Introduction
"We made 800 calls this week." That sentence could describe an excellent week of prospecting or a completely wasted one: it tells you nothing either way. This is the core problem with managing outbound phone prospecting. The numbers that are easiest to produce are the ones that inform the least. We unified five telephony platforms into a single database and analyzed roughly 50,000 calls. That work taught us one thing above all: before you measure, you have to define. Here are the metrics that deserve a place on your dashboard, the ones that lie, and the discipline required to stop fooling yourself.
Define Your Terms: "Call" Means Nothing
The word "call" covers at least three distinct realities. Confusing them distorts everything else:
- Dialed call: the SDR dialed a number. This includes unanswered rings, voicemails, and invalid numbers.
- Pickup: a human answered. That is already useful information (about list quality and reachability), but a "hello" followed by a hang-up ten seconds later still counts as a pickup.
- Conversation: an actual exchange took place. Our operational threshold is a conversation of at least one minute. Below that, you are statistically looking at an immediate refusal or a voicemail, not a real exchange.
Why be this precise? Because every telephony platform counts differently. One reports all dialed calls, another marks only connected calls as "completed," a third has its own definition of "connected." When we unified our five platforms, we had to establish a single rule and name it explicitly in every report. "Conversations of 1 min or more" and "pickups" are two separate columns, never merged. A number you cannot precisely describe is not a KPI: it is a rumor.
Reachability and Attempts: The Curve That Decides When to Stop
The second useful family of metrics: how many attempts does it take to reach a prospect, and when should you give up? Our data produces a clear curve. Twenty-two percent of prospects are reached on the first attempt. Cumulatively, 30.7% after two attempts, 34.8% after three, 36.8% after four. After that, the curve flattens: beyond the fifth attempt, each additional try reaches less than 1% of the remaining contacts, and the eighth attempt reaches only 0.2%.
This is not a decorative statistic; it is an allocation rule. We stop at around four to five attempts and reinvest that calling time on fresh contacts. Without this curve, two symmetrical mistakes become likely: giving up after a single try (and leaving nearly half of all reachable prospects on the table) or pushing to ten attempts (and paying SDR days for crumbs). The average number of attempts per prospect is therefore a discipline KPI. If it drifts upward, the team is grinding; if it collapses, the team is skimming.
Meetings Booked vs. Meetings Held: The Only Number That Pays
A booked meeting is just a promise. The number that matters to the end client is the meeting that actually happens. The gap between the two (no-shows) is a metric in its own right, and it is actionable: confirmation reminders, qualification quality, the time between the call and the scheduled meeting. A team measured only on "meetings booked" quickly learns to book fragile ones. A team measured on meetings held learns to qualify properly. This is also a question of internal alignment. Our SDRs are compensated on meetings delivered (not on call volume), precisely so that the individual metric and the value produced are the same thing.
Vanity Metrics: Recognizing the Numbers That Lie
A few classics, and why they mislead:
- Total calls dialed. This measures activity, not output. It goes up when the list is poor (you dial more because no one answers), meaning it appears to improve when the situation is actually getting worse.
- "Connection rate" without a definition. If voicemails count as connections, the rate looks impressive and means nothing. Always demand the definition before admiring the number.
- Global averages. An average conversation rate over twelve months smooths out everything interesting: differences between lists, time slots, and job titles. We systematically break data down by list, by day and hour, and by function. Decisions live in the gaps, not the averages.
- "Generated" pipeline. A sum of hypothetical deal values weighted by self-declared probabilities. Impressive in a meeting, unfalsifiable, and therefore unusable.
The simple test: a good KPI must be able to go down. If no realistic scenario causes your favorite metric to drop, it is not a metric; it is a sales argument.
Instrumenting Without Fooling Yourself
Three practices that meaningfully improve the reliability of prospecting reporting:
Measure your own coverage. Of the 49,898 calls we analyzed, 33,821 (roughly 68%) can be matched to a known prospect in our lists. The rest comes primarily from campaigns loaded directly with external vendors that do not match our database. We display this coverage rate alongside every analysis. A reachability rate calculated on 68% of calls is not wrong, but you need to know it covers 68% of calls. An honest dashboard says what it cannot see.
One definition, everywhere. The rule "conversation of 1 min or more" is the same across our list analyses, our per-SDR statistics, and our hourly curves. The moment two reports use two definitions of the same word, meetings turn into interpretation debates.
Instrument at the source, not in a spreadsheet. Numbers reworked by hand on a Friday evening always end up telling the story you want to hear. At Scalefast, every call lands automatically in the database with its platform, duration, and status. The dashboard reads the database directly; nobody "prepares" the numbers.
Conclusion
Good prospecting management relies on a small number of metrics, but each one defined with surgical precision: real conversations (with an explicit threshold), pickups, attempts per prospect (with a stop rule), meetings held, and the coverage rate of the measurement itself. Everything else (raw volumes, undefined rates, smoothed averages) decorates slides and degrades decisions. The question to ask about every number stays the same: "what does this actually count?" If the answer takes more than one sentence, be suspicious.
