Telesales
Telesales Metrics: Which Denominator Should Your Team Use?
Separate call attempts, live conversations, follow-ups and orders. Use one fictional example to distinguish activity counts from KPI rates, choose explicit denominators and report completed observation windows.
อ่านเป็นภาษาไทย
A tale of one customer called twice
Imagine a small Thai online cosmetics shop — call it Malee Glow — whose telesales team spends every weekday afternoon calling customers who have not reordered in sixty days. At a later reporting snapshot for one calling week, the team lead pulls up a single number from the call log: 200 calls. She divides 20 purchases by 200 and announces a 10 percent conversion rate. Her manager nods. Everyone feels reasonable about the result.
Except the numbers hide a structural problem. Of the 200 logged attempts, 80 are unanswered repeat dials made to 40 customers who were never successfully reached — roughly two unanswered attempts per unreached customer. Every one of those 80 dials produced no contact. The remaining 120 attempts each reached a distinct customer in a live conversation, giving the team 120 contacts. Among those 120 contacts, 40 have pending follow-up appointments whose outcome window has not yet closed — they form an immature cohort. The remaining 80 contacts are mature: their attribution window has closed, and 20 completed orders resulted, giving a purchase-per-mature-contact rate of 25 percent, not 10. None of the 40 customers with pending follow-up appointments appear in the purchase numerator yet, because their cohort is not yet mature.
The practical decision is which denominator to use when reviewing a telesales campaign. This guide separates counts for attempts, contacts, scheduled follow-ups and orders, distinguishes those counts from KPI rates, and states the eligible population and observation window. The fictional worked example shows why 25 percent per mature contact and a preliminary 10 percent per attempt answer different questions; neither establishes causal lift or a performance benchmark.
The four layers of a telesales funnel and why each needs its own denominator
A telesales funnel is not a single conversion event. It is a sequence of distinct transitions, and each transition has its own population — its own denominator. Using the wrong denominator for a given transition produces a rate that is arithmetically valid but managerially meaningless.
The four layers are: call attempts, contacts, follow-ups scheduled, and purchases. A call attempt is any logged outbound dial, regardless of outcome. A contact is a conversation in which a real, identifiable customer was reached and heard the offer. A follow-up is a scheduled future action — a callback, a second demonstration, a promised quote — that the customer agreed to during a contact. A purchase is a completed transaction that can be traced to the telesales interaction.
Each layer feeds the next, but the populations are not interchangeable. Attempts include multiple dials to the same customer. Contacts count distinct conversations, not distinct customers unless your logging explicitly deduplicates by customer ID across the cohort window. Follow-ups are a subset of contacts, not of attempts. Purchases are a subset of whichever denominator your team agrees to use — and that choice must be stated explicitly alongside every rate you publish.
Salesforce describes KPIs as metrics used to measure how closely sales-team performance tracks to predetermined goals, distinct from general activity counts; the critical discipline is that each KPI aligns with a specific objective and carries a clear denominator definition. Mixing layers violates that discipline before any data is even collected.
Sources: Salesforce: Sales KPIs — definition, importance, and how to track them
Why attempts and unique customers are not interchangeable
The conflation most damaging to telesales reporting is using attempt count as a proxy for the number of customers reached. Consider the following fictional scenario, stated here as an explicit teaching example with invented numbers.
Malee Glow's team makes 200 logged call attempts in one week. Of those 200 attempts, 80 are unanswered repeat dials made to 40 customers who were never successfully reached — approximately two unanswered attempts per unreached customer — producing no contacts. The remaining 120 attempts each reached a distinct customer in a live conversation, so the team contacted 120 distinct customers. In other words, the 200 attempts comprised 120 that each produced a contact with a distinct customer, and 80 unanswered repeat dials to 40 customers who were never successfully reached.
The same fictional snapshot contains 20 completed orders from 20 buyers. Its preliminary orders per attempt are 20 ÷ 200 × 100 = 10 percent. Its observed buyer conversion among the 80 mature contacted customers is 20 ÷ 80 × 100 = 25 percent. Dividing the recorded orders by all 120 contacts gives 16.7 percent orders per contact, also preliminary because 40 windows remain open. These named measures use different units and observation populations; changing the denominator does not improve results.
