Four Tests to Run Before Adding Another Intent Data Source
Another source can improve coverage, add context, or simply duplicate information you already have. Before expanding an intent program, test whether the source changes a useful business decision. More records and more fields are not automatically more intelligence.
The four tests are relevance, novelty, usability, and outcome value. They provide a practical way to review an expansion without assuming that every new feed deserves a permanent place in the model.
Test one: does it represent the purchase you serve?
Compare the source’s topics and observed behavior with your actual offer. Identify adjacent audiences and alternative explanations. A source can be accurate about category interest while remaining poorly aligned with your product.
Review examples with sales. Ask what question the signal would help answer and which eligibility checks remain necessary. Confirm availability and permitted use for the intended workflow.
If the source cannot be explained in terms of a buyer decision, the proposed value may be too vague to evaluate.
Test two: does it add independent information?
Inspect overlap with existing sources and, where possible, the origin of events. Several feeds may reflect the same underlying observation. Counting them as independent corroboration can inflate confidence.
Novelty can mean new suitable accounts, a different buying-stage clue, or better timing. It does not have to mean entirely unique contacts. Define which kind of novelty matters for your business.
Keep identity resolution consistent when assessing overlap. Duplicate records under different identifiers can make a source appear more incremental than it is.
Test three: can the team act on it?
A signal is useful only if it arrives with enough context and within a relevant time window. Review latency, resolution level, required research, and destination eligibility.
If sales cannot interpret the signal or the intended activation route is unavailable, the source may need a different use case. It could support aggregate content planning rather than a direct prospecting queue.
Estimate the operating effort. A source that creates extensive cleanup may still be worthwhile, but the cost belongs in the decision.
Test four: does it improve a defined outcome?
Choose a practical comparison and mature outcome. Depending on the purpose, that might be relevant accounts per review hour, qualified opportunities at fixed capacity, or incremental campaign contribution under a suitable test.
Do not reward the source solely for delivering more records. Volume can dilute attention and create incomplete feedback. Compare outcomes under reasonably consistent treatment.
If the evidence is limited, retain the source as exploratory rather than promoting it immediately into the core model.
A hypothetical expansion review
An agency adds a second external intent feed. The new feed supplies many records, but most overlap existing accounts and arrive later. A smaller portion contains a useful application-specific signal that the original source lacks.
The agency might retain only that distinct use case, subject to permissions and economics, rather than treating the entire feed as equally valuable. The decision follows the information added, not the headline database size.
This example is hypothetical and does not describe a Cortex source or customer result.
Preserve the experiment in the feedback
Record which source contributed to each selected cohort, the configuration date, and the intended action. Keep customer outcomes and rejection reasons linked under the approved workflow.
If sources overlap, avoid assigning every win independently to each source and then adding the totals. Attribution rules should make the overlap visible.
Document changes to the source mix so later model shifts can be interpreted. A new score distribution may reflect the added feed rather than a sudden change in buyer behavior.
Make source control meaningful with Cortex
Cortex’s custom buyer intent approach gives clients a role in choosing relevant sources, websites, topics, and buying signals. The value of that control depends on disciplined review and feedback.
The proprietary learning should help identify which combinations matter for your niche and offer. It should not become a reason to collect more data without a clear purpose.
Review your source mix with Cortex. Bring the decision each current source supports and the gap you want a new one to fill. A smaller, more informative mix can be easier to use and evaluate than a larger unexplained stack.
