The 30-day window: recency as a score component — Open Pulse

At a glance

Published

2026-10-12

Written for

Anyone deciding how fresh a buying signal has to be before it is worth a reply

Category

informative

Reading time

8 minutes

Keywords

recency lead scoring, buying signal decay, lead response time, intent data recency, lead scoring components

Overview

Every signal Open Pulse stores carries a timestamp, and the timestamp is not decoration. It is one of seven score components, worth 10 of a signal's 100 points, and it is also a hard filter: posts older than the listener's data window never reach the classifier at all. This piece explains why the window exists, why it is 30 days for most listeners, and what the research says about how fast a buying signal goes stale.

The honest admission first. There is no single study that proves "30 days" is the right window for every market. The research we can cite is on response time and signal decay, and it points in the same direction: interest is perishable, and the decay is steepest early. The window is our reading of that evidence, tuned for public buying signals rather than inbound form fills. We state the reasoning so you can disagree with it.

Why buying intent is perishable

A person complaining about their billing vendor in public is describing a moment, not a permanent condition. Moments resolve. They renew with the incumbent, they buy a competitor, they deprioritise the initiative, or the person posting leaves the role. Every day that passes between the post and your reply is a day in which one of those resolutions becomes more likely.

The oldest hard data on this is from the Lead Response Management studies summarised in Harvard Business Review by James Oldroyd, Kristina McElheran and David Elkington in "The Short Life of Online Sales Leads" (March 2011). Auditing 2,241 US companies, they found 37% replied to an online lead within an hour, 16% within one to 24 hours, 24% after more than 24 hours, and 23% never replied at all. In a separate analysis of 1.25 million sales leads at 42 US companies, reported in the same article, contacting a lead within an hour was nearly seven times as likely to qualify it as contacting it an hour later, and more than 60 times as likely as waiting 24 hours or more. A related analysis found contact within five minutes carried 21 times the qualification odds of contact at 30 minutes.

The footnotes matter, and we will not bury them. The study is from 2011, it measured inbound sales leads rather than public buying signals, and the numbers describe qualification odds rather than closed revenue. What has survived is the shape of the decay: steepest in the first hours, compounding with every day. Practitioner playbooks on intent data arrive at the same shape. A 2026 practitioner playbook on intent activation recommends full weight for signals under 14 days, progressive discounting to 30 days, and treating anything past 60 days as expired for most B2B software purchases. That is a vendor-authored playbook, not peer-reviewed research, but it is written by people who watch thousands of signals decay in real pipelines, and it agrees with the academic curve.

Thirty days sits inside that consensus: long enough to catch the cycles where people research on weekends and discuss with colleagues midweek, short enough that the list you triage on a Monday morning is still mostly alive.

What the window does to a run

A listener's data window is applied during the cheap filtering stage, before any model is called. Each run does its work in the same four steps: search every source in the plan, pull the post text wherever the platform allows it, filter cheaply, and only then judge what survives. "Cheaply first" means page shape, off-topic, your exclusions, blocked domains, duplicates, and anything outside the data window are all discarded before the classifier sees them. Old posts never reach the classifier at all.

Rejections from those filters are kept with their reasons for 14 days, and "outside the window" is one of the recorded reasons. When a listener feels quiet, the quality report shows, per query, how many candidates were kept and how many were cut, and why. The distinction matters: an empty search is a plan problem, a full rejection log is a window problem, and the fix for each is different.

When a listener is new, you can look further back for a single run from its panel, so it is not staring at an empty screen until its window fills. That is the exception, not the rule. The run takes the wider window once, and every scheduled run after it keeps the standing window.

Why local prospects get 90 days

One listener objective works on a 90-day window instead of 30: Local prospects, which reads Google reviews for the businesses you sell to. The exception is deliberate. A review is a slower signal than a post. A business owner reads a one-star review on a Tuesday and acts on it the following month, or never. The buying moment for the supplier selling to that business, "your reviews are bleeding customers," stays valid longer because the underlying condition, a reputation problem with a 90-day trail of evidence, does not resolve overnight the way a migration question does.