HubSpot distinguishes categories of sales activity measures and outcome measures. The repeated-dial example in this guide illustrates a separate point: another call to the same person can add activity without adding a unique customer. That is reasoning from this fictional example, not a rationale attributed to the source. Keep call events and distinct customer counts visible so readers know what is being measured.
Contact rate: the first outcome metric and its correct denominator
Contact rate answers the question: of all the calls we dialed, what share resulted in a real conversation? Its formula is straightforward.
Contact rate = (contacts ÷ attempts) × 100
The denominator here is attempts, not customers. This is intentional. Contact rate measures the efficiency of your dialing effort — whether your list quality, call timing, and channel are producing conversations. In the Malee Glow example, the team dialed 200 times. Of those 200 dials, 120 resulted in a live conversation with a distinct customer; the remaining 80 dials were unanswered repeat attempts to 40 customers who were never successfully reached. Contact rate: (120 ÷ 200) × 100 = 60 percent.
A low contact rate is an illustrative warning sign that something upstream may need attention — list quality, call timing, or channel selection — rather than a signal about script persuasiveness. As a heuristic example only: if most dials go unanswered, improving your script will not help because most of your audience never hears it. Diagnosing the correct layer before acting saves resources regardless of where your specific threshold sits.
Follow-up rate: the denominator shifts to contacts
Once a customer has been reached, the next question is whether the conversation produced a committed next step. Follow-up rate answers: of everyone we actually spoke with, what share agreed to a scheduled follow-up?
Follow-up rate = (follow-ups scheduled ÷ contacts) × 100
Notice that the denominator is now contacts, not attempts. This is because customers who were never reached cannot logically schedule a follow-up. Including uncontacted attempts in the denominator would artificially deflate the follow-up rate and make your team's conversational effectiveness look worse than it is.
In the Malee Glow example: 40 follow-ups were scheduled from 120 contacts, giving (40 ÷ 120) × 100 = 33.3 percent. If this number is low relative to your team's own baseline, the problem is in the conversation itself — perhaps the offer is unclear, the timing of the call within the customer's purchase cycle is poor, or the product being offered does not match the customer's recent purchase history. These are problems a better script, a different product selection, or a different customer segment may address. They are not problems a faster dialing rate will fix.
A follow-up also has a specific definition that must be agreed upon before tracking begins. A vague 'call me later' is not a scheduled follow-up. A logged appointment with a date and a customer identifier is. Without that definition, two team members will count follow-ups differently, making the KPI meaningless for comparison.
Purchases: mature-customer conversion and preliminary orders per attempt
Purchase reporting needs two clearly named measures here: observed buyer conversion among mature contacted customers, and preliminary orders per attempt across the whole calling week. Forty of the 120 contacts still have open seven-day windows. The earlier 80 have completed windows. Do not label a full-campaign snapshot as a final conversion rate while part of its observation window is unfinished.
Metric A, mature-customer buyer conversion = distinct buyers among mature contacted customers ÷ distinct mature contacted customers × 100. Here 20 ÷ 80 × 100 = 25 percent. Deduplicate people by customer ID within the defined cohort; extra conversations do not add extra people. In this fictional case each of those 80 customers had one live contact and each buyer placed one order, so the buyer and order numerators happen to match. If a buyer orders twice, retain one buyer for this conversion measure and report the extra order separately.
Metric B, preliminary orders per attempt = completed orders recorded so far ÷ all logged attempts in the calling-week cohort × 100. Here 20 ÷ 200 × 100 = 10 percent. This is an operational yield snapshot, not a final customer conversion rate. It remains preliminary until the original attribution windows for the entire cohort have closed; then report the completed-cohort order yield using the same counting and attribution rules.
