, delivered through our Odoo consulting services.
- Custom scoring: when the shipped variables are not enough, we add custom signals or an AI layer alongside Odoo's model.
- Odoo 20 AI: Live Chat agents and AI fields that fill lead data before scoring, with credit budgets agreed upfront.
To size a CRM or AI project, try the Odoo project estimate tool for a quick first number.
Lead scoring case study: DataTile
Lead scoring needs leads captured cleanly in CRM first, and that is what we built for DataTile Ltd. DataTile is a UK survey-data provider in SaaS and market research technology.
Before the project, DataTile had no CRM. Website inquiries and demo requests were tracked by hand, and follow-ups slipped.
We connected the Contact Us and "Try for Free" forms directly to Odoo CRM, with validation, so each inquiry creates a lead immediately. Odoo Live Chat was added sitewide as a second capture channel, alongside GA4 and Google Tag Manager tracking.
The 12-week project delivered 2 lead forms that auto-create CRM opportunities, the clean lead flow a scoring model needs. Read the DataTile Odoo website, CRM and subscriptions case study.
The bottom line on Odoo predictive lead scoring
Odoo predictive lead scoring is a sound, free starting point, but it only becomes useful once your team closes leads honestly and you tune the variables per team. Treat it as a guide until each team has real history.
Start by cleaning open leads, setting the right start date and checking the AI tooltip on live deals. Then add probability-based assignment, and only build custom scoring or AI where the built-in model clearly falls short.
Ready to make your Odoo lead scores worth trusting?

Get a scoring and pipeline review from a team that configures Odoo CRM every week. Request an Odoo lead scoring review, call +91-90235-15208, or email business@iventureteam.com.
Frequently Asked Questions
How does Odoo calculate lead probability?
+
Odoo uses a naive Bayes model. It compares the lead's stage, team and chosen variables with past won and lost leads of the same team, then divides the win score by the win plus lost score to get a percentage.
Is predictive lead scoring available in Odoo Community?
+
Yes. The scoring code sits in the standard CRM module of Odoo's open-source code base, so it runs in Community and Enterprise. Related services such as lead mining are paid IAP services.
Does Odoo lead scoring use IAP credits?
+
No. The scoring model runs inside your database on your own CRM history. Odoo AI features, lead mining and partner autocomplete are IAP services and do consume credits.
How many leads does Odoo need for accurate scoring?
+
Odoo sets no published minimum. A team needs at least one won and one lost lead before scores compute, and tags count only after 50 closed leads. In our practice, a few hundred closed leads per team gives usable results.
Can I add custom fields to Odoo predictive lead scoring?
+
Not from the settings screen. You can choose only stage, team and seven shipped variables such as source, country and tags. Custom fields need development work or a separate scoring layer.
Why is my Odoo lead probability not changing?
+
Usually a rep typed a probability by hand, which stops automatic updates for that lead. Click the AI icon next to the probability to switch back. Missing won or lost history for the team can also block scoring.
How often does Odoo update lead probabilities?
+
Scoring tables update each time a lead is won or lost, and open leads recompute when their stage, team or scoring fields change. A daily full rebuild action exists but ships inactive in Odoo 20.
Can Odoo assign leads based on probability?
+
Yes. Turn on Rule-Based Assignment in CRM settings and add a Probability condition, such as Probability >= 20, to a sales team or member domain. It can run manually or on a repeating schedule.
Does Odoo 20 AI replace predictive lead scoring?
+
No. The naive Bayes model still sets probability in Odoo 20. Odoo AI adds Live Chat agents that qualify and create leads and AI fields that fill lead data, which can improve scoring inputs.
Free AI use-case map for your Odoo
Bringing AI into your Odoo?
We map where AI can save time inside your Odoo processes, what it needs from your data and what it takes to build.
Free, with no obligation. A senior consultant replies within 12 business hours.
Free Odoo tools
All tools- Odoo Health CheckVersion, modules, findings and upgrade cost
- Odoo App SelectorWhich apps you need, by industry
- Odoo Project EstimateIndicative cost and timeline in minutes
- Odoo Implementation Cost CalculatorImplementation cost and timeline, in minutes
- Odoo Migration Cost CalculatorVersion jump or legacy ERP, priced in minutes
Continue reading
Odoo Lead to Invoice Workflow: CRM, Quote, Sales Order and Billing
Set up the Odoo lead to invoice workflow step by step: CRM leads, quotation templates, sales orders, invoicing policy, down payments and Odoo 20 changes.
Odoo 20 Sales Commission: Plans, Setup, Calculation and Payout
Odoo 20 sales commission explained: commission plans, targets vs achievements, paid-invoice and manager commissions, calculation, payout and limits.
Odoo Sales Order Approval: Thresholds, Studio Rules and Audit Trail
Odoo sales order approval for Odoo 19 and 20: Studio approval rules on Confirm, amount thresholds, purchase approval, custom modules and audit trail.
