How do you document the concrete steps for a first-line manager to coach a rep whose forecast accuracy is below 50% in 2027?

The Tuesday morning that starts every one of these plans

A first-line manager runs a team of seven. Six of them land within a reasonable band of their call each quarter. One does not. Over the last three quarters this rep called $410K, $380K, and $455K in commit, and closed $170K, $190K, and $215K — roughly 44%, 50%, and 47% attainment against the commit number. That is a rep whose forecast accuracy sits below 50%, and it is a specific, diagnosable operating problem, not a character flaw.

The manager's instinct is usually a conversation: pull the rep aside, express concern, ask them to "tighten up the forecast." That conversation happens, the rep agrees, and eleven weeks later the same gap appears. Nothing was documented, so nothing was measurable, so nothing changed. The manager cannot escalate because there is no record. The rep cannot improve because they were never told which specific behavior produced the miss. And RevOps cannot fix the model, because a single rep's noise is invisible inside a team roll-up until someone writes it down.

What actually breaks the cycle is a written coaching document — a plain artifact, usually one to three pages, that a manager builds once and updates weekly. It is not an HR performance improvement plan, though it can become the evidence base for one. It is an operating document with five parts: the measured gap, the diagnosed cause, the specific behavior changes with dates, the inspection cadence, and the checkpoint where the manager decides whether the plan worked.

The distinction matters more than it sounds. A performance improvement plan is a legal and HR instrument with defined language, defined timelines, and defined outcomes. A forecast coaching document is a management tool. It should live in whatever the team uses for one-on-one notes — a shared doc, a Notion page, a call-plan field on the account record — and it should be visible to the rep. The rep should be able to read exactly what number is being measured, what they are being asked to change, and what happens at the 60-day mark. Managers who write these documents privately, as a file-building exercise, produce worse outcomes than managers who hand them to the rep on day one and edit them together every week.

The forecast-accuracy problem is worth this level of process because it compounds outward. A rep who calls $410K and delivers $170K does not just miss their own number. They corrupt the manager's roll-up, which corrupts the regional number, which corrupts the number the CRO takes to the board. In a team of seven, one rep at sub-50% accuracy on a $400K commit introduces roughly $230K of variance into a team forecast that might only be $2.5M. That single rep can move the team's aggregate accuracy by nine points. Finance builds hiring plans on those numbers.

What the coaching document actually contains, section by section

Start with the measurement section, and make it boring and precise. The document opens with a table: for each of the last eight to twelve weeks, what the rep committed at week start, what closed by week end, and the delta. Then the same view by quarter for the last three quarters. Do not editorialize inside the table. The point of putting raw numbers first is that the rep reads them before they read any judgment, and the numbers are not arguable.

Define the accuracy metric explicitly in this section, because "forecast accuracy" means at least four different things depending on who is speaking. The three common definitions: commit attainment (closed ÷ committed, where 100% is perfect and over 100% is sandbagging), absolute variance (|closed − committed| ÷ committed, where 0% is perfect and direction is ignored), and category slippage (what percentage of deals called commit actually closed in the called period, regardless of dollar value). Pick one, write the formula into the document, and never change it mid-plan. Changing the metric mid-plan is the single fastest way to make a coaching document worthless — the rep will correctly point out that the goalposts moved, and they will be right.

Second section: the diagnosis. This is where most managers fail, because they write "needs to improve forecast discipline," which is not a diagnosis, it is a restatement of the symptom. There are only a handful of real root causes for sub-50% accuracy, and they require different coaching.

Timing error. The deals close, just later. The rep's qualification is fine and their win rate is fine, but they consistently believe a deal will close 30 to 60 days before it does. This shows up as a rep whose annual number is decent but whose quarterly calls are always wrong in the same direction. The fix is calendar-based: close-date discipline anchored to a verifiable customer event, not the rep's hope.

Qualification error. The deals do not close at all. The rep is calling deals commit that have no budget confirmation, no identified signing authority, or no articulated business problem. Their pipeline looks healthy and converts terribly. The fix is criteria-based: no deal enters commit without documented evidence against a named framework.

