How I Stress-Test a B2B GTM Plan Before the Board Sees It, With a Pre-Mortem and a Bottom-Up Count

When I finish drafting a go-to-market plan, I assume it is too good to be true. The author is the worst person to find its gaps, because every assumption in it was chosen by someone who wanted it to work.

I write board plans in one format, the Board Plan on a Page, drawn from the McKinsey and SaaStr playbooks. A thesis and headline tiles sit on top, then eight numbered boxes: where to play, why now, what we do, the solution suite, how to win, the market strategy, the numbers, and what it takes. Annexures follow in the same order. The format helps a board read fast, and it hides one kind of failure: each box reads well on its own while two boxes disagree.

So the stress test is mostly about the joins. Does the target fit the count of real accounts? Does the pipeline fit the sales cycle and the calendar? Does the budget fit the cost of a customer today? The table alongside shows the question I put to each box.

The order matters, and the table alongside shows it. The CRAFT GTM analyzer goes first because it is quick and shows which parts of the plan are thin. Its own output says the scores come from "a keyword check", so it is a coverage check, not a judgement.

The bottom-up count goes second whenever the target is a number of new customers, because a board challenges that number first and the fact check needs the count as a source. ICP Intelligence works it out from your own account count, ACV, ICP match rate and Year 1 share. For other targets, skip this pass.

The Claims fact check skill goes third, on both the plan and the analyzer's critique, because a critique makes claims too. It gives every claim a verdict and leaves a placeholder where only you can supply the fact.

The Second opinion review skill goes last. It judges each critique point through three filters (facts, the stated objective, tone), says what the critique missed, and ends with one final version of the plan.

Why a quarter: I bring the board the annual plan with its first quarter written out in full, because that is the part the board can check against actuals at its next meeting, and the part where a wrong assumption costs least to fix. So the example stress-tests Shelfwalk's Q1, the first quarter of its annual plan. The same four passes run on each quarter and on the full year.

Step 1. Run the CRAFT critique

Input to the CRAFT GTM analyzer, as a quarterly plan: "Q1 plan: win 10 new consumer goods brands by running two distributor meets, a LinkedIn campaign for national sales heads and an outbound sequence to heads of sales operations. Owner: marketing lead. Budget: $25K. Success metric: 40 qualified opportunities by 31 March."

What came back: 34 out of 50, "rated GOOD", strong on Character and Timeline. Its fixes, in order: no deliverable, document or channel asset; no audience, persona or segment, no assumptions and no current state; not "a target with a number"; and no review rhythm. It added "No risk is named in the plan." Its sector check said the owner, managing director or head of operations usually signs, that "Sales, distribution and IT review a purchase", and that buyers "prefer a pilot in one region, judged against a comparable region."

How to read it: the analyzer shows the plan's own line for each score, so check every "not found" against your text. Four of its readings do not match this plan, as in the second table alongside. Its other points hold, and the missing risks and review rhythm are the ones a board will raise.

A second run with the audience, the desired outcome and the industry set to vertical SaaS gave the same score and points, with the first two shown as labels at the top.

Step 2. Test the target against a bottom-up count

Input to the ICP Intelligence TAM, SAM and SOM calculator, all example inputs: 1,200 consumer goods brands with 100 or more field reps in the target region, ACV $18,000, a 30% ICP match and a 3% Year 1 share.

What came back: the arithmetic in the table alongside, and "~11 closed customers" for Year 1, with $583.2K of pipeline at a 1 in 3 win rate, which it labels as an example figure. Its sector note says the National Sales Head or Managing Director signs, that distributors "hold the stock and order data", and to count only companies where those roles exist and the problem is measured.

How to read it: the plan asks for 10 brands in one quarter against a count that supports 11 in a year. Either the count is too low (check the 1,200 against an account search and a second source, and hand-check a sample of 100 to set the 30%), or the Q1 target should come from the closed-won count per quarter. Either way, the board sees arithmetic, not an argument.

Show the board the pipeline at your own win rate. The calculator's 1 in 3 is a labelled example. On my own arithmetic, at Shelfwalk's 25% (an example input) the year's $194.4K needs about $777.6K of pipeline, and the plan's 40 opportunities at $18,000, $720K at 25%, are $180K of new ACV in one quarter, close to the whole year's count.

