Lead vs opportunity: how to manage both in your CRM

Lead vs opportunity vs deal

A lead is a person or company that has shown interest and may fit your offer. An opportunity is a prospect who is very likely close to buying: there is a need, a budget, a timeframe and people who decide. HubSpot calls the matching record a deal, Salesforce an opportunity. Either way, it carries the value, the stage and the expected close date.

Lead management generates leads and qualifies them. Opportunity management takes over once sales accepts a lead as a real prospect, ranks it by how likely it is to close and turns the pipeline into a forecast. One does not work without the other:

MQL vs SQL: the stages from lead to opportunity

Marketing owns the MQL, sales everything after it, and each stage needs an exit criterion the CRM can check. There is little point in handing leads to sales while they are still at the MQL stage: give every lead the time it needs to become a real opportunity.

Why sales rejects marketing leads

When marketing reports plenty of leads and sales says none of them are any good, both are usually right. They are working with different definitions. The leads in between sit idle or get called too late, and by then a buyer with a concrete need may already be talking to a competitor.

The fix is an agreement between marketing and sales, often called a service level agreement. It settles what counts as an MQL, how quickly sales responds, and what happens to a lead that sales returns. Without it, every rep reads the stages differently, which is where many sales teams start.

In 2022, Forrester put numbers on the problem: lead-centric processes built on MQLs convert less than one percent of enquiries into closed deals, while more than 80 percent of buying decisions are made by groups of more than three people. The MQL still has its place, but the goal is the buying group and the opportunity.

Lead qualification with BANT and MEDDICC

BANT (budget, authority, need, timeline) is quick and works well where purchases are smaller and decided by few people. Enterprise purchases rarely are. According to Forrester's State of Business Buying 2026, a typical buying decision involves 13 internal stakeholders and nine external influencers, and procurement is among the decision makers in 53 percent of buying cycles.

For those, MEDDICC fits better. It extends MEDDIC, developed at PTC in 1996, with a second C for competition, and asks about metrics, the economic buyer, decision criteria and process, the pain you solve, a champion on the customer's side and the competition.

Pick one model and use the same criteria, in the same CRM fields, for every lead. Three questions are worth adding whichever model you choose: how much support will the customer need, what could the deal be worth beyond the first order, and who else is the customer talking to?

Lead scoring with fit, engagement and AI

Scoring turns those criteria into a ranking along two axes. Fit asks whether the company and person match your ideal customer. Engagement asks what they have done, from website visits to replies. High on both goes to sales; high fit with low engagement goes to nurturing.

Rule-based scores are transparent but age with the behaviour they were built on. AI scoring learns from past deals which signals actually preceded a win, provided there is enough history and the team can see why a lead scored high. Under Article 21(2) GDPR, people can object to profiling for direct marketing at any time, so document your scoring and make sure an objection actually stops it.

A pipeline forecast built on criteria

A forecast is only as good as its stage definitions and its data. Each deal stage needs criteria you can check, such as "decision maker met", "requirements confirmed" or "procurement involved", instead of a probability somebody estimated. And each customer needs to exist once: when the same company lives in the sales CRM, the service system and a spreadsheet, its history and its score split in two.

Criteria won't make a forecast right. When it misses, though, you can see which stage was wrong. The figures worth watching are leading indicators: time from new lead to qualification, conversion from stage to stage, time spent in each stage, and the share of deals that run entirely in the system. The last one tells you how far to trust the others.

AI agents in lead and opportunity management

An AI agent does a task itself rather than suggesting it. Take a new enquiry from your website: the agent identifies the company, checks industry and size against your target profile, finds existing contacts in the CRM and hands the lead to the right rep with a short summary and the questions still open. The rep starts the call informed instead of from scratch. The same pattern covers:

• Research and enrichment: company data and recent news, before anyone picks up the phone.
• Outreach and follow-ups: messages within rules your team sets.
• Deal progression: reading emails and call notes and showing which criteria are confirmed and which are open.

The qualifying conversation and the close stay with people. Buyers want less of the sales rep, but not none: in a Gartner survey published in 2026, 67 percent of B2B buyers preferred a buying experience without a sales rep, yet 69 percent preferred to check AI-generated information with one.

Agents need what a good sales team needs, only more strictly: defined criteria, clean data, clear rules for what they may change, a person accountable for the result and a record of every action. Where an agent talks to prospects directly, the EU AI Act has required since August 2026 that people know they are dealing with an AI. Whether these conditions are in place is what our NATIVE framework measures before an agent goes live.

How HubSpot stores leads and opportunities

As a Diamond HubSpot Partner, we build most of these processes in HubSpot. As of September 2026, the concepts map like this:

• Lifecycle stages on contacts and companies run from subscriber through MQL and SQL to opportunity and customer. By default, a record becomes an opportunity once a deal is associated with it.
• The leads object (Sales Hub Professional and Enterprise) holds the prospecting work in its own pipeline: new, attempting, connected, qualified, disqualified. The first steps update automatically from outreach and replies.
• Lead scoring combines fit and engagement scores, with AI scoring in Marketing Hub Enterprise.
• Deals carry an AI estimate of the win probability, and deal progression (beta) works with frameworks such as BANT or MEDDIC.
• Agents such as the prospecting agent run in HubSpot's Agent Hub.

More on the platform: HubSpot Sales Hub.

Where to start

Buyers notice early whether a company knows what to do with their interest. Get the criteria right first, then the data, then scoring, then agents. An agent on an unclear process just repeats the confusion faster.

If your pipeline stages are open to interpretation, start with CRM Strategy & Process Design. Once the criteria hold, agents can take over qualification and quote preparation.