Sales Follow-Up Statistics: 15 Claims, 8 Untraceable

Search for sales follow-up statistics and the same four numbers come back at you from every result on the page. Eighty percent of sales need five follow-ups. Forty-four percent of reps quit after one. Ninety-two percent quit by the fourth. Ninety-five percent of converted leads are reached by the sixth call.

We tried to open the studies behind them. Fifteen claims went into the audit, and eight of them do not trace to any original research: seven have no locatable source at all, and one turned out to be arithmetic performed on another. The single most-quoted figure in B2B sales comes from a survey of fewer than 40 people conducted in 1942.

This page is the audit. Every claim below carries its original publisher, its year, its sample size and, most importantly, its denominator, because the number of leads you can reach on the fourth call and the number of already-converted customers who happened to take a fourth call are different measurements that get printed as the same sentence.

What the follow-up numbers say once you attach their scope

A follow-up statistic is only usable when four things travel with it: who published it, when the data was collected, how many observations it covers, and what population the percentage is calculated over. Strip any one of those away and the number becomes unfalsifiable.

Here is every claim in this audit that survived tracing, with its scope attached. Nothing in this table is an IVRIS measurement. These are third-party studies, reported with the boundaries their authors set.

Two things stand out. The only claim that describes contact rates by attempt number is conditioned on conversion, and the only claim that describes a full five-to-twelve contact curve is 84 years old. There is no modern, independent, published curve of reachability by attempt number in B2B. That absence is the actual state of the evidence.

This page measures what happens across attempts two onward. How fast the first attempt should happen, and the eighteen different clocks the industry calls speed to lead, is audited separately in our speed-to-lead statistics review, and the operating model for setting those targets sits on speed to lead.

Five different things are all called a follow-up

Most contradictions in this topic dissolve once you separate the units. A touch, an attempt, a contact and a reach are four different events, and the studies quoted alongside each other rarely count the same one.

The last row is where most damage happens. Counting backwards from closed deals cannot tell you anything about the leads that never closed, because they are not in the sample. A statistic built that way describes the history of your winners, not the odds facing your reps on Monday morning.

The 1942 survey behind “80% of sales need five follow-ups”

The most repeated persistence statistic in B2B sales has a traceable origin, and it is older than the transistor. Sales & Marketing Executives International, the organisation that published it, documents the provenance on its own site.

In 1942, the Long Island, NY chapter of NSEA (now SMEI) surveyed their members to determine the ratio of calls made to sales made.

According to SMEI’s own account, published by Willis Turner in August 2021, the survey produced the familiar ladder: 2% of sales on the first contact, 3% on the second, 5% on the third, 10% on the fourth, and 80% across the fifth to twelfth. The same page adds the detail that no roundup carries: “Another thing that might surprise you is that the sample size was less than 40.”

A “call” in 1942 meant a salesperson physically visiting a business. The survey measured the ratio of visits made to sales made, self-reported by fewer than 40 members of a single local chapter, in an economy without direct-dial long distance, let alone email. It is not a measurement of B2B email touches, and it was never presented as one.

SMEI is not defending the figures. The page exists to caution readers, asking directly: “Who did the research? Is the research unbiased? Was the research done recently, or is it out of date? Was the sample size meaningful?”

One further correction is worth recording, because published debunkings of this statistic have concluded that the National Sales Executive Association never existed. SMEI’s own site states that it was “named in earlier days as National Sales Executives Association.” The association is real and still operating. What does not exist is a modern study by it. The 1942 chapter survey is the whole evidentiary base.

Follow-up statistics that do not survive an audit

Eight of the fifteen claims failed. Five could not be traced to any original study, one is arithmetic on another, and two are traceable but consistently misquoted.

Claims with no locatable original

That last detail is worth pausing on. Six separate statistics on the current top-ranking result for this keyword carry the string “(source pending verification)” in production, including the 44% figure and the voicemail pair. An internal editorial note asking a colleague to check a number has been published as though it were the number’s citation, and it has been live long enough to be indexed and quoted onward.

