Micro-Audience Hunt Expansion Controls
Expand a hunt by widening recency and growing its signal vocabulary
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 97%
This framework controls the breadth of a signal-based audience hunt through two variables: the recency window and the signal seed list. A short job-ad window, such as 45 days, favors fresh intent but may return only a few companies. Expanding it to 90, 180, or 365 days increases coverage while gradually weakening timeliness. Separately, marketers can ask the research system to return novel KPI phrases found during the hunt, review that vocabulary, and generate additional related phrases. Repeating the search with a larger, normalized seed list exposes more companies and clusters. These controls should be adjusted deliberately and measured separately, because increasing either one can create volume at the expense of relevance. The process ends when marginal additions stop producing coherent, actionable audiences.
Origin
Extracted from Marketing Against The Grain as the host explains how to broaden a KPI hunt beyond its initial three-company clusters.
Core principles
- 01Treat audience breadth as a controllable research parameter.
- 02Use recency windows to balance timeliness against sample size.
- 03Expand signal vocabulary iteratively rather than guessing every phrase upfront.
- 04Preserve cluster quality while increasing coverage.
- 05Review newly generated seed phrases before reusing them.
How to run it
- 1
Measure the initial hunt
Record the current time window, seed phrases, number of qualifying companies, and cluster sizes.
Pro tip Preserve the initial run as a baseline for comparing later expansion.
Watch out Do not change multiple variables without retaining enough information to explain the effect.
- 2
Adjust the recency window
Widen the job-posting lookback from a short fresh window to 90, 180, or 365 days when more companies are needed.
Pro tip Increase the window one interval at a time and label the age of every signal.
Watch out Older postings may represent needs that have already been resolved.
- 3
Harvest discovered phrases
Ask the system to return KPI phrases it found that were not in the original seed list.
Pro tip Keep the source posting attached to every proposed phrase.
Watch out Model-generated phrases without source evidence should not enter the seed list.
- 4
Expand the vocabulary
Review the discovered list and generate a bounded number of additional phrases that belong to the same semantic cluster.
Pro tip Normalize synonyms while preserving materially different metrics as separate signals.
Watch out Broad thematic terms can create noisy matches that do not express a KPI.
- 5
Rerun and compare
Repeat the hunt, then compare added companies, signal age, cluster coherence, and product relevance with the baseline.
Pro tip Use marginal cluster quality, not raw audience size, as the stopping criterion.
Watch out Expansion that produces more names but weaker intent can reduce campaign performance.
In the wild
An initial 45-day search returns three companies in a useful cluster, which is too small for a selected advertising channel. The marketer repeats the search at 90 and then 180 days, labels older evidence, and keeps only companies whose postings still reflect a credible current priority.
→ The cluster gains usable scale while preserving visibility into signal freshness.
The marketer starts with 15 KPI phrases, retrieves newly observed variants, reviews them, and adds related terms in bounded batches until the search covers roughly 60 validated phrases.
→ The hunt identifies additional clusters that the original vocabulary missed.
Common mistakes
Expanding recency without labels
Old and new signals should not be treated as equally timely when audiences are prioritized.
Accepting invented seed phrases
Phrase expansion must remain tied to real job-ad language or it can manufacture false audiences.
Optimizing only for audience size
A larger cluster is not better if its members no longer share a coherent need or product fit.
Is it for you?
Best for
Teams whose initial micro-audience hunt produces credible clusters that are too small for the intended campaign channel.
Not ideal for
Teams that have not yet validated whether the original signal predicts relevance or buying interest.
From the transcript
“So you could obviously go back 90 days, 180 days, a whole year would actually give you a lot more companies.”
“And then you could actually change your seed list, right? The amount of KPI key phrases.”
“So the number of days since the job ad has been posted and the number of key phrases in that seed list, these kind of…”
From the episode
How I Turned Perplexity Labs into a Marketing Machine