Single-thread error. The rep has one champion and no other contact in the account. The deal is real and the champion is genuine, but the champion cannot sign, gets reorganized, or goes quiet. This produces erratic accuracy — some quarters fine, some catastrophic. The fix is relationship-mapping: a documented multi-threading requirement before commit.

Pressure error. The rep knows the deal is shaky but calls it anyway because the manager, or the culture, punishes a low number more than an inaccurate one. This one is uncomfortable because the diagnosis implicates the manager. It shows up when a rep's private assessment of a deal, if you ask them casually, differs from their CRM stage. The fix is not rep-side at all.

Hygiene error. The rep's CRM is simply stale. Deals sit in the wrong stage, close dates roll forward automatically, and the "forecast" is a snapshot of a data set nobody maintains. This is the most common cause in teams under 20 reps and the easiest to fix, and it is frequently misdiagnosed as a judgment problem.

Write the diagnosis with evidence. "Of the 14 deals called commit in Q2, 9 slipped at least one quarter and 5 were lost. Of the 9 that slipped, 8 had a close date that had already been pushed at least twice. Of the 5 lost, 4 had a single contact on the opportunity." That is a diagnosis. It points at timing error and single-thread error, and it tells you exactly what to coach.

Third section: the specific behavior changes. Concrete means a rep could hand this to a peer and the peer could execute it identically. Each item gets an owner, a due date, and an observable output. "Multi-thread the Henderson account" is not concrete. "By Friday March 12, identify and get a first meeting with the VP of Operations at Henderson; log the meeting and the notes on the opportunity record" is concrete.

Cap it at three to five behaviors. A document with eleven required changes produces zero changes. Pick the two or three that map directly to the diagnosed cause, and let the rest go for this cycle.

Fourth section: the inspection cadence. Write down when the manager and the rep will look at this together, what they will look at, and how long it takes. The standard shape is a 30-minute weekly deal inspection separate from the regular one-on-one — separate because if you fold it into the one-on-one, it gets displaced by whatever is on fire that week, every week.

Fifth section: the checkpoint. A date, a target, and a written statement of what happens at that date under each outcome. Sixty days is the common interval for a full-quarter sales cycle. The checkpoint should specify the metric value that counts as success, the value that counts as partial progress, and the point at which the manager escalates to a formal HR process.

Running the weekly inspection so the document stays alive

The document is worthless if it is written once and filed. What makes it work is the weekly ritual, and the ritual has a shape worth being specific about.

Before the meeting, the manager pulls the rep's commit list and looks at three things: which deals changed stage since last week, which close dates moved, and which deals have had no activity logged in the past seven days. That takes about ten minutes in most CRMs and can be a saved report or dashboard that RevOps builds once for the whole team.

In the meeting, go deal by deal through the commit list only — not the full pipeline. For each deal, the manager asks the same four questions, and the sameness is the point, because it trains the rep to pre-answer them:

What has to be true for this to close on the date you have? Who, by name and title, signs it? What did the customer do in the last seven days that proves this is still moving? If this slips, what is the reason, and do we already know it?

That fourth question is the one that surfaces timing error. A rep with good judgment can usually name the slip risk immediately. A rep with sub-50% accuracy often cannot, which tells the manager the rep has not thought past the optimistic path.

The manager writes the answers into the document — or into a deal-inspection field, if RevOps has built one — and next week, reads last week's answer back before asking again. That single loop, comparing what the rep said would happen to what happened, is the entire mechanism by which forecast judgment improves. It is not the coaching conversation that produces the improvement; it is the repeated confrontation with the rep's own prior prediction.

Keep the meeting to thirty minutes. If the commit list is too long to inspect in thirty minutes, the commit list is the problem, and that is itself a finding worth writing into the document.

Numbers worth calibrating against

Be careful with benchmarks here, because published forecast-accuracy figures vary enormously by segment, deal size, and — critically — by which definition of accuracy the publisher used. Rather than treating any external number as authoritative, calibrate against your own team.

