Request a callHow to audit Demandbase intent keyword sets before you activate them
Audit each intent keyword set by removing duplicates and off-topic terms, testing a refined version against an untouched control, and judging both by the accounts they surface and what happens to those accounts in the CRM. Demandbase documents two MCP tools that help: Rank Keyword Sets, which finds the best-fitting existing set for a campaign or page, and Evaluate Keyword Set, which measures how each keyword lines up with your opportunity pipeline.
Key takeaways
- Intent programs accumulate broad, duplicated and outdated keywords that create noisy account signals.
- Keep the original set as a control and compare it with a refined version over the same period.
- Demandbase MCP tools Rank Keyword Sets and Evaluate Keyword Set can show fit and pipeline alignment, and they are documented as tools you can call from an AI assistant.
- AI recommendations are a starting hypothesis. Only CRM outcomes show whether a keyword set improves pipeline.
Why keyword sets drift
Intent keyword sets start tight and grow. New campaigns add terms, competitors change positioning and products are renamed, while old terms stay in place. Over time a set mixes high-intent terms with generic ones, and the accounts it flags become harder to prioritize.
Adding a keyword to a Demandbase set brings in intent activity from the past 13 months, and removing one deletes its associated historical activity, according to Demandbase support documentation. Changes therefore affect the history you see, which is a reason to duplicate a set before editing it.
What the Demandbase tools do
- Rank Keyword Sets: finds which of your existing sets best fit a campaign, product or landing page, and whether to reuse one or create a new set.
- Evaluate Keyword Set: measures how well each keyword in a set lines up with your opportunity pipeline.
- Account-Level Intent: finds accounts showing intent by set, keyword, strength, date or trend.
These are described in the Demandbase MCP tools overview, which was last updated on October 19, 2026. Check the live tool list for current names and behavior before building a process on them.
A clean test design
- Pick one established intent set and duplicate it. Do not edit the original.
- In the copy, remove duplicates and terms that no buyer would use, and add terms from buyer research and sales calls.
- Run both sets for the same period against the same account universe.
- Compare the accounts each version surfaces and what happens to them afterward in the CRM.
- Keep the version that finds more qualified accounts with fewer false positives, and retire the other.
Measures that decide the test
| Measure | What it tells you |
|---|---|
| Account overlap | How much the two sets agree. |
| Net-new qualified accounts | Whether the refined set finds accounts the control missed. |
| Opportunity rate by intent tier | Whether higher intent actually converts better. |
| False-positive rate | How many flagged accounts sales rejects as irrelevant. |
| Pipeline per activated account | Whether activation spend is going to the right accounts. |
Keep a human in the loop
Treat any AI suggestion about keywords as a hypothesis from a model that has not seen your sales conversations. Review additions and removals with sales, and record why each change was made so the set does not drift again.
Frequently asked questions
Should we edit our live keyword set?
Duplicate it first. Removing a keyword deletes its historical intent activity, so keep the original as a control.
Are these MCP tools a replacement for an analyst?
No. They give you faster evidence, but deciding what to change and testing it against CRM outcomes still takes judgment.
How long should the test run?
Long enough to cover a normal buying cycle stage for your accounts. Agree the period before you start.
Sources
Related service: Demandbase programs. See also pricing and client results.
