Sparks now actively records what it learns about your accounts
The Sparks agent will pin candidate patterns much more aggressively. Previously it was waiting to be certain before recording an observation as a pattern — which meant it almost never recorded anything, and the Patterns panel stayed empty. Now it records tentative observations on the first sighting and only surfaces them once a later run independently confirms them. Aggressive on the input side, conservative on what reaches you.
You'll see Patterns build up faster on accounts Pace has investigated more than once. Within a couple of agent runs, accounts with consistent behaviour ("dips on Mondays," "branded out-converts generic," "weekend ROAS is structurally lower") should start showing dots on the Patterns panel where they previously showed "No patterns yet."
Acute issues still go through Sparks, not Patterns. A broken landing page or a sudden conversion drop is a Spark — something happening now. Patterns is for enduring traits of the account that hold week after week.
🚀 New
Sparks have a calmer voice
No more "CRITICAL:" or "URGENT REPEAT ISSUE:" headlines. Sparks now write in the voice of a senior analyst leaving a note, not an alarm system: sentence case, declarative, no shouting. The severity badge on each card already tells you how serious the finding is, so the title can describe what happened without yelling about it.
Recommendations live in the description, not the title. A title like "Pause Shopping campaign immediately" becomes "Shopping campaign CPA reached $124 this week," with the suggested action in the body of the alert.
Critical findings are still clearly critical. The red badge, the placement at the top of the feed, the Slack ping — none of that changed. What changed is that the writing trusts you to read it.
🚀 New
Patterns surface faster — Pace recognises the same observation even when worded differently
Pattern recognition no longer depends on word-for-word wording. When Sparks notices the same behaviour about your account across multiple runs — say a Monday slump, or branded campaigns out-converting generic — Pace now treats those observations as the same pattern even if the agent phrases them slightly differently each time. The bar to promote a tentative observation into a confirmed pattern (two sightings across separate runs) is unchanged, but it's now actually reachable.
Existing accounts catch up immediately. A one-shot cleanup runs across every workspace and consolidates the patterns that were sitting in limbo because they'd been recorded in slightly different words. Accounts that previously showed "No patterns yet" should start surfacing what Pace has actually been noticing.
🚀 New
Sparks are quieter when an issue is ongoing
Same finding, told once. When Sparks flags a problem on an account, you'll hear about it once and then it goes quiet for three days, even if the underlying issue is still there. No more daily Slack pings about the same broken landing page or the same paused campaign.
If the issue is still active after three days, Sparks will re-flag it. That way you find out a problem hasn't been resolved without being reminded every single day.
Severity escalations always come through. If something gets worse — say a warning becomes critical because the account paused itself — the new Spark fires immediately.
Genuinely new findings are unaffected. Different problems on the same account still surface independently. The cooldown only applies to repeats of the same finding.
The Sparks tab and the home dashboard will still show the original finding for as long as it's relevant. The change is to delivery: one alert per issue, not one per day.