The most useful benchmark a manager has is the internal distribution. Take the last four quarters, compute the chosen accuracy metric for every rep on the team, and rank them. The median rep on the team is the realistic target for a struggling rep, not the top performer and not an industry figure from a report about a different segment. If the team median commit attainment is 82% and the struggling rep is at 47%, the 60-day target is not 95%. It is something like 65–70%, which is meaningful progress toward the median without being a target the rep will read as impossible and stop trying for.

Set the improvement target as a fraction of the gap. The gap here is 35 points. Asking for a third to a half of the gap in the first 60 days — so 12 to 18 points, landing the rep at 59–65% — is a target a rep can believe in. Targets that close the full gap in one cycle get abandoned in week three.

On sample size, be honest about what the numbers can tell you. A rep carrying four to six commit deals per quarter produces an accuracy figure with enormous variance. One large deal slipping can move a quarterly number by 30 points with no change in the rep's judgment at all. For low-volume, high-ACV reps, prefer deal-level accuracy — of the deals you called commit, what fraction closed in the called period — over dollar-weighted accuracy, and look across three or four quarters rather than one. For high-volume transactional reps closing 20+ deals a quarter, dollar-weighted accuracy over a single quarter is a reasonable signal.

On time-to-improvement, the honest range is one to two full sales cycles. If the average sales cycle is 90 days, a 60-day checkpoint is measuring behavior change, not outcome change — most of the deals that will close in that window were already in flight and already mis-forecast before the plan started. Write that into the document explicitly. The 60-day checkpoint asks: did the behaviors change? Did close dates get anchored to customer events? Did multi-threading happen? The outcome question — did accuracy actually improve — needs 120 to 180 days for a long-cycle team.

On inspection load, a manager with seven to nine reps running a 30-minute weekly inspection each spends roughly four hours a week on deal inspection. That is real and it is why managers skip it. If the manager has more than ten reps, the honest recommendation is to run the full inspection only for reps below the accuracy threshold and a lighter 15-minute version for everyone else.

On escalation timing, if a rep shows no behavior change by week four — not no outcome change, no *behavior* change, meaning the documented actions simply did not happen — that is the signal to escalate the conversation, well before the 60-day checkpoint. Behavior that has not started by week four rarely starts in week seven.

Where this trades off against other approaches, and when to pick something else

Individual coaching documentation is not always the right instrument, and a manager who reaches for it reflexively will burn time on problems it cannot solve.

If more than roughly a third of the team is below the accuracy threshold, the problem is systemic and individual coaching plans are the wrong tool. Three reps out of seven under 50% means the stage definitions are ambiguous, the commit criteria are unwritten, the comp plan rewards pipeline volume over accuracy, or the forecast is being called under pressure from above. Writing three coaching documents in that situation treats symptoms and burns manager time that should go to fixing the definitions. The diagnostic question is simple: is this rep an outlier on this team, or is this team an outlier in this company?

The alternative to individual documentation is systemic tooling — forecast categories with hard entry criteria enforced in the CRM, required fields that block a deal from entering commit without a documented economic buyer, or an AI-driven forecast model that scores deals independently of rep judgment. These scale better than manager coaching. They also fail in a specific way: they change what the rep types, not what the rep believes. A rep who does not understand why a deal is not real will fill in the required fields with plausible-sounding text and call it commit anyway. Systemic controls raise the floor on data quality; only coaching changes judgment. Most teams need both, and the sequencing usually runs systemic-first — fix the definitions and the required fields, then coach the reps who are still wrong afterward, because that second group is a much smaller and much better-diagnosed set.

There is also the manager-forecast-override alternative. Some organizations solve rep inaccuracy by simply not using rep numbers — the manager applies their own judgment to each deal and calls a number that ignores the rep's call. This works, in the narrow sense that the manager's number is more accurate. It also permanently removes the feedback loop that would have taught the rep anything, and it does not scale past the number of deals one manager can personally assess. Use overrides as a short-term patch during the coaching period, and be explicit in the document that the override is temporary and tied to the rep's accuracy recovering.