Step 3. Fact check the plan and the critique

Input to the Claims fact check skill: the plan, the analyzer's output, and the figures I hold, all example inputs: ACV $18,000, a sales cycle of 45 days, a win rate of 25%, and the bottom-up count from step 2.

What came back: a register with a verdict and a smallest fix for each claim; the plan's four sentences became seven rows. The 10 brand target was marked "Weaker than written" against the bottom-up count. The pipeline arithmetic, 40 opportunities at 25%, does give 10 wins, but with a 45 day cycle an opportunity qualified after about 14 February cannot close by 31 March. "Qualified" has no definition, the budget has no split, and 31 March has no year.

On the critique, it gave eight rows: four verified, three contradicted by the plan text and one corrected, the same four readings flagged in step 1.

How to read it: each "[User to add]" is a question the board would otherwise ask you in the room, and "as pasted, not yet checked" is the honest status of a planning assumption.

Step 4. Get a second opinion on the critique

Input to the Second opinion review skill: the plan, the critique, the register, the bottom-up count, the example figures (ACV $18,000, sales cycle 45 days, win rate 25%, and Shelfwalk's example CAC of $9,000) and the purpose: the CEO and the board approve the $25K and the 10 brand target.

What came back: a verdict for each critique point. It accepted five points, some with edits, modified three (on the score: "do not show it to the board as a grade") and denied the two that misread labels. It also said the review was not independent of the fact check run in the same session, and suggested a fresh reviewer.

Then it listed five things the critique missed. The first three came from the register: the target against the bottom-up count, the 14 February line, and a single owner for both opportunities and wins. The fourth was the cost per win: $25K over 10 wins is $2,500, against the example CAC of $9,000, so either the budget is programme spend only or the plan assumes a far cheaper customer. The fifth was who comes to a distributor meet: a room full of distributors produces distributor contacts, not brand opportunities.

That fifth point shows why every reading needs checking against its source. The skill credited it to "the analyzer's sector note", but the wording it quoted, that distributors "hold the stock and order data", is in the calculator's note. The point stands, but a board member who checks the citation would find it wrong.

It ended with a board pre-mortem: it is 31 March 2027 and the plan won 4 brands, so why? Each failure story has an early sign you can read by mid-February, as in the table alongside. Then it wrote one final version of the plan, each gap left as "[User to add]".

Step 5. Put the corrected plan back in the eight boxes

The skill's final version is a list of lines. Before it goes to the board I put it back on the one page, box by box, as in the table alongside. Every line comes from the plan, the tools or the skills above, except the $777.6K pipeline, which is my own arithmetic from step 2.

How to read it: why now and solution suite were empty in the original plan, and no tool filled them, which is right. A critique cannot know what changed for your buyer this year or which features are live, so those boxes are yours to write. Then hand the page to someone who did not write it and ask which box they would challenge first.

1. Put the annual plan on one page in the eight boxes with the first quarter written out, run the CRAFT GTM analyzer on the full text, and check every "not found" against your own text.
2. If the target is a customer count, run the bottom-up count next, and show the pipeline at your own win rate.
3. Run Claims fact check on the plan and the critique together, with last year's win rate, cycle and ACV as your named sources.
4. Run Second opinion review on the critique and the register, check each of its readings against the output it cites, and take its final version as the draft for the board pack.
5. Put the final version back in the eight boxes, fill every "[User to add]", and ask someone who did not write the plan to read it last.

1. Showing the board a critique score as a grade. The analyzer's score measures which parts are covered, not whether the plan will hit its number.
2. Accepting a critique, or a second opinion, without checking it. In this example four of the analyzer's readings did not match the plan, and the second opinion credited a sector note to the wrong tool.
3. Checking each box on its own. Most failures sit between boxes: a target the count cannot support, or pipeline that cannot close inside the period.
4. Filling the gaps with estimates. A placeholder the board can see is safer than a number nobody can source.
5. Letting the author run the final review. Ask a fresh reviewer, a new session or a person, to read the final version last.

• How I Check Product-Market Fit, Churn and ICP Gaps Before Scaling Paid Demand Generation
• How I Pick the Lead B2B Go-To-Market Motion With EPIC, Then Size and Anchor It
• From Selling Tokens or Hours to Selling Outcomes: An Outcome Pilot Workflow for AI-Native and IT Services Firms
• All 15 guides