Most sales organizations have a sales process. Far fewer have one that changes what anyone does on a Tuesday afternoon. The gap between those two states is where most of the value sits, and it is worth being precise about what a sales process actually is before arguing about whether it works.

A sales process is not a methodology, not a CRM configuration, and not a funnel diagram. Here is a working definition worth adopting: a sales process is a shared, buyer-anchored sequence of verifiable states that lets a seller diagnose where a deal actually stands and decide what to do next. Every clause of that does work. Shared, because a process that lives in one manager's head is a preference. Buyer-anchored, because the only thing that moves a deal forward is a buyer doing something. Verifiable, because if you cannot point to evidence, you are guessing. Diagnose and decide, because a process that produces a stage label but no decision is administrative overhead.

The most-cited number in this space comes from a 2013 survey of pipeline management practices conducted by Vantage Point Performance and the Sales Management Association, later written up by Jason Jordan and Robert Kelly for Harvard Business Review. Companies with a clearly defined sales process reported 18% higher revenue growth than those without. Two other practices came out ahead in the same study: managers spending more than three hours per month per rep on pipeline management (+11%) and training managers specifically in pipeline management (+9%).

That finding is directional, not conclusive, and it is worth saying so plainly. The sample was 62 respondents, the data were self-reported, and the design was correlational. It is entirely plausible that well-run companies both grow faster and happen to write down their process, with no causal arrow between the two. Anyone quoting the 18% as though it were the output of a controlled trial is overselling it.

The stronger evidence is indirect, and it comes from the academic literature on what actually predicts salesperson performance. The most comprehensive synthesis is Verbeke, Dietz and Verwaal's meta-analysis, published in the Journal of the Academy of Marketing Science in 2011, covering 268 studies, 292 samples and 79,747 salespeople across 4,317 organizations from 1982 to 2008. Five drivers emerged with meaningful standardized effects: selling-related knowledge (β = .28), degree of adaptiveness (β = .27), role ambiguity (β = -.25), cognitive aptitude (β = .23) and work engagement (β = .23). Together they explained roughly 32% of the variance in sales performance.

Two of those five are the mechanism by which a good sales process helps. Role ambiguity is the third-strongest driver and the only negative one: not knowing what you are supposed to be doing measurably degrades performance. A well-designed process is, functionally, a role ambiguity reduction device. It answers "what does good look like at this point in this deal" without a manager in the room. And selling-related knowledge, the strongest driver in the set, is exactly what a process encodes when it is built from what has actually worked rather than from a whiteboard exercise.

The third of those five, degree of adaptiveness, is the constraint. Adaptive selling and rigid process compliance pull in opposite directions if you build the process badly. A process that tells a rep which script to read at which stage suppresses the second-strongest performance driver in the literature. A process that tells a rep what the buyer needs to have accomplished, and leaves the how open, does not. This distinction is not stylistic. It is the difference between a process that helps and a process that costs you.

The single most common design failure is building stages around what the seller does. "Demo delivered." "Proposal sent." "Follow-up call completed." These are activities, and activities are trivially self-certifiable. A rep can send a proposal into a void and advance the stage.

Gartner's buying jobs model is the more useful anchor. Its research frames B2B purchasing as six jobs a buying group has to complete: problem identification ("we need to do something"), solution exploration ("what's out there"), requirements building ("what exactly do we need this to do"), supplier selection ("does this do what we want"), validation ("we think we're right but we need to be sure") and consensus creation ("we need everyone on board"). Crucially, Gartner finds buyers do not march through these in order. They loop, revisiting jobs repeatedly as new stakeholders arrive and new information lands.

The scale of the consensus problem is the part most process designs underweight. Gartner's Future of Sales work put the median enterprise software buying group at 11 stakeholders for deals above roughly $100K in annual contract value, with 2024 figures ranging from 7 in professional services to 18 in life sciences and healthcare. Forrester's 2023 buying study found committees of 14 to 23 people on deals above $1M, spanning about 6.2 distinct functional roles. A process whose late stages assume one decision maker signing one document is modeling a transaction that mostly stopped existing.

Gartner also reports that buyers spend around 17% of the total purchase journey meeting with all potential suppliers combined, and roughly 5-6% with any single rep. Whatever your process says a rep should do, it has to fit inside that budget. This is the practical case for designing the process around buyer jobs: your access is small, so it has to be spent on the job the buyer is currently stuck on rather than the activity your stage list happens to name.

