The Decision Brief · Independent marketing strategy publication

Decide what to test before you commit the budget

A working publication for teams weighing AI-assisted marketing plans. Four short essays show how to frame the decision, sort what you know from what you assume, compare proposals without pretending to a score, and keep the review honest. Then a private worksheet sets up to three proposed initiatives side by side and exports a Markdown brief.

Overhead view of the octagonal La Trobe Reading Room, with wooden reading tables, green lamps and people studying
La Trobe Reading Room, State Library of Victoria, Melbourne: wooden reading tables with green lamps beneath the octagonal dome, seen from above. Photo by DP Singh Bhullar. Cropped, resized and converted to WebP. Bhullargraphic · CC BY-SA 4.0 · source

Two failure modes are useful to examine in a marketing plan. Either the team commits before the question is clear, or it gathers so much material that no one can say what would count as an answer. The worksheet exists to close that gap: you state the decision first, then attach evidence, assumptions, costs and constraints to each candidate.

01

Four essays, in the order they help

Each essay addresses one part of the same job. Read them in sequence if you are starting from nothing; read a single one if you already know which part of the plan is weak.

  1. 1

    Start marketing planning with a decision

    Write the decision question before the proposal. A decision question names the choice, the person who makes it and the deadline, so that every later note has something to attach to. Proposals that cannot name a decision tend to expand instead of resolve.

  2. 2

    Keep evidence separate from assumptions

    Evidence is something you could show a skeptical colleague; an assumption is something you are willing to act on without that. Both belong in the brief, but in different places. Mixing them is how confident-sounding plans lose their footing.

  3. 3

    Compare initiatives without fake precision

    Three proposals, one table. Cost, owner, time constraint and the unknowns each proposal carries are enough to make the trade-off visible. A composite score would hide the disagreement you most need to see.

  4. 4

    Set a useful review cadence

    A review date is a promise to re-read the assumptions, not an occasion for a status meeting. The cadence should match how fast the underlying evidence could plausibly change.

02

What a filled-in brief actually contains

The worksheet does not generate recommendations and it does not score anything. It holds what you type and returns it in a readable order. That is a deliberate limit: the judgment stays with the people who hold the constraints.

Fields the worksheet collects for each of up to three proposed initiatives
Field What belongs there What does not
Decision question The single choice the brief is meant to resolve, with the person deciding and the date. A goal statement, a slogan, or several unrelated questions stacked together.
Evidence Named sources, internal records or observations you can point a colleague to. Confidence, hunches, or a competitor's visibility presented as proof of demand.
Assumptions What you are taking for granted to make the proposal work, stated plainly. Assumptions smuggled into the evidence column to make the case look settled.
Cost Direct spend and the internal hours you expect to consume. A single round number that erases the difference between spend and staff time.
Owner and time constraint Who is accountable, and the deadline or capacity limit that governs the work. Shared ownership with no name attached, or an implied "as soon as possible".
Open unknowns Questions the brief cannot answer yet, and what would settle each one. Silence where the answer is simply not available.

The table is also the shape of the exported Markdown file. Because nothing is computed, two people reading the same export see the same unresolved questions — which is usually the point of writing it down.

Printed reference volumes standing on shelves in the Main Reading Room of the Library of Congress
Printed reference volumes on the shelves of the Main Reading Room, Library of Congress — an example of physical research collections. Photo by Djembayz. Cropped, resized and converted to WebP. Djembayz · CC BY-SA 3.0 · source

03

Why the evidence column has to stay honest

Two published sources shape the standard this publication applies to its own material and to the plans it helps readers assemble.

Substantiation, stated plainly

The US Federal Trade Commission states that advertising must be truthful and non-deceptive, that advertisers must have evidence to back up their claims, and that before running an ad an advertiser needs a "reasonable basis" — objective evidence supporting the claim. The FTC also states that statements from satisfied customers usually are not enough to support a claim that requires objective evaluation, and that a money-back guarantee is not a substitute for that substantiation.

The practical translation for a decision brief: if a proposed initiative rests on a claim about what the work will achieve, write down the objective evidence behind that claim, or move it into the assumptions column and say so. The FTC further notes that it looks at both express and implied claims, and that its staff cannot clear ads in advance — which means the substantiation has to live in your own records, not in an approval.

AI-assisted plans and stated risk

NIST describes its AI Risk Management Framework (AI RMF 1.0) as intended for voluntary use, to improve the ability to incorporate trustworthiness considerations into the design, development, use and evaluation of AI products and services. NIST states the framework was released on January 26, 2023, was developed through a consensus-driven, open and collaborative process, and that AI RMF 1.0 is being revised as part of the White House AI Action Plan.

NIST also released a Generative AI Profile (NIST AI 600-1) on July 26, 2024, which it describes as helping organizations identify risks unique to generative AI and proposing actions for managing them; NIST states the profile centres on 13 risks and more than 400 actions. Where a proposed initiative depends on an AI-assisted step, those risks belong in the unknowns and assumptions of the brief, not in a footnote added after the decision.

Content made to be useful

Google states that its automated ranking systems are designed to prioritize helpful, reliable information created to benefit people rather than content made to manipulate rankings, and that E-E-A-T is not itself a specific ranking factor, with trust the most important of those aspects. Google also states that fabricating creator profiles — AI-generated headshots, made-up names, false credentials — is a form of deception and a signal of a low-quality page, and that using automation to produce content primarily to manipulate rankings violates its spam policies.

This publication therefore keeps its authorship, its sources and its limits in view, and treats the worksheet as a tool that helps a page do its job rather than as a substitute for judgment. Nothing here promises a ranking, a result or a review.

Source documentation for these statements is listed on the sources and credits page.

04

Use it as a private working document

Everything you enter stays with you. The worksheet runs no external models and produces no score; it arranges your own words into a brief you can circulate, revise or discard. Because it exports Markdown, the brief can live next to the rest of your planning notes and be diffed when circumstances change.

If you are unsure where to begin, start with the decision question. If the question is already clear, go straight to the worksheet and let the unknowns column do its work.

Long wooden tables and bookcases in the Suzzallo Library reading room, University of Washington
Reading-room tables and wooden bookcases, Suzzallo Library, University of Washington, Seattle. Photo by Tom Harpel. Cropped, resized and converted to WebP. Tom Harpel from Seattle, Washington, United States · CC BY 2.0 · source