How to Stop a Marketing Strategy From Going Stale

Five symptoms show up before anyone admits the strategy is out of date, and each is observable rather than a matter of opinion:

• Nobody can say when a claim in it was last checked against a competitor or a customer.
• The competitor set names companies you no longer meet in deals, and omits the ones you lose to.
• A differentiator in the document now appears in three competitors' ads, so it has become table stakes.
• The channel mix reflects where the team had skills two years ago rather than where buyers are now.
• Nobody has changed a line of it in a year, which in a moving market means the review never happened.

Work an AI does reliably when it is given your own dated observations:

• Producing a changed-versus-unchanged list from a quarter of monitoring, which is the slowest manual part of any review.
• Extracting the repeated claims from fifty competitor ad texts and grouping them by promise and objection.
• Drafting the diff between the current strategy document and what the evidence now supports, as a proposal to argue with.
• Noticing that the same observation has appeared in three consecutive reviews without anyone acting on it.

Decisions that have to stay with a named human:

• Which of the changes is worth a response, since responding to all of them is how a strategy becomes a list of reactions.
• Whether the evidence is sufficient, which is a judgement about sample size and source quality rather than a summarisation task.
• Anything involving price, legal claims or a public comparison, where the cost of a confident error is external.
• The final wording of positioning, because the sentence a team can repeat is not the sentence a model optimises for.

The order matters more than the tooling, because the common failure is discussing opinions before anyone has listed the facts.

1. Collect what changed, before opening the strategy. Pull the competitor changes, your own channel numbers and the deals won and lost since the last review into one list. Do this first, because reading the strategy document first anchors the discussion on defending it rather than testing it.
2. Map each change to the assumption it threatens. Every item on the list either threatens an assumption or does not. A competitor adding a free tier threatens a pricing assumption. A competitor posting more often usually threatens nothing. This step removes most of the list, which is the point of it.
3. Check the survivors against the evidence bar. For each threatened assumption, ask whether the evidence meets the bar for its part of the strategy. Positioning needs two quarters of consistent signal. A content theme needs three attempts. If the bar is not met, the item stays open rather than becoming a change.
4. Decide, and name what you are giving up. Write each decision as what changes and what it costs: which channel loses budget, which claim gets dropped, which segment gets less attention. A decision with no cost attached is an addition, and a strategy that only accumulates is how teams end up doing everything badly.If nothing is being given up, the review has produced a wish list rather than a strategy change.
5. Edit the document and log the change. Change the strategy text itself, not a slide about it, and record the date, the evidence and the decision in a change log. Without that record the next review cannot tell a deliberate bet from a drift, and the same argument gets had again.

The output is small, deliberately, and it is the same three items every time:

1. The edited strategy itself, with the changed sentences actually changed. A memo describing what should change is not a change.
2. One log entry per change: the date, what changed, the evidence, and what you expect to see if the decision was right.
3. A watch list of assumptions where the evidence was not yet sufficient, so the next review starts from them rather than from scratch.