The 25 percent among mature contacts is an observed outcome rate. It can be an outcome measure in a controlled script test, but alone it does not establish that a script caused purchases: assignment, audience, offers and observation windows must be comparable. The preliminary 10 percent orders per attempt is operational yield across the full calling week, not a standalone profitability rule. Deciding whether a campaign is worthwhile also requires order value, margin, staffing and telephony costs, discounts and other relevant costs. Compare the same metric over comparable completed windows; changing the denominator does not improve performance.
Salesforce's guidance on sales conversion rate notes that the denominator definition — the number of leads or prospects in the base — critically determines what the resulting percentage means. That denominator-choice principle applies at every funnel layer: define your denominator before you calculate, and state it alongside every published rate.
Sources: Salesforce: Sales conversion rate — formula and best practices
A complete fictional funnel: Malee Glow's weekly telesales numbers
This is one fictional teaching example; all figures are invented. The table lists four activity/outcome stages plus a separate cohort-status note. The 80 mature contacts are a filter on the 120 contacts for purchase-rate calculation, not a fifth funnel stage. A purchase does not require a follow-up appointment.
Definitions used in this example. Attempt: one logged outbound dial event, identified by a call-log entry with a timestamp and a customer ID. Multiple entries with the same customer ID count as multiple attempts. Contact: a call-log entry where the outcome field is 'answered' and a conversation of at least 30 seconds occurred with the named customer. In this example, 120 of the 200 attempts each reached a distinct customer in a live conversation; the remaining 80 attempts were unanswered repeat dials to 40 customers who were never successfully reached — producing no contacts. Follow-up: a calendar entry created during or immediately after a contact, with a scheduled date, a customer ID, and a responsible agent. The 40 follow-up appointments are a subset of the 120 contacts — these 40 customers both had a live conversation and agreed to a scheduled callback. Their purchase outcomes are not yet resolved at the time of reporting. Cohort-status filter: at this later reporting snapshot, the first 80 contacts are at least seven days old and their seven-day outcome windows have closed; the later 40 contacts are less than seven days old and their windows remain open. In this fictional example those later 40 also have pending appointments, but maturity depends on elapsed time, not on having an appointment. Purchase: a completed order, identified by a unique order ID in the order management system, where the order was placed within seven days of a telesales contact and the contact is recorded as the initiating event. In this fictional example, the 20 completed orders came from 20 distinct customers among the 80 mature contacts — none came from the 40 whose follow-up window is still open.
The four layers have these fictional counts: 200 attempts, 120 live contacts, 40 booked follow-ups and 20 completed orders from 20 buyers. Contact rate is 120 ÷ 200 = 60 percent; follow-up booking rate is 40 ÷ 120 = 33.3 percent. Applying the cohort filter gives observed buyer conversion of 20 ÷ 80 = 25 percent among the 80 mature contacted customers. Preliminary orders per attempt are 20 ÷ 200 = 10 percent across the full calling week. None of the 20 orders comes from the 40 pending contacts. Follow-up appointments are a separate branch, not a required step before purchase. A later order belongs to this example only within its original seven-day window; otherwise report it separately under a stated rule.
Two important caveats apply. First, if one customer had placed two separate orders, the purchase count by order ID would differ from the count of distinct purchasing customers — the team must decide whether their KPI tracks distinct order IDs or distinct purchasing customers. Both are valid; neither can be left undefined. In this fictional example each of the 20 purchasing customers placed exactly one order, so both counts are equal. Second, once the 40 follow-up appointments mature, some may convert and the final purchase-per-contact rate across all 120 contacts will likely differ from the interim 25 percent calculated on the 80 mature contacts alone. Clearly label any in-progress rate as preliminary.
