WHO OWNS WHAT  //  six jobs move, one never does, and three of our eight recommendations are not usApply
Done-for-you outbound

Done-for-you outbound, and whether you should buy it

Updated 12 August 2026  //  by Mark Glazer, ReplyLead  //  editorial standards

The short answer: done-for-you outbound is one team owning six of the seven jobs in an outbound programme: targeting, data, infrastructure, copy, sending and reply handling. The seventh, closing, stays with you in every honest version of the arrangement. It is worth buying when three or more of those six have no owner today, your first-year contract value is above roughly $5,000, your ICP runs to a few thousand companies and somebody can take six or more discovery calls a month. When those are not all true, the tools below will tell you so and say what to do instead.

Build your outbound machine

Four questions about your business and who currently owns each job, and this returns one of eight recommendations. Three of them say do not run outbound at all, and two more send you somewhere that is not an agency. The rule that produced your answer is published below the tool, in evaluation order, so you can check it rather than trust it.

  1. 1. Business
  2. 2. Today
  3. 3. Ownership
  4. 4. Result
Step 1. Your business
What one new customer is worth in the first 12 months.
Your reachable market, not the total market.
Calls somebody can actually run, not calls you would like.
What you sell
Used to phrase the result, never to change the arithmetic.
Step 2. How outbound runs today

Picking one pre-fills step 3 with the ownership pattern we see most often for it. Change anything that is wrong.

Step 3. Who owns each job

Six delivery jobs plus the one that never moves. Software counts as an owner for data, infrastructure and sending. It does not count for targeting, copy or reply handling, because a licence cannot decide who to contact, cannot write your angle and cannot qualify a human being: that rule is stated in the table below the tool.

Closing is in the list so you can try assigning it to an agency. No honest provider owns that row: an outside team can produce a qualified conversation, it cannot run your sales call or carry your commercial terms. Any provider that claims it is selling something other than outbound.

Step 4. The recommendation

Based on: $25,000 first-year contract value, 10,000 companies in reach, 20 discovery calls a month, and 3 of 6 delivery jobs with no owner.

Recommended, and it is not us

Outbound is not the right instrument yet

Every model on this page ends at a meeting somebody has to run, and you have told us nobody can run one. Booked meetings that go unrun are the most expensive outcome in outbound: you pay to create them, then pay again in the reputation of a no-show.

What to do instead. Clear calendar capacity first, even two hours a week, then come back. Nothing else on this page matters until that is true. If capacity is permanently the constraint, the honest answer is that outbound is not your bottleneck and spending on it will not move revenue.

Recommended, and it is not us

Outbound is not the right instrument yet

Outbound at scale is a coverage instrument. With a list this small you would exhaust the market inside a month and spend the rest of the year re-contacting the same people, which is how sending reputations are destroyed.

What to do instead. Reach these companies deliberately: research each one, use a warm path, and treat the list as a named-account plan rather than a campaign. That is a founder-led or an account-based motion, and it will beat any agency on a list this size.

Recommended, and it is not us

Outbound is not the right instrument yet

The arithmetic does not clear the floor. The cheapest published infrastructure tier is $2,023 a month at 25,000 emails, which on our measured constant buys roughly 6 qualified meetings a month: about $332 of infrastructure per qualified meeting before any service fee, any margin and any of your own selling time.

What to do instead. Either raise the first-year contract value, by bundling, annualising or moving up-market, or reach these buyers through a channel with a lower cost per conversation. Cold email is not cheap; it is just cheaper than a salesperson at scale, and that only helps when the deal is worth having.

Recommended, and it is not us

Hire an internal SDR

A small, high-value market rewards depth over coverage, and depth is the one thing an outside team cannot buy. One person can hold a few thousand contacts in a quarter, will learn your product properly and will out-converse any agency on a technical first call. Paying for coverage you do not need is the expensive mistake in this quadrant.

What to do instead. Hire, and price the hire honestly: the tooling does not disappear when you do. Put your own salary, ramp and volume numbers through the SDR cost calculator, which shows every line of the arithmetic, then read alternatives to hiring an SDR for the comparison in full.

Recommended, and it is not us

Buy software, not a team

You already own the scarce input, and it is not software: it is attention. Four or more of the six delivery jobs have a real internal owner, so hiring a team to run them again is paying twice for the same work. The licence is the cheapest line in an outbound programme and the least of the work, which is exactly why it is the right purchase for you.

What to do instead. Buy the sending platform, keep the people you have, and spend the difference on the two jobs software genuinely cannot do: deciding who to contact, and reading the replies. If a programme you run stalls later, it will almost certainly stall on deliverability or reply handling rather than on the tool.

