Lead Prioritization: Why Score Alone Picks the Wrong Calls

Sort your database by lead score and the top of the list is the people most likely to pick up the phone. That is not the same list as the people worth calling.

Both things are true at once, and most CRM views show you only one of them. A score is a single number, so it goes in a single column, so you sort by it and start dialing. The contact who will happily chat for twenty minutes about a purchase they are not making until spring outranks the one sitting on a stalled listing agreement — because the first one answers and the second one does not.

A score answers one question, and it is not "who should I call?"

Every propensity score in a real estate CRM answers some version of how likely is this? Likely to reply. Likely to convert. Likely to go cold. None of them answers is this worth an hour of a licensed agent's time?

Value is the second axis, and in this business it has a natural unit: the commission the contact represents if the thing they are trying to do happens. A $210,000 townhouse buyer and a $1.4M listing appointment can carry identical engagement scores and land nowhere near each other on the money.

The working rule: score ranks the order inside a group; value decides which group gets your morning.

One caveat before either number is worth anything: both axes assume the contacts being ranked are prospects. Lenders, recruiters, inspectors, other agents and the title rep sit in the same list as your buyers, and they will score well on engagement forever. We covered separating them in not everyone in your Follow Up Boss database is a lead — do that pass first, or you will spend a very productive hour calling your own lender.

Four kinds of "next call"

A ranked list is more useful when it knows what kind of opportunity it is looking at, because the call is different in each case. The Opportunity Map in Ace Trove buckets ranked contacts into four groups, each implying a different opening sentence:

A contact belongs to one group, not several. The classes are evaluated in a fixed order and the first match wins, so what you see is the most actionable thing true about that person right now. A dormant past client with listing signals who is also drifting shows up as a save-the-deal call, because that call has to happen first. Multi-list membership feels thorough and produces a worklist nobody can work.

Unrecognized signals land in "Other," not in the bin. When a signal type appears that the grouping does not know about, the contact still shows up under Other opportunities rather than vanishing from the list. A system that silently drops what it does not understand is worse than one that admits it.

The 40-minute triage

The routine works with any CRM that will put a score and a dollar estimate on the same screen.

1. Filter to one group. Start with save-the-deal. It is the only group where waiting has a cost that compounds.
2. Sort that group by value, not by score. Inside a group everyone already qualifies; the money breaks the tie.
3. Take the top eight. Not fifteen. Eight calls with real preparation beat thirty dials, and you will actually finish the list.
4. Read the evidence before you dial. Every ranked contact should be able to tell you why it is ranked — which signals fired, and when. If your tool cannot show that, the ranking is a vibe.
5. Open honestly on the cold ones. "I'm going through my list and I realized I've let three weeks go by without an update. That's on me. Where's your head at on the timeline right now?" It works because it is true, and because it hands them the exit they were going to take anyway — the fastest way to find out whether this is still a deal.
6. Log the outcome in the CRM, not in your head. An unlogged call is a call the ranking will never learn from.

Run it weekly — the KPI checklist covers the numbers that belong in the same sitting.

The chart lies when you filter it

Plot your contacts on two axes, click a filter, and watch the picture reorganize. Charting tools commonly derive the axis from whatever data is in view; in Vega-Lite that is the documented default — "by default, a scale in Vega-Lite draws domain values directly from a channel's encoded field" (Vega-Lite scale docs). Filter to your smallest, cheapest segment and the axis obligingly shrinks to fit it. The three contacts that looked like a rounding error a second ago now fill the frame.

You would think knowing about this protects you. It does not. Across five studies, Yang and colleagues found that y-axis truncation "leads viewers to perceive illustrated differences as larger," with 83.5% of participants showing the effect — which, in their words, "persisted after viewers were taught about the effects of y-axis truncation" (Yang et al., 2021, Journal of Applied Research in Memory and Cognition). Graph literacy failed to predict the size of an individual's effect. Correll and colleagues found the impact "persistent across visualizations designs, even for designs with explicit visual cues that indicate truncation has taken place" (Correll et al., CHI 2020) — labeling the problem on the chart did not undo it. And it runs both ways: a 2025 study reports observers "perceive larger differences between values when there is truncation and smaller differences when there is expansion at either end of the y-axis" (Zang and Cousineau, 2025, Journal of Experimental Psychology: Learning, Memory, and Cognition).

