How long should you give an outbound agency before deciding it doesn't work?

For cold email, give it about 10 to 12 weeks from kickoff, and treat the first four as setup, not a test. For LinkedIn the clock is shorter. That is our suggested window, not an industry norm (we couldn't find a verified one), and it only holds if the agency can show you something specific at every checkpoint below.

"Be patient" is a fair answer in week 3. It is a bad answer in week 10 if nobody can show you what happened in between.

Weeks 1 to 4: nobody can judge anything yet

A new email program starts with new domains and fresh mailboxes, and they have to earn a sending reputation before they carry real volume. MDR's standing practice is brand-new domains and about four weeks of warm-up.

That matches what the sending platforms publish. Lemlist says roughly 3 to 5 weeks, Smartlead says at least 4 weeks before serious volume, and Instantly's 2026 benchmark report describes 4 to 6 weeks of gradual warm-up. Google names no number of weeks, only the principle: start with low volume and ramp up slowly. The three tool vendors all sell sending software, so read the range as guidance.

So in weeks 1 to 4 there are no replies to judge. What you can ask for instead:

1. Domains and mailboxes set up, authentication records in place.
2. Warm-up running, with health you can look at yourself.
3. The list built, and a sample of it you can inspect.
4. Copy drafted, with the angle explained.

If it's week 4 and none of that exists, you have your signal. It doesn't need a calendar.

Weeks 5 to 6: first sends, and the first thing to check is delivery

Sending starts small. That is too little to expect meetings and too small a sample for a reply rate to mean anything.

Ask three things. Are the emails reaching the inbox? Are bounces low? Did any reply come in at all? Smartlead's rule of thumb is that zero replies on a properly warmed mailbox points to a deliverability problem, not a copy problem. That is the whole argument of deliverability before copy. Zero replies here means check the infrastructure, not fire anyone and not rewrite the message yet.

Weeks 7 to 8: full pace, and the first readable numbers

In the process shape we plan for, full sending pace arrives around 6 to 7 weeks after day one. Now the reply rate starts to say something.

Instantly's report puts the average cold email reply rate at 3.43%, and Smartlead gives 2 to 8% for a warmed mailbox. Both are vendor data and wide ranges, so use them as a sanity band, not a target.

Ask for positive replies separately from all replies. We found no reliable benchmark for positive reply rate, so don't accept a percentage on faith. Ask to read the actual replies.

Also ask what people are saying no to. Wrong title, wrong timing, already have a vendor: an objection is information, and a good agency turns it into a change to the list or the angle.

One caution. Instantly reports that 42% of replies come from follow-ups, not the first email. A sequence that hasn't finished running hasn't been judged.

Weeks 9 to 12: the first fair verdict

By now full sequences have run to enough people that the numbers mean something. Ask for the reply rate, the positive replies, what the objections changed, and the meetings, held and qualified, not just booked. What counts as a result is its own argument, and we made it in what's behind "94 meetings booked".

Then read the pattern:

1. Replies look healthy but no meetings: the problem is the offer, the targeting or the handoff. That is fixable and it is not the same as failing.
2. Replies near zero with placement confirmed fine: the list or the message is off. Ask what changed since week 6. If changes were made and nothing moved, this is when you can walk away without guilt.
3. Nothing shown to you at any checkpoint: that is your answer, whatever the calendar says.

LinkedIn runs on a shorter clock

There is no domain to warm, so the clock starts earlier. The first weeks still go to the list, the angle and the profile.

Acceptance rate is the first number you can read. Published averages run from 26% to 51%, and the largest datasets cluster around 26 to 30% (a roundup of vendor studies, so again vendor data). Read acceptance in the first two weeks of sending, then replies and positive replies by about week 4 to 6 of sending. If acceptance is far below that band and nobody can say why, ask.

Nobody honest can promise you a date

No agency can guarantee a meeting by week 8, and we don't. What an agency can commit to is what you will see, and when. Ask for that checkpoint list before you sign, and hold them to it.