Recommended

Done-for-you outbound

There is a genuine capability gap rather than a preference: three or more of the six delivery jobs have no owner at all. The market is large enough to be worth systematic coverage, the contract value carries the programme cost, and somebody can take the meetings. That is the case done-for-you outbound exists for, and it is the only combination on this page where we would argue for ourselves.

What it would look like. We take targeting, data, infrastructure, copy, sending and reply handling. You take the meeting and the deal. Our infrastructure amount is published by volume tier on pricing, and the rest of our compensation is a share of revenue that actually closes, so in a quarter where nothing closes no share is owed.

Want us to build this for you?

Recommended, and it is not us

A hybrid split, not a full build

You are not missing a team, you are missing two or three specific things, and in almost every case they are the unglamorous ones: verified data and sending infrastructure. Buying the whole chain would mean paying somebody to redo the targeting and the copy you already do well, and those are the two jobs that most reward knowing your own market.

What to do instead. Buy the gap, not the chain. Keep targeting and copy in house, put domains, mailboxes, warmup and live verification on somebody whose full-time job that is, and keep every reply in front of a human being on your side. If you would rather hand over the whole thing later, that door stays open; it does not work in reverse.

Recommended

Done-for-you outbound, at pilot scale

The fundamentals are sound, but at least one threshold is marginal, so a full build would be buying certainty the numbers do not support yet.

What to do instead. Run the smallest version that can still produce a readable result. One sequence, one segment, enough volume to clear the noise floor: below roughly 500 contacted leads a reply rate is not a measurement, which is why our own variance study excludes campaigns under that size. Get a real number first, then decide.

Whichever of these is true of your answers is the marginal one:
  • fewer than three of the six delivery jobs are unowned
  • contract value is under $5,000
  • the ICP list is under 2,000 companies
  • call capacity is under six a month

Want us to build this for you?

What would change this answer.
  • If nobody on your side could take the calls, this becomes Outbound is not the right instrument yet.
  • If your ICP list were under 500 companies, this becomes Outbound is not the right instrument yet.
  • If your first-year contract value were under $2,000, this becomes Outbound is not the right instrument yet.
  • If your ICP were 5,000 companies or fewer at $25,000 or more a deal, this becomes Hire an internal SDR.
  • If four or more of the six delivery jobs already had an internal owner, this becomes Buy software, not a team.
  • If three or more delivery jobs were unowned, at $5,000 a deal, 2,000 companies and six calls a month, this becomes Done-for-you outbound.
  • If only one or two jobs were missing and two already had an internal owner, this becomes A hybrid split, not a full build.
  • If the fundamentals held but one threshold were marginal, this becomes Done-for-you outbound, at pilot scale.

All four steps are shown above because JavaScript is not running. The recommendation is the one produced by the values shown, which are the defaults. The complete rule set is in the table below, so the answer for any other combination can be read off directly.

The rules this tool runs, in evaluation order

The first rule that matches wins. Nothing else is consulted. We publish it because a recommendation engine that always recommends the company that built it is a brochure, and the only way to prove this one is not is to let you audit it.

The gap count used by rules 5 to 8 counts a job with no owner at all, and also counts software placed on targeting, copy or reply handling.
#Fires whenVerdictWhy that is the honest answer
1Discovery calls you can take per month is 0Not yet. No one can take the meetingsEvery model on this page ends at a meeting somebody has to run. Booked meetings that nobody runs are the most expensive possible outcome, because you pay for them twice: once to create them and once in the reputation of a no-show.
2Companies matching your ICP is under 500Not yet. The market is too small for volumeAt the median measured reply rate of 2.54 percent per contacted lead, 500 companies contacted once return about 13 replies in total, ever. Outbound at scale is the wrong instrument for a market you could write down on one page. Reach those companies deliberately instead.
3Average first-year contract value is under $2,000Not yet. The contract value cannot carry the machineThe cheapest published infrastructure tier is $2,023 a month at 25,000 emails, which on the measured constant of one qualified meeting per 4,100 emails buys about 6.1 qualified meetings a month. That is roughly $332 of infrastructure per qualified meeting before any service fee, any agency margin and any of your own selling time.
4ICP is 5,000 companies or fewer AND contract value is $25,000 or moreHire. Internal SDRA small, valuable market rewards depth over coverage. One person can hold a few thousand contacts in a quarter, learns the product properly and will out-converse any outside team on a technical first call. Buying coverage you do not need is the expensive mistake here.
5Four or more of the six delivery jobs already have a real internal ownerBuy a tool. Software-ledYou already own the scarce input, which is attention, not software. The licence is the cheapest line in an outbound programme. Hiring an agency to run steps you are already running well is paying twice for the same work.
6Three or more of the six delivery jobs are unowned AND contract value is $5,000 or more AND ICP is 2,000 companies or more AND you can take 6 or more calls a monthOutsource. Done-for-you outboundThere is a real capability gap, the market is big enough to be worth systematic coverage and the contract value carries the cost. This is the case done-for-you outbound exists for.
7One or two jobs are unowned and at least two already have an internal ownerSplit it. HybridYou are not missing a team, you are missing two or three specific things, most often infrastructure and data. Buying the whole chain would mean paying an agency to redo work you already do well.
8Anything that reaches this lineStart small. Done-for-you outbound, at pilot scaleThe fundamentals are sound but at least one threshold above is marginal, so the honest recommendation is the smallest version of the programme that can still produce a readable result, not a full build.

