"How is the quarter looking?"

"Good. I think we will hit it."

That exchange happens in most B2B companies under 50 people, and it is a mood rather than a forecast. The quarter then closes 30 percent short and everybody is surprised, which is the part that should not happen. The information needed to see the gap in week three was sitting in the CRM the whole time.

You do not need a RevOps hire or a forecasting platform to fix this. You need stage definitions that mean something and your own conversion rates instead of the ones your CRM shipped with.

Why the weighted pipeline number is wrong

Open your CRM and look at the percentage sitting next to each deal stage. Proposal might say 60 percent. Negotiation might say 80 percent.

Ask where those came from. In almost every company I look at, the answer is that they came with the software. HubSpot ships defaults. Pipedrive ships defaults. Salesforce ships defaults. They are generic starting values, and hardly anyone replaces them with numbers from their own business.

So when the dashboard multiplies deal value by stage percentage and sums it into a weighted pipeline total, it produces a number that looks precise and is built on someone else's assumptions. A company that closes 30 percent of its proposals is running a forecast that assumes 60. That single gap is enough to miss a quarter while the dashboard stays green.

The math is fine. The inputs are borrowed. Replace the inputs.

Fix the stage definitions first

Before any percentage can be trusted, the stages have to mean the same thing to everyone. This is where most SMB pipelines fall apart.

Stages named Interested, Qualified, or Hot describe a feeling inside your salesperson's head. Two reps looking at the same deal will file it differently, and one rep will file it differently on a Monday than on a Friday after a good call. Conversion rates calculated on top of that are measuring mood, not progress.

Define every stage by something the buyer did. It has to be verifiable by a third party looking at the record.

  • Discovery held. A scheduled call happened with someone who has budget authority or direct access to it.
  • Requirements confirmed. The buyer has stated in writing what they need and roughly when.
  • Proposal sent. A priced document went out.
  • Decision process known. You know who signs, what else has to happen, and what date they are working toward.
  • Verbal commitment. The buyer has said yes and the remaining steps are contractual.

Notice that none of these require judgement. A deal either had the call or it did not. When stages work this way, two salespeople put the same deal in the same place, and the conversion rates you calculate on top finally describe something real.

The most useful stage on that list is decision process known, and it is the one most SMB pipelines skip. A deal where you cannot name the signer and the date is not late stage, no matter how friendly the last call was.

Calculate your real conversion rates

Export every deal from the last twelve months that reached a final outcome, won or lost. You want 30 to 50 minimum. Under about 20 and single deals swing the percentages so hard the result is noise, in which case forecast on deal count and named accounts instead and come back to this when you have the history.

For each stage, count how many deals ever entered it, then how many of those eventually closed won. Divide the second by the first.

If 80 deals reached proposal and 24 closed, your proposal-to-close rate is 30 percent. Not 60. Do this for every stage and you have a set of numbers that belong to your business. Put them into the CRM in place of the defaults, and the weighted pipeline figure becomes worth reading.

Recalculate quarterly. Rates drift as pricing, positioning, and lead sources change, and a rate calculated eighteen months ago describes a company you no longer are.

Build the forecast three ways

One number hides the risk. Report three.

Commit is what you would stake your reputation on. Only deals where the decision process is known, a real decision date exists, and the buyer has confirmed the timeline. This number should be boring and it should be right.

Best case adds everything that could realistically land if things go well. The gap between commit and best case is your risk, and it is the most informative number in the review. When commit alone covers the target, you are in good shape. When only best case reaches it, you have a problem that needs action this month rather than an explanation next quarter.

Coverage is total open pipeline divided by the target. Need $400,000 and holding $1.2 million open gives 3x coverage. The right multiple is the inverse of your own win rate, rather than a figure borrowed from an article. Close 25 percent of pipeline and you need about 4x. Close 40 percent and 2.5x is fine. Calculate yours.

The rot that breaks every forecast

Stale deals are the biggest source of forecast error in small companies, and they are entirely self-inflicted.

A deal with no buyer-side activity for 30 days is not a deal. Nobody wants to close it lost, because closing it lost makes the pipeline look smaller and someone will ask about it. So it sits there, quietly inflating coverage and the weighted total, for months.

Set an automated rule. Any deal with no buyer activity in 30 days gets flagged, and the flag forces one of three outcomes: advance it with evidence, set a specific next step with a date, or close it lost. "Following up" is not a next step. Both HubSpot and Pipedrive can run this rule natively, and the same CRM hygiene discipline that keeps contact records clean applies here.

Watch close dates too. When a deal's close date moves to the last day of the month three times running, the date was never real. Tie every close date to something the buyer said about their own timing, and write that quote in the notes. If nobody on the buying side has stated a decision date, the deal does not belong in commit.

Where marketing fits

Once the conversion rates are real, run them backwards. This is the part that turns a sales exercise into a marketing plan.

Suppose next quarter needs $400,000 in new business. Average deal size is $20,000, so that is 20 wins. If 4 percent of qualified leads become customers, you need about 500 qualified leads entering the funnel. Sales cycle is 90 days, so those leads are needed now, not next quarter.

That single calculation is more useful than any lead goal set by taking last quarter and adding 20 percent. It tells you the number, the timing, and whether the current marketing budget can realistically produce it. When the answer is no, you find out with a quarter of runway instead of at the post-mortem.

Review weekly for the current quarter and monthly for the next. Thirty minutes, covering only what changed, what went stale, and whether coverage still supports the target. A gap found in week three is fixable. The same gap found in week ten of a 90-day sales cycle is arithmetic you cannot argue with.

The takeaway

Forecasting is a definitions problem wearing a tooling problem's clothes. Define stages by what the buyer did. Calculate conversion rates from your own closed deals. Close out the deals that are already dead.

Do those three and a spreadsheet next to your CRM will out-forecast a platform running on vendor defaults. The number stops being a mood and starts being a plan.