And there is the honest option nobody writes down: sometimes the rep is not going to develop forecast judgment, and the right answer is a role change or an exit. Documentation matters more, not less, in that case. The document is the difference between a defensible decision and a surprise. If a manager is 70% sure by week three that the outcome is an exit, the document should still be written and run in good faith — because the 30% case is real, and because a plan run as theater is visible to the rep and poisons the rest of the team's trust in the process.

The failure modes that make these documents useless

Writing the document without the rep in the room. A coaching document assembled privately and delivered as a verdict produces defensiveness, and defensiveness produces compliance without change. Build the measurement section yourself, then walk through the diagnosis with the rep and let them argue. If the rep says "those four deals slipped because procurement changed their process, not because I misjudged them," check it. They might be right, and if they are, the diagnosis was wrong and the coaching would have been aimed at the wrong behavior.

Confusing the coaching document with the HR document. These have different purposes, different audiences, and different language. Mixing them means either the coaching document reads like a legal threat — which kills the collaboration that makes it work — or the HR document is too soft to support any actual decision. Keep them separate. When escalation happens, the coaching document becomes evidence *for* the HR document; it does not become the HR document.

Measuring the metric the rep can game rather than the one that matters. If you measure only commit attainment (closed ÷ committed), a rep can hit 100% by committing almost nothing. Their forecast becomes technically accurate and operationally useless — nobody can plan against a rep who commits $50K and closes $50K on a $400K quota. Pair any attainment metric with a coverage or commit-as-percentage-of-quota check so sandbagging shows up as clearly as over-calling.

Letting the definition drift. Six weeks in, the manager decides that "actually we should measure best-case accuracy too," or the metric quietly changes from dollar-weighted to deal-count. Every prior week's data becomes incomparable and the rep loses trust in the whole exercise. Lock the definition in week one, write the formula down, and if the metric genuinely needs to change, start a new plan with a new baseline rather than editing the old one.

Skipping the inspection when the quarter gets busy. The weeks a manager most wants to skip the inspection — end of quarter, big deal in flight — are exactly the weeks with the most forecast signal. A cadence that survives only the easy weeks trains the rep that the plan is optional. If the meeting genuinely cannot happen, do it asynchronously in the document rather than dropping it.

Coaching the forecast instead of the deal. The forecast is an output. A rep whose deals are badly qualified will produce an inaccurate forecast no matter how carefully they are taught to categorize. If the diagnosis says qualification error, the coaching content is discovery skill and business-case construction — the forecast fixes itself downstream. Managers who spend twelve weeks teaching a rep to categorize deals more conservatively have produced a rep who accurately predicts that their bad deals will lose.

No named consequence at the checkpoint. A plan with a date and a target but no stated outcome for missing it is a suggestion. The consequences do not have to be punitive — "we move to a formal performance conversation," "we reassign the enterprise segment and you focus on mid-market," "we extend 30 days with a narrower focus" are all legitimate. What matters is that it is written before the checkpoint, not decided after.

Treating the plan as finished when accuracy hits the target. Accuracy that improves under weekly inspection and collapses the month after inspection stops was never the rep's judgment improving — it was the manager's judgment substituting for theirs. Taper the cadence rather than cutting it: weekly for 60 days, biweekly for the next 60, then monthly. If accuracy holds through the taper, the judgment is real.

Ignoring what the pattern says about hiring and enablement. If the third rep in a row from the same hiring cohort or the same onboarding class lands under 50% accuracy, the coaching document has stopped being a rep problem and started being a signal about how reps are trained on what "commit" means in this company. RevOps and enablement should see the aggregate pattern across coaching documents, with names stripped, precisely so systemic causes surface before the fourth rep hits the same wall.

Related questions

How long should a forecast coaching plan run before you decide it failed?

Sixty days measures behavior change; 120 to 180 days measures outcome change on a 90-day sales cycle. Escalate early — around week four — if the documented behaviors simply have not happened, since behavior that has not started by then rarely starts later.

Should the rep see the coaching document?

Yes, and they should help edit it. Documents written privately produce compliance without judgment change. The rep needs to know the exact metric, the exact behaviors, and the exact checkpoint date, or they cannot self-correct between inspections.

What if the whole team's forecast accuracy is low?