The eight stages, and who owns each one

Done-for-you outbound is not a service, it is a list of jobs that move across a line. This is the line. For each stage: what actually happens, who normally owns it, the failure we see most, the number that tells you it is working, and what we do. Closing never moves.

Stage 1 of 8

Targeting

Deciding which companies and which job titles are worth an email at all, and writing that down as a rule a list can be built from.

Usually owned by
The founder, almost always, because it is the only step that requires knowing why anyone buys.
Where it fails
A definition so wide it cannot be wrong. "B2B companies in North America" is not a target, it is a permission slip to email anyone.
The metric
Share of replies that are from the intended title. If most replies come from people who cannot buy, targeting is the fault, not copy.
How we run it
We write the ICP rule with you and it is the one thing we will not change without asking, because every other number on this page is downstream of it.
Stage 2 of 8

Data

Turning that rule into verified contact records: finding the people, finding the addresses, and proving the addresses are real immediately before sending.

Usually owned by
Nobody, in most stalled programmes. A list gets bought once and then decays.
Where it fails
Trusting a vendor's "verified" flag. Addresses go dead continuously, and a flag written months ago is a claim about the past.
The metric
Bounce rate. Across the 248,443 sends in our 73-campaign set it was 1.44 percent of sends.
How we run it
Live verification pass immediately before each send, never a stored flag. It is the single cheapest insurance in outbound.
Stage 3 of 8

Infrastructure

Sending domains, mailboxes, authentication, warmup and the sending platform itself: the machinery that decides whether a well-written email is ever seen.

Usually owned by
A tool, in name. In practice nobody owns it until something breaks.
Where it fails
Sending from the company's primary domain. A deliverability problem then becomes an everyone-cannot-receive-our-invoices problem.
The metric
Inbox placement, and the mailbox count a given daily volume needs. Each mailbox needs two to two and a half weeks of warmup before it carries volume; the thresholds we run to are on deliverability.
How we run it
Dedicated domains and mailboxes that are not your primary domain. Across the seven programmes measured here that is 372 sending domains and 1,149 mailboxes.
Stage 4 of 8

Copy

The angle, the first email, the follow-ups, and the test that decides which of them survives.

Usually owned by
The founder writes version one. Then it is never rewritten, which is the actual problem.
Where it fails
Writing about yourself. The second most common is writing five variants of the same idea and calling it a test.
The metric
Reply rate per contacted lead, read against a distribution rather than a single number. Our measured spread across 47 campaigns runs 0.56 to 7.83 percent.
How we run it
We write it, test it and keep the sequence in place long enough for the result to mean something.
Stage 5 of 8

Sending

Actually putting the volume out, every working day, at a rate the infrastructure can carry.

Usually owned by
Software. This is the one step a licence genuinely does own.
Where it fails
Treating list size as the constraint. Across the 67 of our campaigns with more than one sequence step, 247,812 sends split into 60,311 first touches and 187,501 follow-ups: 75.7 percent of everything sent was a follow-up. Reaching 10,000 more people costs roughly four times 10,000 sends, not 10,000.
The metric
Sends per contacted lead. We measured 1.75 across the 73-campaign set.
How we run it
Held steady. Volume that lurches is a deliverability event waiting to happen.
Stage 6 of 8

Reply handling

Reading every reply by hand, answering the real ones, and deciding which are worth your calendar.

Usually owned by
Nobody, and this is where most programmes quietly fail.
Where it fails
Automating it. A reply is the first moment a human being has volunteered attention, and it is the worst possible place to save money.
The metric
Replies per sales-qualified lead. On the 12-month programme it took 19 replies to produce one.
How we run it
Every reply is read by a person. It is the most expensive part of what we do and the reason the model works.
Stage 7 of 8

Meeting

Getting a qualified, interested person to accept a specific time, and getting them to turn up.