The chain behind the headline claim is just as short, and it can be walked in two clicks. That same page sources “80% of sales require at least 5 follow-up contacts before closing” to invespcro.com. That page carries twelve follow-up statistics, including this one, with no citations, no footnotes, no bibliography and no outbound links of any kind. The chain terminates there, because there is no study at the end of it to point at.

The claim that is arithmetic, not research

“92% of sales reps give up after four attempts or fewer” is presented everywhere as an independent finding, usually one sentence away from the 44 / 22 / 14 / 12 ladder. Add that ladder up: 44 + 22 + 14 + 12 = 92. The two are the same untraceable artefact, printed twice, and quoting them together does not corroborate either. It double-counts a claim nobody has sourced once.

Claims that are traceable but misquoted

The sixth-call statistic is the clearest case in this audit, because the original is specific and the circulating version is wrong in three ways at once. Velocify’s 2013 study analysed nearly 3.5 million leads from more than 400 companies, generated in the first half of 2012 and tracked through the third quarter, and reported that “93% of all converted leads are contacted by the 6th call attempt.”

It circulates as 95%. It is frequently credited to InsideSales or to Inc. rather than to Velocify. And the denominator is dropped entirely.

The study is explicit that it describes converted leads: “only 7% of all leads that eventually convert are reached for the first time after seven or more call attempts.” Because the population is defined by having converted, the figure cannot tell a rep how likely a sixth call is to reach anyone. Both the first and third results currently ranking for this keyword publish it as 95%, one of them crediting a publisher that did not produce it.

The second misquote is subtler. RAIN Group’s prospecting research, based on 488 buyers representing $4.2 billion in purchases across 25 industries plus 489 sellers, is regularly cited for “8 touches to secure a meeting” alongside the untraceable “8 cold call attempts to reach a prospect.” Those are different constructs. Touches to secure a meeting counts outbound actions toward an outcome; attempts to reach counts dials until someone answers. We could confirm the sample on RAIN’s public page but not the touch figure or the fieldwork year, both of which sit behind a download gate, so this claim is recorded as partially traced.

What modern platform data measures by attempt number

Three vendors publish genuine first-party datasets with methodology attached. Each measures something real. None of them measures the same thing.

Email sequence data

Belkins analysed 7,530,489 emails sent across twelve months of 2025, alongside 112,005 LinkedIn invite sequences from 33 client projects and more than 46,000 call records. It defines its reply rate as unique replies divided by emails sent, excluding auto-replies and bounce notifications, and it disabled open tracking for the period rather than report a metric it considered unreliable.

Per-step reply rates are low and flat: 0.59% at step one, 0.30% at step three, 0.34% at step four, 0.28% at step six. The aggregate is the interesting part. Steps two through six together account for 58.6% of all replies, against 41.4% for the opening email, and step three alone drives 35.6% of email-sourced meetings, more than steps one and two combined.

Woodpecker, drawing on more than 20 million emails, reports a 3.43% average reply on the first email, campaigns with three to five follow-up steps replying at 8.3% against 4.1% for sequences with no follow-up at all, and an optimum of four to seven total touchpoints. It also states plainly that Apple Mail Privacy Protection now accounts for roughly 49% of recorded opens, which is why open rate is “a directional signal in 2026, not a precise performance measure.” Woodpecker does not state the collection window for its dataset, which is the one gap in an otherwise well-documented set. Channel-specific execution for this sits in our guide to cold email follow-up.

Dial and connect data

Cognism logged 451,895 calls against 196,470 prospects across the UK, US, EMEA, DACH and France during 2025, from 110 account executives. SDR cold calls were answered 13.3% of the time and AE calls to warm contacts 14.4%. Teams averaged 3.36 touches per prospect, with calls making up 57% of all tasks, LinkedIn 27% and email 15%.

