Evidence
The raw material of good decisions — and how Synloom keeps it honest.
Evidence is anything that tells you something true about your users or market: support tickets, interview transcripts, survey results, analytics, competitor intel, internal docs. In Synloom, each piece becomes an evidence item you can search, filter, and feed to agents.
Credibility
Not all evidence is equal. Each item carries a credibility level (high / medium / low). When agents reason over your evidence, stronger sources carry more weight — so a single noisy comment doesn’t outweigh a clear pattern across many interviews.
Staleness
Evidence ages. Synloom models this with a staleness weight — older evidence gradually counts for less, with a roughly six-month half-life. Crucially, this is dynamic: evidence that led to a successful outcome decays slower (it earned trust), and evidence tied to a failed prediction decays faster.
Lesson: not all evidence deserves equal weight
- What it is
- Every evidence item carries a credibility level and a staleness weight. Together they decide how much a piece of evidence influences synthesis — a recent, high-credibility interview counts for far more than a two-year-old anonymous comment.
- Why it matters
- Teams routinely over-index on whatever is loudest or most recent in memory. Weighting by credibility and freshness keeps rankings grounded in signal, not recency bias or a vocal minority.
When you import, spend the extra few seconds to set credibility. It’s the cheapest way to make every future synthesis sharper.
A problem that “solved itself” after a redesign
Dana, PM on a consumer app
The situation. Eighteen months ago, onboarding confusion dominated Dana’s feedback. The team shipped a redesign, and complaints tapered off — but the old tickets are still in the evidence corpus.
What they do
- 1Dana runs synthesis on the full history.
- 2The old onboarding tickets are past their staleness half-life, so they count for much less than they did.
- 3Recent evidence — now about a different friction point — outweighs them.
The result. Onboarding drops down the ranking on its own, and the current top problem reflects where users struggle today. Dana isn’t chasing a problem she already fixed.
Why this matters
Over time, the evidence that actually predicts good outcomes rises to the top automatically. That’s the foundation of Synloom’s feedback loop.