New: Odoo and Dynamics 365 cost calculators.Estimate your costs

Lead Scoring in Dynamics 365 Sales and Odoo CRM: Setup, Limits and Cost for 20 to 100 Users

Dynamics 365 Sales needs 40 qualified and 40 disqualified leads for predictive scoring. Odoo scores from won and lost history. Compare setup, limits and cost.

Both platformsCRM7 min readUpdated 6 sources

Key takeaways

  • Lead scoring ranks each prospect by how likely it is to buy, using fit (who they are) and intent (what they do), so the people on your team call the best leads first.
  • Dynamics 365 Sales scores open leads with a predictive machine learning model trained on your own closed leads.
  • Odoo CRM scores leads with predictive lead scoring built on a naive Bayes model.
  • Use the platform minimums as your first filter.

What lead scoring is and which platform scores out of the box

Lead scoring ranks each prospect by how likely it is to buy, using fit (who they are) and intent (what they do), so the people on your team call the best leads first. The score can be points on a 0 to 100 scale or a win probability.

Dynamics 365 Sales offers predictive lead scoring as a trained model, with a fixed minimum of closed leads before it will build one. Odoo CRM offers predictive scoring as a win probability on every lead and opportunity, and lets you override it by hand. Neither replaces a clear definition of a qualified lead, which is why this guide starts with a rule-based model and treats predictive scoring as the second step.

Lead scoring in Dynamics 365 Sales: predictive model and rule-based build

Dynamics 365 Sales scores open leads with a predictive machine learning model trained on your own closed leads. The score appears in a Lead score column in views and in a widget on the lead form, with the top factors that raised or lowered it. Sellers can see why Lead B scores 50 while Lead A scores 80 instead of trusting a bare number.

Setup sits in the Sales Hub app under Sales Insights settings, then Predictive models, then Lead scoring. Advanced Sales Insights features must be enabled first. You choose a business process flow, the status values that mean qualified and disqualified, a filter for which leads to score, and a training window from three months to two years (the default is two years). You can keep up to 10 models, which suits separate regions, business units or product lines. Leads that abandoned the selected business process flow do not count toward training or the minimum.

The data floor is firm. Microsoft requires at least 40 qualified and 40 disqualified leads created and closed inside the training window you pick, and the Get started button stays disabled until you meet it. Newly closed leads take about four hours to reach the analysis store. You can tick Retrain automatically to refresh the model every 15 days. If the model's area under curve (AUC) score falls below the threshold, the app shows Not ready to Publish; you can still publish it, but Microsoft warns that it will perform poorly.

On licensing, Microsoft states that a Dynamics 365 Sales Enterprise license can turn on predictive scoring through the lead and opportunity scoring quick setup, with 1,500 scored records per month. For Sales Professional and for volumes above that cap, check Microsoft's plan comparison table and licensing guide before you plan, because this is where scoring budgets get surprised. Scores are written to the Predictive Score (msdyn_predictivescore) table, not to the lead row itself, which matters if you build reports or flows on the score.

The rule-based path works on any plan and needs no history. Add a whole-number Lead Score field to the lead, build a Power Automate flow that adds fit and intent points when a lead is created or changes, and publish a lead view sorted by that field. Route leads over your sales-ready threshold to an owner with the flow, using the routing rules covered in CRM automation. Rules break in three predictable ways. Weights drift out of date, a field the formula depends on goes blank, and the flow runs on records it should skip. Name an owner for the score and review the weights on a schedule.

Lead scoring in Odoo CRM: predictive probability and manual rules

Odoo CRM scores leads with predictive lead scoring built on a naive Bayes model. It uses the historical data in your own CRM to estimate the probability that an open lead or opportunity will be won. Each opportunity shows that win probability on its form, and it updates as the opportunity moves through the pipeline.

Setup is under CRM, then Configuration, then Settings, then Predictive Lead Scoring, then Update Probabilities. Stage and Team always feed the model. You can switch on any of State, Country, Phone Quality, Email Quality, Source, Language and Tags, and set the date from which leads are considered, so you can drop old data from before a process change. Leave out fields your team fills in unevenly, because a noisy field trains a noisy score.

