← All blogs

How to build a sales process

How to build a sales process: which stages earn a place, how to write exit criteria two people read the same way, and how to keep it current.

A sales process is the agreed sequence of stages a deal passes through, with a written rule stating what has to be true before it leaves each one. The stages give a deal a position. Those rules are what stop two people looking at the same deal and placing it differently.

Instructions on how to build a sales process usually arrive attached to something being sold: a trademarked methodology, a training platform, an agency's proprietary system. The decisions underneath all of them are the same handful.

What is a sales process?

A stage is a position a deal holds until a stated condition is met, and that condition is the exit criterion. Between them they answer two questions about every deal in the business: where it sits, and what has to happen before it moves.

The document is short. A business selling one thing to one kind of buyer fits its process on a single side of paper, and a business selling three things to three kinds of buyer usually needs three short pages rather than one long one.

Length is a warning sign. A process running to twenty pages has become a training manual, and nobody consults a training manual while updating a deal between calls.

What does a written sales process contain?

Five things, and nothing else is required:

Everything else that gets attached to a process document belongs somewhere else. Call scripts, objection handling and discovery question sets are technique, and technique is revised far more often than a sequence is.

Does building a sales process require new software?

No. The process lives wherever the deals already live, which is usually the CRM the business is already paying for and is sometimes a spreadsheet.

Software holds the stages and enforces nothing. No tool will decide what qualified means for this business, and a tool bought to fix an undefined process inherits the definition problem and adds a migration on top of it.

One capability is worth checking for, and almost every option has it: the ability to list deals filtered by stage and sorted by how long each has been sitting there.

What is the difference between a sales process, a pipeline, a sales funnel and a methodology?

Four terms describe four different objects and get used as though they described one. The process is the rule set. Each of the others describes the current contents, the shape of the numbers, or the technique used inside a single stage.

The confusion is expensive in one direction in particular. Methodology training is the most common purchase made in response to a process problem, and it improves what happens inside a conversation while leaving the question of which conversation to have next exactly where it was.

What does a business lose without a written sales process?

Three things, and each one tends to get diagnosed as something else first.

The first quarter of a new hire is where the cost is most visible. A new salesperson asks three colleagues when a deal should move to the next stage, gets three answers, and adopts the most generous one, because a generous reading makes their pipeline look healthier in week six.

Written criteria collapse that to a single reading, which is most of the value on its own. The second effect is that the person reviewing that pipeline is reviewing the same object the new hire built, rather than translating it in their head.

What are the working parts of a sales process?

Three, and a process holds up only when all three are present. Call them the three working parts.

  1. Stages that mark a change in the buyer's position. Not a change in what the selling team has done, which is a separate thing and is always available.
  2. Exit criteria written as observable events. Something that either happened or did not, recorded somewhere a third person can check.
  3. A review that changes both when they stop matching reality. Scheduled, short, and owned by a named person.

Drop the third part and the document stays accurate for as long as the business stays still. Without the second, stage names become labels each person fills in from memory. Dropping the first is the hardest to notice, because a board built from seller activity still looks like a process and still produces a number every week.

Where do the stages come from?

From the points at which the buyer's position changes, not from the list of things the sales team does.

Writing the stage list from the activity side is the faster route and the one nearly every template takes. Contacted, demo booked, quote sent, following up. Each of those names something the seller did, and every one of them can be true while the buyer has done nothing at all.

The buyer side produces a different list, and every item on it is something the buyer conceded rather than something the seller performed:

None of those can be reached by working harder at the one before it.

How many stages does a sales process need?

As many as pass the change test, and no target number is worth adopting in advance.

The test has two halves. What changed for the buyer when the deal entered this stage, and what does the selling team do differently as a result?

A stage that cannot answer the first half is describing effort. One that answers the first and not the second is real, and needs no column of its own, because nothing is routed differently once a deal arrives in it.

Two stages producing the same next action are one stage with two names. Merging them removes an argument about where a deal sits and removes no information, since the exit criterion of the survivor can carry both conditions.

Boards grow because adding a stage is cheaper than rewriting a rule. An unusual deal arrives, it fits nowhere, and a new column absorbs it. Running the change test on a proposed stage before it is created is what stops that, and it takes about a minute.

What are the stages of a typical B2B sales process?