| Funnel stage or cohort note | Count | Denominator | Rate | What to investigate, not a diagnosis |
|---|---|---|---|---|
| Call attempts | 200 | — | — | No signal at this layer alone |
| Contacts | 120 | 200 attempts | 60% | List quality, timing, or channel issue |
| Follow-ups scheduled | 40 | 120 contacts | 33.3% | Script, offer, or customer-fit issue |
| Cohort filter only: mature contacts (not a fifth stage) | 80 | 120 contacts | 66.7% of contacts are mature | 80 windows closed; 40 still open. Maturity is elapsed time, not appointment status. |
| Purchases within 7 days (from mature contacts) | 20 completed orders from 20 distinct customers | 80 mature contacts for 25%; 200 attempts for 10% | 25% per mature contact; 10% preliminary orders per attempt | Offer relevance, urgency, or follow-through issue |
The denominator trap: how mixing layers distorts every rate
The most common mistake in telesales reporting is not choosing the wrong formula — it is changing the denominator mid-analysis without realizing it. Here are four specific scenarios where this happens and what each one produces.
Scenario one: reporting 20 ÷ 200 = 10 percent preliminary orders per attempt while discussing the 20 ÷ 80 = 25 percent observed purchase outcome among mature contacts. The two measures use different populations and units. Neither proves that a bottleneck is the script, list or calling time. Investigate comparable records and use a controlled design before treating an observed difference as causal lift.
Scenario two: reporting follow-ups ÷ attempts instead of follow-ups ÷ contacts. This gives (40 ÷ 200) × 100 = 20 percent instead of 33.3 percent. The team looks like it is booking follow-ups at a low rate when in fact it is converting a third of every conversation into a committed appointment.
Scenario three: confusing distinct order IDs with distinct purchasing customers when counting purchases. If one customer placed two separate orders after one contact, the order-ID count would exceed the purchasing-customer count. Revenue impact (which scales with order count) differs from customer conversion (which scales with customer count). Neither figure comes from a call-log row count; purchases must be reconciled against your order management system.
Scenario four: including immature cohorts in a purchase rate. The 40 follow-up appointments in the Malee Glow example have not yet had time to convert. Calculating a purchase rate that treats those 40 contacts as resolved non-buyers would give (20 ÷ 120) × 100 = 16.7 percent — lower than the 25 percent that reflects only the 80 contacts whose outcome window has actually closed. Measure each funnel transition using only the population whose stage outcome is already known, and label any in-progress rate as preliminary.
Activity metrics versus outcome KPIs: a practical distinction
Attempts and raw contact counts are activity metrics. Contact rate, follow-up rate, and purchase conversion rate are outcome KPIs. This distinction matters because the corrective actions for each are completely different.
When an activity metric is below target — say, your team is making only 150 attempts per day against a target of 200 — the fix is operational: more agents, longer calling windows, better list preparation, or automated dialing tools. The solution lives in resources and scheduling.
When an outcome KPI is below target — say, your contact rate drops noticeably — the fix is investigative: why are fewer people answering? Has your number been flagged as spam? Is the call time wrong for your customer base? Did a list segment expire? The solution lives in list quality, channel selection, and call timing.
When your follow-up rate drops, the fix is conversational: the offer may have changed, the script may not address the right objection, or agents may be calling customers whose purchase cycle has not yet prompted a need. The solution lives in offer design, segment selection, and agent training.
HubSpot supports distinguishing activity measurement from outcome measurement. The different troubleshooting actions suggested in this guide are hypotheses to check against your records, not explanations established by that source. A low rate alone does not identify its cause; verify definitions, data quality and comparable observation windows before choosing an intervention.
Incomplete cohorts versus mature cohorts: a critical timing warning
Every purchase conversion rate calculation implicitly assumes that the observation window for all contacts in the denominator has closed. In practice, telesales teams call customers throughout the week, and a customer called on Friday may not purchase until the following Wednesday — well within a seven-day attribution window.
If you calculate your weekly purchase rate on Friday afternoon, every contact made on Thursday and Friday is in an immature cohort: their outcome window is still open. Including them in the denominator as if they are resolved non-buyers deflates the rate. Excluding them — or waiting until the attribution window closes before finalizing the rate — gives a more accurate picture.
The same logic applies to follow-ups. A follow-up appointment scheduled for next Tuesday is a pending outcome, not a resolved one. If you want to measure the purchase rate among customers who completed a follow-up call, you must wait until those appointments have occurred and their outcomes are logged before including them in any conversion calculation.