Usually owned by
Split. The agency books it, your team runs it.
Where it fails
Booking anything with a pulse. A meeting that clears a loose definition still consumes an hour of your sales capacity.
The metric
Emails per sales-qualified lead. Measured at one per 4,100 on the 12-month programme. Note what that counts: a decision maker who accepted a specific time. It is an accepted meeting, not a held one, and we publish no show-up rate because we have not measured one. Appointment setting covers the bar in full.
How we run it
We qualify against your written criteria before a time is offered, and a meeting that does not meet the bar is not booked.
Stage 8 of 8

Closing

Running the call, handling the objection, and winning the deal.

Usually owned by
You. In every honest version of this arrangement.
Where it fails
Nobody available to take the call. Booked meetings that go unrun are the most expensive outcome in outbound.
The metric
Close rate of held meetings, which is yours to measure and ours to have no opinion about.
How we run it
We do not do this, we do not price it, and any provider claiming this row is selling something else.

The five ways this gets done

No interaction needed. The honest difference between these is not price, it is who does the work and who is out of pocket when a quarter produces nothing.

Founder-led

You do all of it
Who does the work
You, in the gaps between everything else
Cash cost
Infrastructure only, from $2,023 a month
First meetings
As soon as you find the time, which is the problem
Breaks when
You get busy. Outbound stops the week a deal needs closing
Best when
Pre-product-market-fit, when the conversations are the research

Software only

You operate a tool
Who does the work
You, faster
Cash cost
Same infrastructure, from $2,023 a month
First meetings
Two to three weeks, after warmup
Breaks when
Targeting, deliverability or replies. Never the tool itself
Best when
You already own targeting, copy and reply handling

Internal SDR

You hire one person
Who does the work
One person, for the hours they are awake
Cash cost
$80,000 on-target plus employer cost plus the same infrastructure
First meetings
Months. Hiring, then ramp
Breaks when
They leave, and the ramp clock restarts
Best when
A small, high-value market that rewards depth over coverage

Retainer agency

You buy a team
Who does the work
Their team, on their process
Cash cost
A fixed monthly fee. Only two of seven providers we checked publish one
First meetings
Three to six weeks
Breaks when
Results dip. The invoice does not
Best when
You want capacity now and can carry the performance risk

Performance-aligned

You buy a team, and share the outcome
Who does the work
Their team, on their infrastructure
Cash cost
A published infrastructure amount plus a share of revenue that closes
First meetings
Three to six weeks. Warmup is not negotiable for anyone
Breaks when
Attribution is vague. Agree it in writing before anything is sent
Best when
The deal is worth enough that a share of it beats a retainer for both sides

Cost, workload and who carries the risk

Four commercial models on the same volume assumption. Change anything you disagree with. For the build-versus-buy decision on its own, in more depth and with the breakeven, use the SDR cost calculator; this one exists to put four models beside each other on the dimensions that are not cash.

Your assumptions

Sets the published infrastructure tier.
Default: US median on-target earnings, The Bridge Group, 2025 measured
Payroll tax, benefits, equipment judgement
Default: the lowest retainer floor published by any of the seven providers we checked on 1 August 2026. Replace it with your actual quote.
Annual figures. Cost per qualified opportunity uses the same conservative measured constant for every model, so the comparison is like for like: one qualified meeting per 4,100 emails, from a single 12-month programme. Your own conversion will differ, which is exactly why the row below it exists.
 In-house SDRSoftware onlyRetainer agencyPerformance-aligned
Annual cash cost$134,092$34,092$48,000$34,092 plus a share of revenue that closes
A quarter with nothing closed$33,523$8,523$12,000$8,523
Cost per qualified opportunity$916$233 plus your time$328$233 plus the share
Your workloadManage, coach, review. Not modelled hereYou run all six delivery jobsApprove and take meetingsApprove and take meetings
Time to first meetingHire plus ramp, measured in monthsTwo to three weeks after warmupThree to six weeksThree to six weeks
Who owns infrastructureYouYouUsually them, askThem, on domains that are not yours
Who carries performance riskYouYouYouShared
What a quarter with no closed revenue costs, by commercial modelFour horizontal bars showing the cash a single quarter costs when nothing closes, for an in-house SDR, software only, a retainer agency and a performance-aligned model, at the volume and salary assumptions currently selected.In-house SDR$33,523Software only$8,523Retainer agency$12,000Performance-aligned$8,523
Cash cost of one quarter in which no revenue closes, at the inputs above. This is the row the pricing model actually decides: under a retainer a closed-lost quarter costs exactly what a record quarter costs. Under a performance-aligned model the infrastructure amount remains, because mailboxes, domains and verification are real costs either way.