Odoo documentation sets no minimum count of won and lost leads the way Microsoft does, so the model starts early. Quality still depends on your outcomes. A pipeline where leads are rarely marked lost gives the model nothing to contrast with the won deals, and probabilities cluster around the same value.

You can override the probability by clicking the number on a record. Odoo blocks setting it to 100% by hand. A manual value replaces automatic updates for that record, the model's value stays visible, and the AI icon next to the percentage restores automatic scoring. That makes Odoo a natural place for a hybrid: let the model rank the pool, and let a rep or manager pin a probability on a deal where they know something the data does not.

Compared with Dynamics 365 Sales, Odoo gives you one probability per record and a start-date setting rather than separate models per region. A team with very different markets can run separate pipelines so each one learns from its own history, and Flectic can set that up during discovery.

Do you have enough leads for predictive scoring?

Use the platform minimums as your first filter. In Dynamics 365 Sales the minimum is 40 qualified and 40 disqualified leads created and closed inside the training window. Odoo has no published minimum, but a model trained on a handful of outcomes ranks leads close to randomly.

Count the closed leads in your CRM before you decide. Filter to leads created and resolved in the last 12 months and split them into qualified and disqualified. If either bucket is below 40, or if reps leave most leads open rather than closing them, start with rules. Flectic's guidance, not a vendor rule, is to also require that your team marks leads disqualified with a reason, because a model learns nothing from leads that were abandoned.

Rules and prediction work together. Use rules for hard disqualifiers such as the wrong country, a competitor or a student, and let the predictive model rank the remaining pool. When you train a predictive model, train it on closed-won outcomes rather than on which leads a sales rep accepted. Models trained on rep taste copy rep bias. For Copilot-based features in Dynamics, see Sales Copilot.

Rule-based and predictive lead scoring for a 20 to 100 person team
QuestionRule-basedPredictive
What it needsAn agreed definition of a good leadDynamics 365 Sales: 40 qualified and 40 disqualified closed leads. Odoo: won and lost history
Who can explain itAnyone, every point is visibleSellers see top factors in Dynamics; Odoo shows a probability
UpkeepSomeone re-weights points on a scheduleDynamics retrains every 15 days if you tick the option; Odoo updates as outcomes close
Main riskScore inflation and stale weightsPoor data in, poor ranking out
Start here ifYou have thin history or messy lead statusesYou close and disqualify leads consistently

A starter scoring model: fit, intent and negative points

Copy the structure below, not the exact weights. Set the points from your own closed-won and closed-lost history and involve sales when you do, because a marketing-only score floods the queue and loses trust fast. Published B2B examples put senior titles around 20 to 30 points and demo or quote requests far above ebook downloads.

A practical starting scale for a small team is 0 to 100, with the MQL line near 60 and the sales-ready line near 75, matching the operating bands below. Industry guides often set the MQL line so roughly the top fifth of leads qualify, so sales capacity is not flooded. Adjust monthly against MQL to SQL conversion and win rate. Treat BANT (budget, authority, need, timing) as a checklist for the sales conversation after handoff, not as form fields at first touch.

Starter model for a 20 to 100 person B2B sales team. Adjust the weights to your ICP and win data.
SignalTypePointsWhy it matters
Owner, VP or director in a buying roleFit20 to 25Authority over budget
Company size inside your ICP bandFit15 to 20Matches the economics of your offer
Target industryFit10 to 15Firmographic fit
Customer or partner referralSource20 to 40Trust is already built
Demo, trial or quote requestIntent30 to 50Strong buying signal
Repeat pricing page visitIntent15 to 25Purchase-stage research
Comparison page or calculator useIntent10 to 15Evaluating options
Generic ebook download onlyIntent5Awareness, not readiness
Personal email for a B2B offerNegativeminus 10 to minus 15Weaker firmographic signal
Wrong industry, student or competitorNegativeminus 20 or disqualifyHard bad-fit filter
No engagement for 30 daysNegativeminus 10 or decayStale score

Build a rule-based model first

Start with 5 to 7 criteria that map to your ideal customer profile. Models with more than 15 to 20 rules become hard to maintain, so expand only when win data justifies it.

  1. Write the ICP in measurable terms

    Name industry, company size, geography and decision-maker title. Every weight is judged against this profile.