This is not a minor technical detail. In any week where a material share of contacts have pending follow-up appointments, reporting on all contacts as if their outcomes are known will consistently understate conversion performance. Teams may then never understand why their numbers always seem to improve when they look back at older cohorts. Clearly label any in-progress rate as preliminary and finalize cohort rates only after the attribution window has closed for every contact in the denominator.
What to do when each KPI layer underperforms
Low contact rate (below your team's own baseline or target). First, check whether your dialing number has been flagged by carrier spam filters — customers who see an unknown number labeled 'spam risk' are unlikely to answer. Second, audit the time-of-day distribution of your attempts and compare answer rates by time slot for your specific customer base, since optimal windows vary. Third, review list freshness: stale phone numbers with no recent purchase activity tend to have higher disconnect and wrong-number rates, though the exact threshold will differ by market. These are illustrative starting points for investigation, not guaranteed causes. The fix is list hygiene and dialing hygiene, not script revision.
Low follow-up rate (below your team's own baseline or target). First, listen to recorded contacts and identify the most common reason customers decline a follow-up. Is it 'I'll think about it and call you' — a sign the offer lacks urgency? Is it 'I already bought from somewhere else' — a sign the segment selection is too late in the purchase cycle? Is it 'I don't need this right now' — a sign the timing or the RFM segment is mismatched to the offer? Each answer points to a different fix: urgency design, earlier segment trigger, or a different product match.
Low purchase conversion rate among mature contacts (below your team's own baseline or target). This layer can conflate two different populations: customers who bought immediately and customers who scheduled a follow-up and then bought (or did not). Splitting these into separate sub-rates — immediate purchase rate among contacts who did not schedule a follow-up, and purchase rate among customers who completed a follow-up appointment — reveals whether your script's immediate close is weak or whether your follow-up execution is failing.
Low purchase rate among completed follow-ups specifically. If customers who keep their follow-up appointments still are not buying, the problem may lie in the follow-up call itself: the agent is not referencing the original conversation, the offer has changed without the customer being informed, or there is a checkout friction that the telesales channel cannot resolve. This is where integrating telesales logging with your order management system — so agents can confirm stock, apply a promised discount code, or complete the order on the customer's behalf — may make a measurable difference.
How CasperBrain fits into this process
CasperBrain is a retention analysis and planning product. It segments your customer base using RFM scoring — Recency, Frequency, and Monetary value — using four quartiles where 1 represents the best quartile and 4 represents the weakest. A composite score like '114' means the customer is in the best recency quartile, the best frequency quartile, and the weakest monetary quartile. This tells you something specific about who to prioritize for a telesales campaign and why.
Within CasperBrain, authorized users can export customer data as CSV files, with field visibility determined by account permissions. A telesales team planning a lapsed-customer reactivation campaign might work from a customer list filtered to customers with a recency quartile of 3 or 4 (have not purchased recently) and a frequency quartile of 1 or 2 (relatively higher-frequency buyers within this shop's data). Within that shop's distribution, those customers tend to have purchased more often than others — though the quartile alone does not tell you their absolute order count or guarantee they are not one-time buyers. Applying an additional filter on minimum order count, if your data supports it, can sharpen the selection further.
That segmentation shapes the input to your telesales funnel. A more tightly targeted input list may, in practice, produce better contact rates and follow-up rates than a broad or stale list — but this depends on your specific data quality and campaign execution, not on any automatic product behavior. CasperBrain produces the segment analysis and the export; it does not make the calls. Attempts, contacts, and follow-ups must be tracked in your call-logging system; purchases must be reconciled against your order management system and linked to contacts by customer ID. Report the counts for all four layers separately and give each KPI rate an explicit denominator.