Every line of the arithmetic. Infrastructure = the published tier at 50,000 emails a month, $2,841 x 12 = $34,092. In-house SDR = on-target earnings plus employer cost plus that same infrastructure = $80,000 + $20,000 + $34,092 = $134,092. Retainer agency = $4,000 x 12 = $48,000. Sales-qualified leads a year = 50,000 x 12 / 4,100 = 146. Cost per qualified opportunity is each row above divided by that count. The performance-aligned column shows the fixed floor only: the revenue share is a percentage of revenue that actually closes, it is agreed per client, and inventing a number for it here would be the exact thing this page criticises.

Two of these four have the same fixed floor. Software-only and a performance-aligned model both sit on the same infrastructure line, because it is the same machine either way: the mailboxes, domains, warmup, verification and sequencer do not become cheaper when you operate them yourself. What you are choosing between is not the cost of the machine, it is who runs it and who is out of pocket when a quarter produces nothing.
The seven published infrastructure tiers this uses
Published monthly technology cost by sending volume, itemised line by line on what a cold email agency costs. These are our own list prices, not an estimate of the category.
Emails a monthMonthly total
25,000$2,023
50,000$2,841
75,000$4,315
100,000$5,531
250,000$12,036
500,000$22,493
1,000,000$44,090

If the performance-aligned column is the one you keep looking at, the next step is a conversation about attribution, not a demo. Tell us the deal size and the ICP and we will map it.

Apply and choose a time

What a month of sending actually returns

This returns a range, not a number, because our own measured campaigns range. Every rate below is a published first-party figure and every denominator is named. Nothing here is a projection of what your campaign will do; it is what ours did.

The published volume tiers, so this lines up with the cost table above.
28,520Unique leads contacted a month, at the measured 1.75 sends per contacted lead
724Replies at the median campaign, from 582 at the 25th percentile to 1,010 at the 75th
12.2Sales-qualified leads a month, Route A. The number to plan on
38.1Sales-qualified leads a month, Route B. The optimistic route, shown so you can see the disagreement
Reply rate per contacted lead across 47 ReplyLead campaignsA box plot on a nought to eight and a half percent axis. The whiskers run from a minimum of 0.56 percent to a maximum of 7.83 percent. The box spans the twenty fifth percentile at 2.04 percent to the seventy fifth at 3.54 percent, with the median marked at 2.54 percent. A separate dashed marker at 0.93 percent shows where the single twelve month programme quoted elsewhere on this site would sit: between the minimum and the twenty fifth percentile.0.562.042.543.547.830%2%4%6%8%0.93% - the one 12-month programme quoted across this siten = 47 campaigns with 500 or more contacted leads, of 73 with sends, across 7 client programmes
Reply rate per contacted lead. Each campaign is counted once. The box is the middle half of the campaigns, the line inside it is the median, and the whiskers are the observed minimum and maximum, not a confidence interval. The best campaign in the set beat the worst by about 14 times under one operator on one methodology. Source and full method: cold email benchmarks, pulled 12 August 2026.
Two published routes to a monthly qualified-meeting estimateTwo horizontal bars comparing the qualified meetings a month implied by each of two independently published ReplyLead datasets at the selected sending volume. The lower estimate is the one to plan on.Route Aone programme12.2 a monthRoute B73 campaigns38.1 a month
Route A divides sending volume by the measured constant of one qualified meeting per 4,100 emails, from the 12-month programme. Route B runs the volume through the 73-campaign median reply rate and then through the measured 19 replies per qualified meeting. They disagree because the 12-month programme replied at 0.93 percent per lead, near the bottom of the campaign distribution above. Plan on Route A.
Every row is a published count or a single division of published counts. The two datasets are kept apart: rows 1 and 2 come from the 73-campaign multi-client set, row 3 from the single 12-month programme. Row 3 is the only place they are joined, and it is the only estimate on this page that crosses datasets.
StepRate usedWhere it comes fromDenominator
Emails sent to leads contacted1.75 sends per contacted lead73 campaigns, 7 client programmes, pulled 12 August 2026248,443 sends / 141,713 contacted
Leads contacted to replies2.04% / 2.54% / 3.54% (25th, median, 75th)The 47 campaigns that contacted 500 or more leadsunique replies / leads contacted, per campaign
Replies to qualified meetings1 per 19 repliesOne portal-tracked 12-month programme5,028 replies / 263 qualified
Emails to qualified meetings (Route A)1 per 4,100 emailsThe same 12-month programme, measured end to end1,080,000 emails / 263 qualified
Why the two routes disagree, and which to believe. Route A is one programme measured end to end, and that programme replied at 0.93% of the leads it contacted, which sits between the minimum and the 25th percentile of the 47-campaign distribution above. It is a low-percentile programme, so a route built on it is conservative by construction. Route B uses the median of a much wider set, but that set has no measured qualification step at all, so it has to borrow the 19-replies-per-meeting ratio from Route A's programme, where reply quality may be quite different. Plan on Route A. If a provider quotes you something closer to Route B, ask which of those two datasets it came from and what the denominator was.

