CRM settings: predictive lead scoring, rule-based assignment and lead mining

Every opportunity carries a probability. Unless someone types over it, that number is produced by a statistical model: WindoorERP counts, across your own won and lost history, how often each value of each chosen field appeared on a win versus a loss, and combines those counts.

The block shows a sentence naming the fields in use. Stage is always one of them and cannot be removed. The shipped configuration adds phone quality, e-mail quality, state, country, source, language and tags, and considers deals created in the last eight days onwards.

Two consequences worth knowing. First, the model learns from your data, so on a young database with few closed deals the numbers mean very little. Second, the moment anyone types a probability by hand on a deal, that deal stops following the model — its value is pinned until someone clears it.

Update Probabilities needs administration rights, and the check is made twice. The button is hidden from anyone without the Access Rights administration group, and the wizard tests the same right again when Update is pressed. That second test is silent: reach the dialog by any other route — a saved link, a shortcut — and pressing Update writes nothing, recomputes nothing and shows no error.

There is a scheduled job that would do the same work nightly, but it ships switched off. Until an administrator enables it, the only thing that refreshes the scores is this button.

These controls write directly onto the underlying scheduled job. Like the scoring job, it ships inactive: switching Running to Repeatedly and saving is what activates it.

All three are paid Odoo IAP services and consume credits: one credit per company found, and one more per contact. A mining request may ask for at most 200 leads and at most 5 contacts per company. Without credits the request simply ends in an error state with the reason Insufficient Credits — the screens themselves give no hint beforehand that they are metered.

A request is a saved search, not a one-off dialog. It carries the number of leads wanted, whether to fetch companies alone or companies with their contacts, whether the result should be leads or opportunities, and the sales team, salesperson and tags to stamp on whatever comes back.

The targeting filters are company size, countries, states and industries; when contacts are requested, a preferred role, other roles or a seniority. The request keeps a link to everything it generated, so a request that ran badly can be traced and its results removed.