
Evidence needs to be fresh enough that the reason still holds when the buyer reads it. The working question is whether a dated observation still supports outreach, needs a recheck, or should be retired. Syft provides a dated source behind each reason to engage an account, which makes that decision visible before a draft goes out.
There is no universal expiry window for buying signals because different kinds of evidence describe events that last for different lengths of time. A short-lived burst of research can go cold in days. A public initiative can remain a fair reason for months if the work is still underway.
A single rule such as "anything older than 30 days is stale" will retire good reasons too early and keep weak ones too long. Judge freshness by whether the reason still holds. For definitions of signal types, see What Are Go-To-Market Signals.
Evidence of short-lived attention ages quickly, and evidence of a still-active initiative holds longer. Use categories, then check whether the underlying situation is still true.
These are starting categories. Ask whether a buyer would still recognize the reason today. For fit, timing, and corroboration, see Qualified Signals: Fit, Timing, and Corroboration.
You should recheck evidence before outreach whenever the date is old relative to its category, the source is thin, two records disagree, or a seller cannot explain the reason in a sentence.
A practical recheck:
If the reason still holds, keep it and send. If you cannot tell, hold the draft until you can. AI agents writing in Claude, ChatGPT, and Grok Bot should pause when the date or source is missing. For why agents fall back on stale context, see Why AI Sales Agents Use Stale Data and Invent Reasons.
You should retire a reason to engage when the situation has ended, a newer source contradicts it, the owner is gone, or you can no longer open and verify the original evidence. Retirement is how you stop a former truth from becoming this week's outreach hook.
For example, say an account announced a rollout in January, and a March update says the program was paused. The January reason should come out of the live queue, even if the original source is still online. Keep it in history if you want the narrative, and do not use it as why now.
Retire also when the link to the problem you solve was only an assumption. A true, dated event that does not connect to what you sell is not a reason to engage.
You tell whether stale evidence is hurting outbound by measuring what happens after you send. Compare outreach that used a current, dated reason with outreach that used an old or undated one, over the same segment and period.
Useful metrics:
If the outdated-reason replies or the rewrite rate are high, freshness is a sales problem. Tighten recheck and retirement rules, then measure the same metrics again.
A 30-day-old signal is not automatically too old. Short-lived attention may already be past its usefulness at that age, while an in-progress initiative may still be a fair reason if a source shows it is active. Judge the signal by whether the situation still holds today.
AI agents should check dates themselves when the evidence includes a date and a source they can show. If the date or source is missing, the agent should hold the outreach and ask for a recheck rather than guess. Agents in Claude, ChatGPT, and Grok Bot can only apply a freshness rule to evidence that arrives with a date.
When two sources disagree, treat the reason as unverified and recheck before you send. Prefer the source you can open, that names the same company, and that has the more recent last-checked date. If you still cannot reconcile them, retire the reason from live outreach until you can.
Account context should be rechecked right before outreach, and on a regular cadence in between so old reasons do not pile up. Fast-changing categories need more frequent review than in-progress initiatives. Set a cadence by category and recheck again at send time.
If you want to see dated, sourced reasons on accounts in your territory, book a custom demo.
Better context in, better actions out.