The full method, the definitions and the things these datasets deliberately do not report are on cold email benchmarks. The short version: a reply is any human response including a no, a qualified meeting is a decision maker who matched written criteria and accepted a specific time, and no figure on either page is modelled, weighted or extrapolated.

Decode a reply rate somebody quoted you

The two most-cited public cold email datasets report average reply rates of 0.45% and 3.7%. That is roughly an eightfold gap between two competent studies, and almost all of it is definition rather than performance. This converts a quoted rate between denominators and shows where it would sit in our own measured campaigns.

A percentage. Enter 2.5 for two and a half percent.
What it is measured against
If they cannot answer this, you have learned the more important thing.
Default 1.75 is our own measured figure across 73 campaigns (248,443 sends / 141,713 contacted). Theirs will differ, and this is the number that moves the answer most. measured
2.50%Equivalent rate per lead contacted
1.43%Equivalent rate per email sent
between our 25th percentile and our medianWhere the per-lead figure sits against our 47 campaigns with 500 or more contacted leads

The arithmetic. A rate per lead contacted divided by the emails sent per lead gives the rate per email sent: 2.50% / 1.75 = 1.43%. Multiply to go the other way. A rate per email sent multiplied by the emails sent per lead gives the rate per lead contacted: 2.50% x 1.75 = 4.38%. Divide to go the other way. Nothing else changes: the same replies are being divided by a different denominator.

What conversion cannot fix. This assumes both sides count the same thing on top: distinct leads that replied. If their number counts total reply messages, or counts a thread twice, or excludes automatic replies while yours includes them, no arithmetic reconciles the two. Ask for the numerator and the denominator in the same sentence. Our own definitions are on the reply rate benchmark and every metric on this site is defined on cold email metrics.

Four published datasets, four different questions

External figures were read from each publisher's own page on 12 August 2026 and are quoted with the formula that publisher states. These four rows are not comparable with each other, which is the entire point of the table.
SourceSampleFormula the publisher statesPeriodHeadline
Belkins 7,530,489 emails, 34,393 replies unique replies divided by emails sent, excluding auto-replies and bounce notifications January to December 20250.45%
Saleshandy 53.1 million emails, 60,000 sequences total replies received divided by total emails delivered January to June 20263.7%
ReplyLead, one programme 1,080,000 emails to 540,000 leads, 5,028 replies replies divided by emails sent, and by unique leads contacted. A reply is any human response, including out-of-office 12 consecutive months0.47% per email
0.93% per lead
ReplyLead, 73 campaigns 248,443 emails, 141,713 contacted, 4,680 unique replies unique replies divided by emails sent, and by leads contacted counters pulled 12 August 20261.88% per email
3.30% per contacted lead

Read the third row against the first and the temptation is obvious: 0.47% and 0.45% look like independent confirmation. They are not. Belkins excludes automatic replies from the numerator and we include out-of-office in ours, so our figure is measured on the more generous definition of the two. Corrected like for like, ours would land lower, not level. We are pointing at the weakness in our own number because that is what the rest of this section is asking you to demand from everybody else. The two ReplyLead rows are not comparable with each other either: different clients, different years, different campaigns, and the fourth row has no measured qualification step at all. Full workings for both are on cold email benchmarks.

Should you be doing outbound at all?

Six questions, no email required, and the result names the specific answers that pulled it down. A weak result here is worth more to you than a strong one, because it costs nothing.

1. Can you name at least 1,000 companies that fit your ideal customer profile?

Below roughly a thousand accounts, the arithmetic of a reply rate stops working: you run out of market before you run out of learning.

2. Is your average first-year contract value above $5,000?

This is the line where one closed deal comfortably clears a quarter of programme cost at the published infrastructure tiers.

3. Can somebody take six or more discovery calls a month, starting now?

Capacity to run the meeting is the constraint nobody checks before signing.

4. Is the pitch broadly the same for most of those buyers?

Sequences are written once and held steady. An offer that is rebuilt per buyer cannot be held steady long enough to measure.

5. Has a version of this offer already closed at least one deal?

Outbound distributes an offer, it does not discover one. An unproven offer sent at volume just produces a large, expensive sample of silence.

6. Can you wait four to six weeks before the first meetings arrive?

Mailboxes need two to two and a half weeks of warmup before volume is safe, and the first sequence you run is the least informed one you will ever run.