Cognism attaches the caveat almost nobody else in this field does: “All data reflects Cognism’s internal outbound teams and may not generalise across all industries.” That single sentence is worth more than most of the statistics it sits underneath, and it is the correct way to publish first-party data.

Why these results cannot be averaged into one persistence benchmark

It is tempting to combine the surviving numbers into one cumulative reach curve, plot attempts one to twelve, and call it the answer. It cannot be done honestly, and the reason is the denominators.

Velocify counts call attempts against leads that converted. Belkins counts scheduled email steps against emails sent. Woodpecker counts campaign touchpoints against campaigns. Cognism counts dials against prospects at one company.

That is four numerators, four denominators, three channels, and collection windows thirteen years apart. Averaging them produces a figure with no population behind it, which is the exact mechanism that generated the folklore this page is auditing.

What can be done is to plot each source separately, on its own axis, with its own sample and denominator printed alongside it. That is the honest version of the chart everyone wants, and the fact that the panels do not line up is the finding rather than a failure of the data.

How to build your own reach-by-attempt curve

Since no public dataset provides a reachability curve by attempt number, the only trustworthy version is the one you measure from your own CRM. It is a smaller job than it sounds, and the output is a benchmark that actually describes your market rather than someone else’s platform.

1. Fix the denominator before you query anything. Decide whether you are measuring dials per lead, sequence steps per contact, or interactions before closed-won. Write the definition down first, because every disagreement about persistence data is really a disagreement about this.
2. Count attempts per lead, not per campaign. Campaign-level averages hide the distribution completely. A team averaging four touches may be making one touch on most leads and twelve on a handful.
3. Keep unreached leads in the sample. This is the single error that produced the 1942 ladder and the sixth-call statistic. If a lead that was never reached drops out of your denominator, your curve will describe your winners and forecast nothing.
4. Split by channel before you interpret. A dial and a sequence step are not interchangeable units, and combining them is how incompatible benchmarks get created in the first place.
5. State the collection window. Connect rates move with regulation, spam filtering and buyer behaviour. A curve without a date is a curve with a shelf life you cannot see.
6. Publish the sample size next to the result. If the number is going to be quoted internally, it needs the same scope you would demand of an external study.

Run that once and you will have something none of the pages competing for this keyword can offer: a persistence benchmark with a known population, a known channel and a known date.

How to quote each statistic safely

These claims are not all unusable. Most become usable the moment they are stated with their scope. This register gives the safe wording for each surviving claim and names what it cannot support.

Methodology and revision history

We assembled 15 statistics circulating in the top results for this topic, including every claim reproduced in Google’s AI Overview for this query as of 1 August 2026. For each one we attempted to locate the original publisher, the publication year, the sample size, the channel measured and the population the percentage was calculated over, working backwards through the citation chain rather than accepting the first attribution encountered.

A claim is recorded as traced when the original publisher and its methodology could be opened and read; partially traced when the publisher is identifiable but key scope details sit behind a gate; misquoted when an original exists and the circulating figure or attribution differs from it; and untraceable when no original could be located by any route available to us.

Of the 15, six trace to an original study, one of which is misquoted in nearly every place it currently appears. One is partially traced. One is arithmetic performed on another claim. Seven could not be traced to any original at all. Seven of the fifteen disclosed a sample size, and only five disclosed when the underlying data was collected. IVRIS did not conduct any of the studies described on this page and does not present any figure here as its own measurement.

Two sources could not be opened from our verification environment on 1 August 2026: marketingdonut.co.uk failed DNS resolution on repeated attempts, and one widely referenced 2014 debunking returned HTTP 403. Neither is quoted here, and no claim on this page rests on either. Where a source could not be read, that is recorded as a limitation rather than filled with a secondary account.

Published 1 August 2026. This page will be revised as new first-party datasets are published or as any currently untraceable claim is resolved. The URL will not change.

Frequently Asked Questions