A common set runs: new enquiry, qualified, evaluating, committed, closed. Names vary between businesses and the underlying sequence rarely does.

New enquiry holds anything that has arrived and not been assessed. Qualified means the business has confirmed the buyer has a problem it can solve and the means to pay for solving it.

Evaluating covers the period the buyer spends deciding, which is where technical and financial scrutiny happens. Committed means the buying decision is made and only execution remains.

That list is a starting point to cut from rather than a template to adopt. Two of the five usually survive the change test unaltered, qualified and committed, because both name something the buyer did. Evaluating is where most boards hide their problems, since a deal can sit inside it for a quarter while every party stays polite and nothing is conceded.

How is a stage exit criterion written so two people read it the same way?

By naming three things in one sentence: who did something, what observably happened, and where the evidence of it now sits.

A criterion reading "the prospect is interested" fails all three. Nobody is named, interest is a state rather than an event, and no record exists that anyone could check. A criterion reading "a scoping call has been held with the person who signs off the budget, and the call notes record the problem in their own words" passes all three, and a colleague who was not on that call can confirm or deny it in under a minute.

What makes a criterion observable rather than felt?

An observable criterion describes something that either happened or did not. A felt criterion describes a reading of the situation, and two people routinely take different readings from identical facts.

Five words reliably signal a felt criterion:

What does a rewritten criterion look like?

Same intent, different construction. Each rewrite below turns a state into an event and names where the evidence sits:

Each rewrite is longer than the word it replaces, which is the trade being made. Stage names stay short because they are spoken; criteria run long because somebody has to settle an argument with them.

How many conditions can one criterion carry?

One, or two where the second is genuinely inseparable from the first.

Three conditions joined by "and" produce a stage that deals reach in pieces. A rep holding two out of three either waits, which stalls a live deal for a paperwork reason, or moves it anyway, which teaches everybody that the criteria are approximate. Both outcomes are worse than a simpler rule.

Where three conditions really are required, that is usually two stages compressed into one, and splitting them is the fix.

How to build a sales process in five steps

Five steps, in order, with most of the effort landing in the first two:

  1. Draft it alone, in one sitting, from deals that have already closed.
  2. Test it backwards against three deals that are already finished.
  3. Test it forwards against every deal currently open, with a second reader.
  4. Roll it out at a meeting, by placing live deals against it out loud.
  5. Put the first review in the calendar before the first deal is filed under the new rules.

Where does the first draft come from?

From deals that have already closed, read in sequence.

Pull the last twenty closed deals, wins and losses both, and write down what actually happened in each one in the order it happened. The stages fall out of the repetitions. A step appearing in most of the wins and few of the losses is a stage; a step appearing once is an anecdote.

Reading only the wins produces a stage list that every lost deal also passed through, which describes what the selling team does rather than what separated the outcomes. The same reading of closed deals also produces the qualification criteria the earliest stage needs, and the method for extracting them is set out in how to build an ICP scorecard from closed-won deals.

A draft written by a group arrives late and reads as a compromise between people who each pictured a different deal. One author, one sitting, then hand it to the team as an edit.

How is the draft tested against deals that already closed?

By reconstructing three finished deals against the written stages and checking that the sequence can describe what happened.

Take one clean win, one loss, and one deal that took far longer than usual. Walk each through the draft stage by stage, using only what was recorded at the time. The draft fails when a deal has to skip a stage, sit in one that does not describe it, or move backwards for a reason the criteria do not allow.

Failures here are cheap and informative. A stage no real deal ever occupied is a stage somebody imagined, and a deal that skipped three of them is usually a different sales motion that deserves its own short process rather than a set of exceptions bolted onto this one.

How is it tested against deals that are currently open?

By having two people place every open deal independently, then comparing the two lists.

Every deal both readers put in the same stage is evidence the criterion is readable. Each disagreement is a rewrite, and there are usually a handful. Work through them by asking which readings the wording permits, then narrowing it until one reading survives.

Disagreements cluster, and the clustering is the useful part. When four of five disagreements land on the same boundary, the boundary is the problem rather than the wording, and the two stages either side of it are probably one stage.

How is a process rolled out so the team uses it?

At a meeting where every current deal is placed out loud, against the written criteria, in front of everybody.