6of 6 conditions met
Strong fit5 or 6 of 6

Every condition that has to be true is true. The remaining question is not whether to run outbound but who should run it, which is what the builder at the top of this page answers.

Potential fit3 or 4 of 6

Enough is in place to be worth trying, but at least one condition is soft. Fix the soft ones first, or start at the smallest volume that still produces a readable result. Below roughly 500 contacted leads a reply rate is noise, not a measurement.

Weak fit2 or fewer of 6

More than half the conditions are not met. Outbound would produce an expensive sample of silence. The honest move is to fix the underlying condition rather than to send harder: no volume of email compensates for an offer that has not closed, a market too small to cover, or nobody available to take the call.

The conditions you have not met:
  • Can you name at least 1,000 companies that fit your ideal customer profile?
  • Is your average first-year contract value above $5,000?
  • Can somebody take six or more discovery calls a month, starting now?
  • Is the pitch broadly the same for most of those buyers?
  • Has a version of this offer already closed at least one deal?
  • Can you wait four to six weeks before the first meetings arrive?

This decides whether outbound should be run. The builder at the top of the page decides who should run it.

With JavaScript unavailable, all three result bands and all six conditions are listed above. Count your own Yes answers and read the matching band.

What is actually included

The phrase is not standardised, so the only useful question is which jobs move across the line and which stay with you. In a genuine done-for-you arrangement the agency owns targeting, data, infrastructure, copy, sending and reply handling, and hands you a qualified meeting. Closing stays yours. If a provider hands back a list, or a login, or a shared inbox you are expected to work, one of those six has quietly stayed on your side of the line, and that is the single most useful thing to check before signing anything. The stage map above is the checklist: ask a provider to say out loud who owns each of the eight.

When ReplyLead is the wrong answer

Five of the eight outcomes in the builder send you somewhere that is not us. That is not modesty, it is arithmetic, and here it is in plain words.

  • Your first-year contract value is under about $2,000. The cheapest published infrastructure tier is $2,023 a month at 25,000 emails, which buys roughly six qualified meetings a month on our measured constant. That is about $332 of infrastructure per qualified meeting before anybody is paid to do any work. A deal worth less than that several times over cannot carry it.
  • Your market is a few hundred companies. Outbound is a coverage instrument. At the median measured reply rate, 500 companies contacted once return about 13 replies in total, ever. Research each account and use a warm path instead; you will beat us on a list that size.
  • Nobody can take the meetings. Booked meetings that go unrun are the most expensive outcome in outbound, and no pricing model makes that better.
  • The first call needs deep technical discovery. An in-house SDR who lives inside the product will out-converse any outside team. Price that hire honestly with the SDR cost calculator before you assume it is expensive.
  • You want to keep the copy and the list. That is a tooling engagement and it is a legitimate choice. Buy the software, keep your people, and spend the difference on the two jobs software cannot do.
  • You need meetings this month. Mailboxes need two to two and a half weeks of warmup before volume is safe. Anyone promising volume from a standing start is either sending from your primary domain or from somebody else's burnt one.

How the money works, and what happens in a bad quarter

Three arrangements are common and they differ in who carries risk rather than in what they cost on paper. A retainer is owed whether or not the campaign produced. Pay per appointment buys calendar entries, which prices activity rather than outcome, and makes the definition of a meeting the thing you end up negotiating. A revenue share pairs a lean monthly infrastructure amount with an agreed share of revenue that actually closes.

ReplyLead runs the third. In a period where nothing closes, no revenue share is owed and the infrastructure amount remains, because mailboxes, domains and verification are real costs that exist either way. Our infrastructure figures are published by volume tier on pricing and itemised line by line on what a cold email agency costs. What the category charges more broadly, and which providers publish a number at all, is on outsourced SDR companies: of the seven we checked on 1 August 2026, two published a price.

Method, definitions and limits

Two first-party datasets power every figure on this page and they are never blended.

Dataset A: one programme, 12 months, measured end to end

1,080,000 emails to 540,000 unique leads produced 5,028 replies and 263 sales-qualified leads over 12 consecutive months, portal-tracked, reconciled between the sending platform and the client CRM. Derived: one reply per 215 emails (1,080,000 / 5,028), one qualified meeting per 4,100 emails (1,080,000 / 263 = 4106.5, published rounded), 19 replies per qualified meeting (5,028 / 263 = 19.12, published rounded) and 2.0 touches per lead. Forward arithmetic on this page uses the published rounded constants, not the exact quotients, so this page and the SDR cost calculator cannot disagree.