  2. Pick 3 or 4 fit criteria and 2 or 3 intent criteria

    Give the highest points to the attributes most common in your won deals, and weight buying signals such as demo requests above top-of-funnel content.

  3. Set thresholds and owners

    Define the score where a lead becomes an MQL and the score where it goes to sales. Publish both with a first-touch time for each band.

  4. Add negative points and decay

    Downscore obvious bad fits. Let intent points age, for example by a quarter each month with no new activity, or in recency bands of 30, 60 and 90 days. Fit attributes stay until the fact changes.

  5. Review monthly with sales

    Compare MQL to SQL conversion and win rate by score band, then re-weight from outcomes. Treat the first quarter as a draft.

Thresholds, decay and handoff times

A high score from a six-month-old webinar buries today's pricing page visitor, so decay intent points and keep fit points. Set your cut line by plotting six to twelve months of scores against MQL to SQL conversion and win rate, not by copying a number from a blog.

Tie each band to an owner and a clock. Research found that firms contacting a lead within an hour were about seven times more likely to qualify it than slower ones, so a score that does not start a first-touch clock is decoration. For routing rules and speed-to-lead, see CRM automation; for stage probabilities and pipeline reviews, see CRM pipeline management.

Operating bands on a 0 to 100 scale. Calibrate to your capacity and win rates.
Score bandLabelOwnerFirst touchNext action
0 to 39ColdMarketing automationNo sales targetNurture, keep out of the sales queue
40 to 59WarmMarketingOptional glance in 48 to 72 hoursTargeted content
60 to 74MQLSales developmentSame business dayQualify for the account owner
75 to 89Sales-readyNamed account ownerWithin a few hoursCall and offer a meeting
90 to 100Hand-raiserOwner and backupUnder an hourCall first, email second

Pitfalls: inflation, false negatives and distrust

Score inflation happens when too many leads clear the MQL line and conversion to SQL falls. Cut the points for weak actions and raise the line.

False negatives are won deals that scored low. Review them each quarter, because they show which signal your model underweights.

Distrust is the failure that ends scoring projects. Before you rewrite criteria, pull 90 days of MQLs and check speed to lead, touches before disqualification and whether every lead had a live owner, since routing gaps often look like bad leads. Data quality problems are covered in CRM data hygiene.

When lead scoring is worth doing at 20 to 100 people

Scoring pays when your team has more leads than time and cannot tell which to call first. If you get a few leads a week, a manager reading the list does the same job for free.

In our experience, scoring earns its setup cost in three cases: five or more reps sharing one inbound queue, lead sources with very different win rates, or a marketing team handing leads over without agreement on what ready means. In those cases a rule-based model on your existing CRM is a small project, and the predictive step comes later.

What scoring setup costs

Cost has three parts, starting with licences. Check whether your Dynamics 365 Sales plan includes predictive scoring and its monthly scored-record cap, and note that Odoo CRM scoring is configured inside the CRM settings. The second is configuration time, which is days for a rule-based model and a score field, a flow and a view. The third is the sales review that keeps the weights right, which costs hours each month rather than a project.

Flectic sends a written proposal with a fixed price for each phase before work starts. The free discovery call is 30 minutes with a senior consultant, and you leave with a recommended first phase or a clear reason to wait.

Account-level scoring for multi-contact deals

For deals with several buyers, score the account as well as the contact. Add up engagement across contacts, reward different roles taking part, and apply recency so an old thread does not outrank a live one. Do this after the lead model works, not before.

HubSpot and Salesforce comparisons

This guide stays on Dynamics 365 Sales and Odoo CRM. For feature and price comparisons with other CRMs, read Odoo CRM vs HubSpot and Odoo CRM vs Salesforce. For marketing-side scoring and segmentation in Dynamics, read the Customer Insights guide.

How Flectic can help

Flectic implements Dynamics 365 Sales for sales teams of 20 to 100 people in the US and Canada, including lead routing and predictive scoring that ranks leads and deals. Flectic sends a written proposal with a fixed price for each phase before work starts. We also implement Odoo, so if it fits your team better, we will say so. Read the Dynamics 365 Sales implementation scope, then bring last year's won and lost lead counts and Book a free discovery call.

