Recency, frequency, latency

Three numbers, computed correctly, will tell you more about the state of a deal than the entire CRM stage system that most sales teams rely on. They are the recency, frequency, and latency of meaningful behavior on the deal, and the reason they work is that they describe what is actually happening between buyer and seller rather than what someone has typed into a form.

Recency is how many days have passed since the last meaningful touchpoint. The word meaningful is doing important work in that sentence, because most sales tools count things as touchpoints that should not be counted. A calendar reschedule from the buyer is meaningful because the buyer chose to do it. An email open is not, because nothing about an email opening tells you the buyer engaged with the content. A reply is meaningful, an out-of-office bounce is not. We define meaningful as anything the buyer actively chose to do, and we count those things and only those things, because everything else is noise that has been dressed up to look like signal. Recency is the strongest single predictor of future behavior in the entire framework, the one number Novo's research surfaced as more reliable than anything else after forty years of testing across industries, and it works because recent buyers are likely to keep buying, recent repliers are likely to keep replying, and buyers who have not engaged recently are statistically much more likely to never engage again, regardless of how engaged they were in the past. Recency captures the trend that matters.

Frequency is the total count of meaningful touchpoints across the lifetime of the deal, and the version that actually predicts outcomes is the one segmented by stakeholder role. A deal with twelve champion touches and zero economic-buyer touches is structurally different from a deal with six of each, even if the totals look similar at a glance. Frequency tells you how built out the relationship surface is, how many people at the buyer's organization the seller has actually reached, and that depth is what determines whether a single person leaving the company kills the deal or whether the deal survives. High frequency with one person is a single point of failure, and high frequency across multiple stakeholders is a deal with structural depth. The two look the same in a basic activity report and they predict completely different outcomes.

Latency is the average time between sequential events on the deal, the gap between first call and discovery, between discovery and demo, between demo and proposal touch, between proposal and reply. These intervals, taken together across the events of the deal, form a sequence with a typical shape. For any given deal type at any given company, there is a normal rhythm of how the events unfold over time, and latency captures whether this specific deal is matching the normal rhythm or drifting away from it. The shape itself, the sequence of intervals expanding or contracting, is often more useful than any individual interval, because it tells you whether the deal is gaining momentum or losing it.

Each of the three numbers is useful on its own, but the value compounds when you look at them together.

A deal with high frequency and low recency and lengthening latency is a deal that was engaged and is now decaying. The frequency was built up over weeks of work, the recency is gone because the buyer has stopped responding, and the gaps between events are growing because each interaction is harder to schedule than the last. This is the most expensive failure mode in B2B sales, because it is a deal you have already invested in, paid the acquisition cost for, built the relationship around, and you are watching it slip while the data is screaming that something has changed. Catch it early enough and you can save it, miss it and you lose a deal that was already paid for.

A deal with rising frequency and recent recency and shortening latency is a deal that is accelerating. Each new event is happening sooner than the last, the buyer is pulling forward, and the rhythm of the relationship is compressing into the kind of dense back-and-forth that closing always looks like. The right move on a deal in this state is often to not over-message, to let the momentum carry, and to fire trip wires only on the points of friction that could break the rhythm if left unaddressed.

A deal with low frequency and recent recency and unknown latency is new. Treat it as new, not as stalled and not as accelerating, because the data is not yet there to tell you which it is. The right moves are the ones that reduce time-to-second-meeting and start building the rhythm that the rest of the framework depends on.

This is the data layer underneath the four-quadrant view, the trip wires, the deal scoring, and the next-best-action recommendations. Every output Yuzu produces ultimately resolves down to a function of these three numbers, computed against the population baseline of your historical closed-won cohort and against the per-deal-per-stakeholder baseline of this specific buyer's own pattern, and the math is not the interesting part. The interesting part is that almost no other sales tool computes any of this correctly, because almost no other sales tool starts from the assumption that what counts as a meaningful event needs to be defined carefully before any of the math has a chance of working.

Two practical points sellers tend to find counterintuitive.

Recency should be calculated per-stakeholder, not per-account, because the deal as a whole might look active on the surface while the actual decision-maker has been silent for weeks. The procurement contact might have replied yesterday, which makes the account-level recency look fresh, but the economic buyer has not been heard from in fourteen days, and that is the recency that actually matters. Per-stakeholder recency reveals divergences that the per-account view hides, and those divergences are usually where the deal is going to be won or lost.

Latency baselines need to be both personal and population-level, because they capture different things. Your champion's typical reply gap might be thirty-six hours and mine might be seventy-two, and a fifty-hour silence means something completely different for each of us. Population baselines tell you the cohort norm, what is typical for buyers like this one, and personal baselines tell you what is anomalous for this specific relationship. You need both, and you compare against both, and the divergences each of them surface are different and complementary.

The seller does not have to compute any of this manually. Yuzu computes it continuously, surfaces the deals where the numbers say something meaningful, and stays out of the way on the deals where they do not. The discipline of the framework is in what you choose not to act on, as much as in what you act on.