Teams may choose to review their campaign results alongside CasperBrain segment data when planning the next campaign — for example, manually comparing which RFM segments produced higher contact or follow-up rates — but this review is a planning step your team performs, not an automatic ingestion or in-product loop. One important operational note on sequencing: RFM scores reflect purchase history up to the point the data was last prepared. If orders are recorded with any delay, a customer's segment may not yet reflect their most recent purchase. Always confirm the freshness of your underlying order data before treating an exported segment as the definitive input for a live campaign.
Troubleshooting common logging and reporting errors
Problem: the recorded contact rate is unexpectedly high relative to your own comparable baseline. A high rate is not automatically an error and there is no universal cutoff used here. Inspect whether voicemail, the wrong person or an immediate disconnect was counted as a live conversation with the intended customer. Agree on the contact definition before counting. A minimum duration such as 30 seconds is only an illustrative local logging rule, not an industry standard.
Problem: your purchase count does not match your order management system's count for the same period. Likely cause: telesales is counting distinct order IDs, while the order system counts distinct purchasing customers, or vice versa. One customer ordering twice produces two order IDs but one purchasing customer. Fix: agree on whether the KPI tracks distinct order IDs or distinct purchasing customers, document that definition, and reconcile both systems against it weekly.
Problem: your follow-up rate improves every week even though nothing in the script has changed. Likely cause: agents have learned that follow-up entries are counted in their performance score, so they are logging vague 'will call back' notes as scheduled appointments. Fix: require that a follow-up entry include a specific date, a customer ID, and a confirmed agent name — entries missing any field are excluded from the follow-up KPI.
Problem: this week's purchase rate is always lower than last week's rate when you look back at it. Likely cause: immature cohorts. When you calculated last week's rate, some contacts were still within their attribution window. By this week, those contacts' outcomes have resolved and some purchased, raising the retroactive rate. Fix: finalize cohort rates only after the attribution window has closed for every contact in the denominator. Clearly label any in-progress rate as preliminary.
Problem: two team members use different denominators under the same contact-rate name. In this guide, contact rate means live contact events ÷ call attempts. A person-level reach metric instead uses distinct customers reached ÷ distinct customers dialed, with both sides deduplicated over the same window. Live conversations per distinct customer dialed is another event-per-person yield and can differ again. Give each measure its own name and counted unit; do not publish them as interchangeable definitions of one contact rate.
A practical starting plan for shops setting up telesales KPI tracking for the first time
If your team currently tracks only a single 'conversion rate' from a combined call log, the following sequence gives you a concrete path to separating the four funnel-layer counts and their associated KPI rates without overhauling your systems at once.
Week one: add two fields to your existing call log. First, an outcome field with at least three values: no answer, contacted, and invalid number. Second, a follow-up field: yes or no, and if yes, a scheduled date. These two additions are enough to calculate contact rate and follow-up rate immediately.
Week two: link your call log to your order management system by customer ID. For every order placed within the attribution window following a contact, flag the order with the contact date and agent ID. This gives you the raw data for purchase conversion rate without changing your order system's structure.
Week three: run the four funnel-layer counts and their associated KPI rates in parallel with whatever you were reporting before. Compare them. The gaps between your old single rate and the new layered rates tell you exactly where your previous reporting was misleading you.
Week four onwards: set targets for each KPI separately. Use the first three weeks' baselines as your starting benchmarks. Review contact rate and follow-up rate weekly — they tend to respond relatively quickly to list and script changes. Review purchase conversion rate only after attribution windows have matured for the contacts in each cohort. Review the relationship between RFM segment and each KPI periodically — which customer segments produce the highest contact rates, the highest follow-up rates, and the highest purchase rates may not be the same segment, and that finding should shape your next campaign's targeting.
Fictional teaching example; all numbers are invented and use different units. These bars show counts, not one required sequence or comparable conversion rates. The 20 orders came from the 80 mature contacts, not the 40 pending follow-ups. Maturity filters the contact cohort; it is not another funnel stage. The 25% uses only mature contacts; the 10% yield across all attempts is preliminary while 40 contacts remain open.
Sources and editorial method
Examples are fictional and contain no customer records. Prepared by the CasperBrain team with AI drafting assistance; facts and language are checked before publication.