Here is a framework the Sales Agent can apply to any pipeline, at any stage count. Every stage in a functioning sales process is a contract with four fields, and a stage missing any one of them will fail in a predictable way.

Field one, the buying job. Which of the buyer's jobs does this stage serve? If two adjacent stages serve the same job, you have one stage wearing two names. If a stage serves no buyer job at all, it exists for internal reporting and should be a field, not a stage.

Field two, the exit evidence. What observable thing has the buyer done that proves the job is complete? Not what the rep sent. What the buyer produced, agreed to, introduced, scheduled or committed. "Buyer articulated the cost of the status quo in their own numbers." "Buyer introduced us to the person who controls the budget." "Buyer confirmed the evaluation criteria in writing." Exit evidence is the field that makes a process diagnostic rather than decorative, because it is the only field a rep cannot satisfy alone.

Field three, the seller contribution. What does the seller do to help the buyer complete that job? Note the framing: help complete, not extract. Gartner's data suggest buyers who use supplier-provided tools to do their own buying work are 1.8 times more likely to complete a high-quality, low-regret deal. The seller contribution field is where methodology lives, and where adaptiveness must be preserved. Specify the outcome, offer plays, do not mandate a script.

Field four, the kill criterion. What would tell you this deal should stop here? Most processes have no answer, and the cost shows up as pipeline that looks healthy and converts to nothing. Matthew Dixon and Ted McKenna's analysis of 2.5 million recorded sales conversations, published as The JOLT Effect in 2022, found that 40-60% of qualified pipeline is lost not to a competitor but to no decision, and that 87% of the buyers in their dataset showed moderate to high levels of indecision. A kill criterion converts a slow loss into a fast one and frees the capacity that a slow loss consumes.

Four fields, one per stage. If you can fill them for every stage in your pipeline, you have a process. If you cannot, you have a naming convention.

Keep the stage count low. Every stage boundary is a judgment call, and every judgment call is a place where reps disagree and data degrades. Five or six stages that everyone applies consistently produce better forecasts than nine that are applied idiosyncratically. Add a stage only when the buying job genuinely differs and the exit evidence is genuinely distinguishable.

Let deals move backwards, and measure when they do. If buyers loop, and Gartner's research says they do, then a process that only permits forward movement will lie to you. When a new stakeholder joins in month four and reopens requirements, the deal has returned to requirements building. Recording that is not failure. It is the most valuable signal in the pipeline, because backward movement is an early indicator of exactly the consensus and indecision problems that produce no-decision losses. A pipeline where nothing ever regresses is a pipeline where reps have learned that regression is punished.

Use the process for diagnosis, not compliance. The salesforce control literature going back to Anderson and Oliver's 1987 framework distinguishes behavior-based control, which manages what reps do, from outcome-based control, which manages what they produce. Later work, including Oliveira Santini and colleagues' 2019 meta-analysis in the Canadian Journal of Administrative Sciences, points toward hybrid arrangements rather than either pure form. In practical terms: use the process to structure the conversation about what a deal needs next, not as a checklist that determines someone's compensation. The moment stage accuracy affects pay, stage accuracy disappears.

Separate process from methodology, then run both. The process defines the stages, exit evidence and kill criteria. A methodology such as MEDDIC, SPICED or Challenger defines how a rep works inside those stages: which questions to ask, how to qualify, how to frame value. They are different layers and they fail differently. A team with a methodology and no process gets good conversations and unreliable forecasts. A team with a process and no methodology gets clean CRM data about deals nobody knows how to advance.

Rebuild the process from won and lost deals, on a cycle. The process is a hypothesis about what buyers need to do to buy from you. Test it. Pull the last thirty closed deals, won and lost, and check whether the exit evidence you require actually discriminated between them. Evidence that appears in wins and losses at the same rate is not a gate, it is a formality, and it should be replaced. Annually is enough for most organizations. More often than that and you are churning; less and the process slowly becomes a description of a market you used to sell into.

The direct evidence that formal sales processes cause revenue growth is thinner than the frequency of the claim suggests. The indirect evidence, that clarity of role reduces a known drag on performance while preserving adaptiveness protects a known driver, is considerably stronger. That combination points to a specific kind of process: one that is explicit about what buyers must accomplish and deliberately unopinionated about how sellers help them accomplish it.

Build the stage contracts. Anchor them to buying jobs. Require buyer-produced evidence to advance. Write down what would make you walk away. Then use the whole thing as a diagnostic instrument rather than a compliance regime, and let it get corrected by what your own closed deals tell you.