Then it is not a coaching problem. Roughly a third of the team below threshold points at ambiguous stage definitions, unwritten commit criteria, or a comp plan rewarding pipeline volume. Fix the definitions and CRM controls first, then coach whoever is still wrong.

Does an AI forecast model remove the need for coaching?

No. Independent scoring models raise data-quality floors and catch deals rep judgment misses, but they change what the rep types, not what the rep believes. Use them alongside coaching, not instead of it — most teams need both layers.

How do you keep a coaching document from becoming an HR paper trail?

Keep them structurally separate. The coaching document is an operating tool the rep co-owns; the HR plan is a formal instrument with its own language and timelines. The coaching record can become evidence for an HR process later without being written as one.

FAQ

What exactly goes in the coaching document?

Five sections: a measurement table of weekly commit versus closed for 8–12 weeks with the accuracy formula written out; a diagnosis naming which root cause the evidence supports; three to five concrete behavior changes each with an owner, a date, and an observable output; the weekly inspection cadence with its standing questions; and a checkpoint date with the target value and the named consequence for each outcome. One to three pages total. Longer documents get read once.

Which definition of forecast accuracy should the plan use?

Whichever one the rest of the organization uses, so the rep's number is comparable to their peers'. If there is no organizational standard, commit attainment (closed ÷ committed) is the easiest to explain and the easiest to game, so pair it with a sandbagging check. For low-volume enterprise reps, deal-level accuracy — what fraction of committed deals closed in the called period — is more stable than dollar-weighted figures. Write the formula into the document and do not change it mid-plan.

How do you tell a timing problem from a qualification problem?

Look at where the deals ended up. If deals called commit eventually close but a quarter or two late, that is timing, and the fix is anchoring close dates to a verifiable customer event — a contract expiry, a budget cycle, a system cutover — rather than the rep's expectation. If deals called commit are lost or go dark, that is qualification, and the fix is evidence requirements before a deal can enter commit at all.

Is this a performance improvement plan?

No. A PIP is a formal HR instrument with defined language and legal implications. A forecast coaching document is a management tool the rep can see and help edit. It can become the evidence base for a PIP if the checkpoint fails, but writing it as a PIP from the start kills the collaboration that makes coaching work and produces defensiveness instead of behavior change.

What role does RevOps play in this?

RevOps builds the reporting the manager inspects against — a saved commit-versus-closed view by rep, by week — so the manager is not rebuilding the measurement section by hand every cycle. RevOps also owns the definitions: what commit means, what evidence a deal needs to enter it, what the fields are. And RevOps should see the anonymized pattern across coaching documents, because three reps failing the same way is a systemic finding, not three coaching problems.

What if the rep's accuracy improves and then falls back?

That usually means the inspection was substituting the manager's judgment for the rep's rather than developing it. Taper the cadence instead of stopping it — weekly for 60 days, biweekly for 60 more, then monthly — and watch whether accuracy holds as the support comes off. If it degrades at each taper step, the underlying judgment has not changed and the plan needs a different diagnosis, not a longer runway.

Sources

• https://hbr.org/2010/12/stop-making-excuses-for-your-sales-forecast
• https://www.salesforce.com/resources/articles/sales-forecasting/
• https://www.gartner.com/en/sales/topics/sales-forecasting
• https://www.mckinsey.com/capabilities/growth-marketing-and-sales/our-insights/sales-growth-five-proven-strategies
• https://hbr.org/2017/07/how-to-improve-your-sales-forecast
• https://www.hubspot.com/sales-forecasting
• https://openviewpartners.com/blog/sales-forecasting/
• https://www.shrm.org/topics-tools/tools/how-to-guides/how-to-establish-performance-improvement-plan

Related on PULSE

• How do you define commit, best case, and pipeline so every rep categorizes deals the same way?
• What deal-inspection questions actually surface slip risk before it happens?
• How should RevOps build a commit-versus-closed report managers can run themselves?
• When does low forecast accuracy signal a comp plan problem rather than a rep problem?
• How do you multi-thread an enterprise deal before it enters commit?
• What does a healthy pipeline coverage ratio look like by segment and sales cycle length?