Honest caveat: the literature is not unanimous. A 2022 randomized study of a truncated line chart found "context is the only significant factor in opinion-forming; the misleading graph and graph literacy had no effect" (Driessen et al., 2022, PLoS ONE). All of these studies also test static charts with fixed axes — mostly bar charts, not the scatter a two-axis opportunity view actually is, and Zang and Cousineau found that a plot without bars produced less biased judgments under truncation. Nobody has measured the case we actually care about: a scatter that rescales under your hand when you click a filter. Treat the connection as reasoning about a shared mechanism, not as a measured result.

What we did about it

The Opportunity Map locks the vertical scale to the whole loaded set of contacts, then filters the points. Switch from "All" to "Ready to buy" and the surviving dots stay exactly where they were. Nothing rises. Nothing falls. The only change is how many dots are lit.

A test named for the promise it protects — a class filter keeps the axis anchored to the same loaded slice — means a future change that quietly reintroduces the floating axis fails the build instead of shipping. Small thing to test, easy thing to lose.

A blank is not a zero

The second way a two-axis view misleads you is quieter. Some contacts are missing one of the two measures. Plotting them at zero puts a real person in the "not worth calling" corner of your chart on the strength of a number nobody computed.

The rule we hold: a contact missing either measure is not plotted, is counted, and the count is printed under the chart — how many loaded opportunities have both measures, how many do not, and that those contacts remain in the ranked list below, where you can still work them. The caption ends with "Estimates are not guaranteed revenue," because otherwise someone reads a dollar axis as a forecast.

The stage ring on the Companion dashboard — a donut of your contacts by stage that you tap to isolate one slice — applies the same discipline in a different place. It distinguishes two empty states most dashboards collapse into one: "No contacts in this view yet" means it counted and found nothing, while "Stage counts are unavailable" means it does not know. Different facts, different words.

Same failure class we catalogued in the CRM dashboard audit, which walks seven ways a single dashboard number goes wrong. This post is about the chart on top of those numbers; that one is about the numbers themselves, and it is the better place to start if your tiles have ever disagreed with each other.

What these numbers do not prove

Four limits, stated plainly, because a tool that hides its limits is how you end up trusting it more than it has earned.

• The opportunity score is a ranking index, not a probability. It runs 0 to 100 and it orders contacts within a group. A 74 is not "74% likely" to do anything, and we do not present it as one.
• The dollar axis is a ranking weight, and it is an estimate. Expected commission is model-derived, carried so the list can sort by money, and labeled as an estimate everywhere it appears. It is not a forecast of your closing table.
• The map shows the slice that was loaded, not your whole database. The caption says which. A chart that will not tell you what population it drew from is asking you to assume the most flattering one.
• Some of this is gated, and we would rather say so up front. The stage ring renders for anyone with the Companion dashboard. The "Ask Ace about this stage" step that turns a slice into a conversation, and the account-wide Opportunity Map for team leads, are Trove features.

Separately, two free fields — Ace Win Score and Ace Churn Risk — write into Follow Up Boss on every account. Those are a different tool for a different job: they live in your CRM, where your existing filters can see them. The opportunity view is where you go for the ranking and the reasoning on one screen.

If you only do one thing

Open whatever prioritized list you already trust, find the dollar figure next to each contact, and sort by it. If the order barely moves, your score and your value are measuring the same thing and one of them is decorative. If it moves a lot, you have been calling the wrong half of your list — annoying to learn, and worth more than the twenty seconds it took.

From there: seller-side signals are the most neglected half of most databases, Follow Up Boss's own pipeline reports cover the stage-level view, and the nurture path is what "Other opportunities" is for.

Frequently asked questions

Should I call the highest-scoring lead or the highest-value one?

Neither, alone. Filter to the group where waiting costs you something — a valuable contact going cold — then sort that group by value. Score decides who belongs in the group; value decides the order you work it.

Why does my pipeline chart change shape when I apply a filter?

Because in many charting tools the axis is rebuilt from whatever data is in view — the documented default in Vega-Lite, for instance. Filter to a smaller segment and the axis shrinks to fit it, making the remaining points look larger. The underlying numbers did not move. Check whether your tool's axis is locked to the full set before reading anything into the shape.

Is a contact with no dollar estimate a low-value contact?

No — it is a contact with an unknown value, and the two should never be drawn the same way. Plotting a missing measure at zero is how a real prospect ends up parked in the corner nobody looks at.

What does an "expected commission" number on a contact mean?

In our case, a modeled estimate used to rank the list by money, labeled as an estimate wherever it appears — not a prediction of what you will bank. Ask any vendor showing a per-contact dollar figure the same question, and be wary if the answer arrives without a caveat.