Sources

  1. Dynamics 365 Sales predictive lead scoring requires at least 40 qualified and 40 disqualified leads created and closed in the training window (3 months to 2 years, default 2 years); up to 10 models; automatic retraining every 15 days; about four hours to sync closed leads; Not ready to Publish when AUC is below threshold.

    learn.microsoft.comChecked
  2. A Dynamics 365 Sales Enterprise license can enable predictive lead scoring in quick setup with 1,500 scored records per month; scores are written to the msdyn_predictivescore table.

    learn.microsoft.comChecked
  3. Odoo CRM predictive lead scoring uses a naive Bayes model on historical CRM data; optional variables are State, Country, Phone Quality, Email Quality, Source, Language and Tags, with Stage and Team always applied; a start date can be set; probability can be overridden manually but not set to 100%.

    odoo.comChecked
  4. Firms that tried to contact a lead within the first hour were roughly 7x more likely to qualify it than firms that waited longer.

    hbr.orgChecked
  5. Lead scoring model types include rule-based, demographic/firmographic, behavioural, predictive, and hybrid; rule-based is transparent but brittle above ~15 to 20 rules; sales distrust is a primary failure mode.

    nc-squared.comChecked NC Squared article 'Lead Scoring Models Explained' (Dec 2025) compares model types, data needs, and operational failure modes including complexity and missing feedback loops.
  6. Industry guidance: start with 5 to 7 core criteria; combine positive and negative scoring; MQL thresholds often target roughly the top ~20% of leads by score (commonly ~50 to 75 on a 100-point scale depending on model design); use time-based score decay (e.g. ~25% monthly on inactive behavioral points or recency bands 30/60/90 days).

    monday.comChecked monday.com 'Lead Scoring Rules: Prioritize Leads & Boost Sales In 2026' (Dec 2025) key takeaways and decay/threshold sections document these patterns as operating guidance, not universal constants.

Frequently asked questions

Does Dynamics 365 Sales include lead scoring?

Yes. It offers predictive lead scoring through Sales Insights, which needs Advanced Sales Insights features enabled and at least 40 qualified and 40 disqualified closed leads to train. A Sales Enterprise license can turn it on through quick setup with 1,500 scored records per month, so check your plan before you budget.

Does Odoo CRM score leads automatically?

Yes. Odoo CRM calculates a win probability for each lead and opportunity with a naive Bayes model trained on your won and lost history. You choose which variables it uses under CRM settings, and you can override a probability by hand on a single record.

How many leads do I need for predictive scoring?

Dynamics 365 Sales needs at least 40 qualified and 40 disqualified leads created and closed inside the training window you pick, which runs from three months to two years. Odoo publishes no minimum, but its model is only as good as the outcomes in your CRM. If your team rarely closes leads as lost, start with rules.

How much does it cost to set up lead scoring?

A rule-based model is days of configuration plus a monthly review with sales. Predictive scoring adds licence checks, since Dynamics 365 Sales Enterprise includes 1,500 scored records per month through quick setup. Flectic scopes the work in a written proposal with a fixed price for each phase.

Dynamics 365 or Odoo for lead scoring?

Pick the platform on the whole sales process, not on scoring alone. Dynamics 365 Sales gives more control, such as up to 10 models and a choice of training window, while Odoo is simpler and starts scoring on less history. Teams already on Microsoft 365 usually lean to Dynamics, and teams that want sales, stock and finance in one system lean to Odoo.

What is a good starter lead scoring model?

Use a 0 to 100 scale with 5 to 7 criteria: three or four fit signals from your ICP, two or three intent signals, and negative points for bad fits. Set the MQL line near 60 and the sales-ready line near 75, then adjust against your conversion data every month.

What is score decay and why use it?

Score decay lowers behavioral points when a lead goes quiet, for example by about a quarter each month or in 30, 60 and 90 day bands. It keeps last quarter's download from outranking today's pricing page visit. Fit points usually stay until the underlying fact changes.

How should sales and marketing share ownership of the score?

Agree the criteria together and give one person ownership of the weights. Review MQL to SQL conversion and win rate by band each month with both teams in the room. If sales stops trusting the score, check routing and response time before you rewrite the rules.

More on the blog: Implementation and go-live