The window matches the signal's half-life, not a single number for the whole product. That is the design principle behind all of this: recency is not one threshold applied blindly. It is a judgement about how long a particular kind of moment stays actionable.

Recency is 10 points, and it is the humblest component

Every signal carries a score out of 100, built from seven components. Recency is worth 10 points. Purchase intent gets 30, fit 20, urgency 15, commercial value 10, evidence quality 10, and multi-signal 5. The weighting is a statement of belief: freshness matters, but a fresh post that is not about your problem is worth nothing, and a week-old post describing exactly your problem is worth almost everything.

Notice what recency does not do. It does not decide anything on its own. A signal cannot clear the default score bar of 55 on recency alone, and it cannot be kept on recency plus a coin flip either. The bar is crossed by the combination, intent first, fit second, the rest supporting. Recency is a modifier, not a verdict. A score that treated a three-week-old perfect-fit post as equal to a three-day-old perfect-fit post would be lying about which one you should read first.

That is also why the window and the component exist together. The window decides what is allowed in the room; recency decides the seating order. A post from 29 days ago is allowed in, and scores lower than an identical post from yesterday. A post from 31 days ago is not in the room at all. The cutoff is arbitrary at the margin, and the scoring is continuous everywhere else. We would rather have the arbitrariness in the open than hidden.

The one edge case worth naming

Some posts carry no date the run can read. Those are never assigned a guess: the signal says "no date" rather than inventing one, and when the run has fetched the page and still cannot find a date, the candidate is set aside with the reason "No date we could read". This is the same instinct as the territory view, which reports an unrecognised location as Unknown rather than guessing a country. A wrong date on a signal is worse than an admitted gap, because people make outreach decisions on timing. The instrument reports what it measured and marks what it could not.

The rule of thumb

When you evaluate any listening product, ask two questions about its time handling. First: what is the data window, and is it the same for every signal type or matched to the signal's half-life? A single window for everything is a product that has not thought about how differently a complaint and a review decay. Second: is recency a score component or the score? A product that sorts purely by date is telling you every signal is equal except its age. A product that weights recency alongside intent and fit is telling you when to read, not what to think.

And ask what happens at the boundary. If the tool cannot tell you what happens to a post that is one day too old, that is itself the answer.

If you want to see what a fresh scan of your market looks like before anything decays, run a free report. It is a capped, real scan, and the results arrive ranked with their timestamps attached.

References

• James B. Oldroyd, Kristina McElheran and David Elkington, "The Short Life of Online Sales Leads", Harvard Business Review, March 2011. Source for the audit of 2,241 US companies, the response-time split, and the qualification odds within an hour and within five minutes.
• "B2B Intent Data Activation: Signal to Revenue Playbook", House of Martech, 2026. Practitioner playbook, vendor-authored. Source for full weight under 14 days, progressive discount to 30, and expiry past 60.
• "Intent Signal Outreach: LinkedIn's 3% vs 55% Split", LinkedCamp, 2026. Practitioner guide, vendor-authored, on signal decay by type: job changes hot for about 30 days, funding 7 to 14, content engagement 48 to 72 hours.
• Why your keyword alert is noise: a taxonomy of false positives, Open Pulse blog.

Frequently asked questions

Next

• Start the free trial — Creates a workspace. A card is taken up front and nothing is charged during the trial.
• Pricing — What it costs after the trial.
• All posts — The comparisons and the guides, grouped.

See also

• What Open Pulse is and what it watches
• llms.txt — the index of this site, written for agents
• api.md — the MCP server and the REST API, with scopes
• auth.md — how to authenticate
• openapi.json — the REST API as OpenAPI 3.1
• sitemap.xml — every indexable page