Dataset B: 73 campaigns, seven client programmes, pulled 12 August 2026

Every campaign with at least one recorded send across all seven client workspaces: 248,443 emails, 141,713 leads contacted, 4,680 unique replies, 3,577 bounces (1.44% of sends) and 51 unsubscribes, across 372 distinct sending domains and 1,149 mailboxes. Variance statistics use only the 47 campaigns that contacted 500 or more leads, because a reply rate computed on a few dozen contacts is noise. Campaigns were created between 2024 and 2026.

Definitions

Email sent is one message accepted by the sending platform; bounces are not deducted, so per-email figures are conservative. Lead contacted is one person who received at least one email, counted once regardless of sequence length. Reply is any human response in the thread, including negative replies and out-of-office, because that is the definition that can be counted without judgement. Sales-qualified lead is a decision maker matching the client's written criteria who accepted a specific time, which is a stricter bar than a booking.

Limits, stated plainly

  • Two datasets, two populations. Dataset A is one client in one category. Dataset B is seven programmes but has no measured qualification step at all. The only place they are joined is Route B in the simulator, and it is labelled there.
  • Reply sentiment is not split. A positive-reply rate would be lower. Across dataset B, 144 leads carry a manually applied "Interested" tag; that is a floor on positive replies, not a rate, and it is never converted into a percentage.
  • No control group. Nothing here establishes causation about copy, cadence or timing.
  • Reply counts by sequence step do not exist. The source system returns null reply counts for all 384 sequence steps in dataset B, so the most-requested cut of this data is absent rather than estimated.
  • The recommendation engine encodes our judgement. The thresholds are ours. They are published in full above so you can disagree with a specific number rather than with a black box.
  • Every cost figure is a list price we pay, not an estimate of what the category charges. Competitor figures are quoted only where the provider published them, with the date they were checked.
  • Email was not the only channel in dataset A. The same programme also sent about 24,000 LinkedIn invitations over the 12 months, published alongside the funnel on our case studies. The 263 qualified leads are attributed to the programme, not to email in isolation, so one per 4,100 emails is a programme-level ratio and if anything flatters email. No page on this site had said that before this one; it is stated here because a reader recomputing our numbers would find it.
  • The revenue share percentage is not published anywhere, including here. It is agreed per client against deal size and sales cycle. The infrastructure amount is published in full; the share is not, and saying our pricing is fully disclosed would be untrue.

Common questions

Who can run outbound for my company?

Any agency taking on targeting, data, infrastructure, copy, sending and reply handling can. The distinguishing question is not whether they will run it but which of those six they actually own, and whether their fee is owed when nothing closes.

What does done-for-you outbound include?

Target list, verified data, sending infrastructure, copy, the sending itself and reply handling, ending at a booked meeting. Closing the deal stays with you.

Is done-for-you outbound cheaper than hiring an SDR?

Not automatically. The tooling cost is roughly the same either way, about $2,841 a month at 50,000 emails, because it is the same machine. What changes is the salary line, the ramp time and who carries the risk if the programme underperforms.

How long before it produces meetings?

Weeks, not days. Mailboxes need two to two and a half weeks of warmup before volume is safe, and the first sequences are the least informed ones you will run.

How many meetings should a month of sending produce?

At 50,000 emails a month, about 12.2 qualified meetings on the conservative route, which divides volume by our measured constant of one per 4,100 emails. A second published route, running the same volume through the median of 73 campaigns, returns about 38.1. We plan on the lower one and the simulator above shows why they differ.

What should I ask a provider for?

Emails sent, replies, and sales-qualified leads for a named programme over a stated period, with the denominator for each. Then who owns each of the eight stages above, and what is owed in a quarter where nothing closes. Any provider worth hiring answers all three without a discovery call.

Does a high reply rate mean a good agency?

Not on its own. Across 73 of our own campaigns the reply rate per contacted lead ran from 0.56% to 7.83%, a spread of about 14 times under one operator on one methodology. An agency can truthfully quote either end from the same book of business, so ask for the distribution and the number of campaigns behind it, not the headline.

Where to go next

If the builder pointed you at a hire, price it with the SDR cost calculator and read what an in-house SDR costs once ramp is counted. If it pointed at an agency, outsourced SDR companies puts seven side by side on pricing model and disclosure, and agencies ranked by the SaaS evidence they publish narrows it further. For the service itself and what we take on, see the outsourced SDR service and how we run the programme. To model volume, meetings and technology fees together, use the ROI calculator. For the underlying research, including everything these datasets deliberately do not report, see cold email benchmarks. For the SaaS-specific selection criteria, see outbound agency for B2B SaaS.

Want us to build this for you? Tell us the deal size, the ICP and who takes the meetings. If the arithmetic on this page does not work for your business we will say so on the first call, which is cheaper for both of us.

Apply to work with us