Announcing a process in a document produces agreement and no change in behaviour, because nobody has yet been asked to apply it to a deal they care about. Placing live deals against it in a room removes that gap in one sitting. The first deal somebody wants to call qualified and cannot is the moment the criteria become real.

The criteria then have to live where deals get edited. A process document in a shared drive is opened twice, once when it is published and once when somebody is arguing about it, whereas the same sentences pasted into the stage description field of the CRM are read every time a stage changes.

Somebody then has to hold it, which is a separate job from writing it. In a business with an operations function that job has a title and a person whose week is built around it. Where no such function exists, the same work gets assigned by hand instead, and what changes about ownership, capture and measurement is set out in how to build a sales process without RevOps.

What happens in the first month?

Three or four rewrites, and they are the process working rather than evidence it was drafted badly.

Real deals surface conditions a draft could not anticipate. The useful response is to change the criterion rather than to grant an exception, because an exception granted once becomes the precedent the next similar deal is judged against, and nothing about either gets written down.

Rewrites slow to almost nothing by the second month. Anything still being argued over after that is a stage boundary sitting in the wrong place.

Join the MeetIQ waitlist →

Why do sales processes stop working after they are built?

Five failure modes recur, and every one of them is visible on the deal board before it shows up in the numbers.

Which failure is hardest to see from inside?

The stage that never empties, because everything about it looks like activity.

Deals in it are being worked, calls are being logged, and the count goes up rather than down, which reads as growth on any summary view. The check takes a minute: sort that stage by how long each deal has held it, then look at the oldest quarter of the list.

The fix belongs at the stage, not at the deals inside it. A stage that fills and never drains has an exit criterion nobody can meet, or one so vague that nothing counts as meeting it, and rewriting that criterion clears the backlog faster than chasing every deal in the column.

What keeps a process current once it is running?

A scheduled review, plus a short list of events that force an unscheduled one.

A review that happens when somebody gets frustrated happens after a quarter of decisions have already been made against a stale rule. Putting it in the calendar removes the judgement call about whether it is needed yet.

Five events force a change between reviews:

Changes go in one at a time wherever the effect matters. Two revisions shipped together produce one movement in conversion and two candidate explanations, and the argument about which one caused it cannot be settled.

Stages that mark buyer commitments outlast stages that mark seller activity

Activity stages decay for a structural reason. Sending, calling and following up are always available to a seller, so a deal can travel the whole length of an activity board without the buyer having agreed to anything.

Commitment stages cannot be moved that way. A buyer has either let a technical evaluator into the conversation or has not, and no amount of follow up changes which of those is true. That is what makes a board readable by somebody who was not in the room.

The practical difference shows up in forecasting. Two deals sitting in a stage named after a seller action tell a manager the same thing about wildly different situations, because the action belongs to the seller and is identical in both.

Converting an existing board is a rewrite of criteria rather than a rewrite of names. Take each stage, ask what the buyer must have done for a deal to belong in it, and write that as the exit criterion of the stage before. The names can stay, because the team already knows them, and the criteria are what the board actually runs on.

The standard advice is to shorten the board, and shortening is downstream of this rather than a substitute for it. Five activity stages produce the same unreadable forecast as eleven. Eight commitment stages are verbose and honest, which is the better failure of the two.

What to write down first

One page, five lines, drafted from the last twenty closed deals:

  1. The stage names, in order.
  2. One exit criterion per stage, each naming an actor, an observable event, and where the evidence sits.
  3. What counts as the start of a deal.
  4. The end states other than won, with a plain meaning for each.
  5. A date in the calendar for the first review.

Then place every open deal against that page with a second reader present, and rewrite every criterion the two of them read differently. That exercise takes an hour and finds the three or four definitions that would otherwise have produced a year of numbers nobody can reproduce.

None of it requires a purchase, and it gets deferred for years in businesses that could finish it in an afternoon, because the cost of not having it is paid in small amounts spread across every deal rather than in one visible failure.

The test of whether it worked is narrow. Two people, given the same deal and the same page, put it in the same column for the same stated reason. A business that can do that can say which of its stages leaks, and one that cannot is reading a number assembled from private definitions and calling it a pipeline.

Know which account deserves the next hour, not just which stage it sits in

MeetIQ joins the signals a business already generates across website, email, calendar, calls and ads, scores them against the accounts that close, and drafts the message while the window is still open.

Related