SKILL.md · 8.5 KB
View raw
---name: inbox-capacity-plannerdescription: Works out how many new leads a day a set of sending mailboxes can actually sustain, given how long the sequence is, and flags when a planned import will starve its own follow-ups. Use when sizing a mailbox fleet, planning cold email volume, or diagnosing a campaign that looks healthy on the dashboard but has stopped reaching anyone new.license: MIT--- # Inbox capacity planner Most people size a cold email programme by daily send volume. That is the wrongnumber, and using it is the most common reason a campaign quietly stops workingin its second week. Sending capacity is not the rate at which you can add people. Every lead youadd consumes one send for each step of the sequence, spread over the followingdays. A five step sequence means each lead costs five sends, not one. The rateyou can sustain is therefore capacity divided by sequence length, and it is amuch smaller number than the one on the invoice for the mailboxes. This skill computes that number, shows the two separate ceilings that produceit, and says what will go wrong if the import rate exceeds it. ## What to ask for Ask for all six. If the person does not know one, say which, and give theanswer as a range rather than quietly guessing a single figure. 1. **Mailboxes**: how many sending addresses are in the fleet.2. **Sends per mailbox per day**: the configured daily limit per address.3. **Sequence length**: how many emails one lead receives before the sequence   ends, counting the first one.4. **Days between steps**: the gap the sequence waits between emails.5. **Sending window**: how many minutes a day the fleet is allowed to send in,   usually working hours in the recipient's timezone.6. **Minimum gap between sends from one mailbox**: the spacing setting, in   minutes. Also ask how large the list is, if there is one. It converts the answer intodays of runway, which is the form most people actually want. ## The arithmetic Two ceilings apply, and the lower one wins. The **configured ceiling** is what the sending tool has been set to allow. The**physical ceiling** is how many sends a mailbox can place inside the window,given the spacing: ```slots per mailbox     = floor(sending window minutes / minimum gap minutes)effective per mailbox = min(sends per mailbox per day, slots per mailbox)effective daily sends = mailboxes x effective per mailbox sustainable new leads = floor(effective daily sends / sequence length)days to steady state  = (sequence length - 1) x days between stepsdays of runway        = list size / sustainable new leads``` **Always report both ceilings, not just the answer.** When the physical ceilingis the lower of the two, the configured number is fiction: the fleet cannotplace the sends it has been told to place, and the shortfall never appears in asettings screen. Say so out loud every time it happens. ## The failure mode this is really about If new leads are added faster than the sustainable rate, the programme does notfail loudly. It fails in one of two quiet ways, depending on how the sendingqueue is ordered. **If follow-ups go first**, which is the sane default, the daily allowance isspent on the sequence steps already owed to leads in flight. New leads wait.Import several times the sustainable rate and the new-lead queue stops drainingaltogether: the dashboard reports a full day of sending, every mailbox hit itslimit, and almost nobody new was contacted. Everything looks healthy. This isthe shape of a campaign that appears to send nothing for a day, and the causeis arithmetic upstream rather than a fault in the tool. **If new leads go first**, the newest arrivals consume the allowance and thefollow-ups slip. The three day gap becomes five, then nine. The sequencestretches until the second email no longer reads as a follow-up to the first,and the reply rate falls for a reason that has nothing to do with the copyeveryone is busy rewriting. Say which of the two is happening if the person is diagnosing rather thanplanning. Ask how their queue is ordered. If they do not know, the symptomtells you: full send volume with no new contacts is the first, and sequencegaps drifting longer than configured is the second. ## The levers, in order of cost 1. **Shorten the sequence.** The sustainable rate is inversely proportional to   sequence length, so going from five steps to four raises it by a quarter,   immediately and for nothing.2. **Widen the window or narrow the gap**, but only when the physical ceiling   is the binding one. If the configured ceiling is lower, this changes nothing.3. **Add mailboxes.** Linear, and the only lever that costs money every month.4. **Import more slowly.** Not a lever on capacity, but usually the correct   answer, and the one nobody suggests. ## Ramping A domain and mailbox that has not sent to strangers before should not start atthe target volume. During a ramp the sustainable new-lead rate is proportionallylower, because the capacity it is divided from is lower. The trap is importing at the steady state rate during week one. That guaranteesa backlog before a single follow-up has been sent, and the backlog is invisibleuntil the sequence steps come due. If a ramp is in progress, compute the sustainable rate at the current week'svolume rather than the target, and say which one you used. ## What to output A short table, then the sentences that matter: | Figure | What it tells them ||---|---|| Effective daily sends | the lower ceiling, and which of the two it was || Sustainable new leads per day | the answer || Days to steady state | when the full load actually arrives || Days of runway | if a list size was given | Then, in plain sentences: the sustainable rate with the daily send volume nextto it, so the gap between the two is visible; which ceiling is binding and whatwould move it; and, if a planned import exceeds the sustainable rate, whichfailure mode to expect and roughly when. Do not present the sustainable rate on its own. The number is unremarkable inisolation and persuasive next to the number it is usually confused with. ## Worked example Illustrative figures. These are not measured from any account. Eight mailboxes, 30 sends per mailbox per day, a five step sequence with threedays between steps, a nine hour window, and a fifteen minute minimum gap. ```slots per mailbox     = floor(540 / 15) = 36effective per mailbox = min(30, 36)     = 30      configured ceiling bindseffective daily sends = 8 x 30          = 240sustainable new leads = floor(240 / 5)  = 48days to steady state  = (5 - 1) x 3     = 12runway on 2,000 leads = 2000 / 48       ~ 42 working days``` The fleet sends 240 emails a day and sustains 48 new people a day. Anyoneplanning against the 240 is planning at five times the real rate. Now change one setting. With a twenty minute gap instead of fifteen: ```slots per mailbox     = floor(540 / 20) = 27effective per mailbox = min(30, 27)     = 27      physical ceiling now bindseffective daily sends = 8 x 27          = 216sustainable new leads = floor(216 / 5)  = 43``` The tool still says 30 a day. The mailbox can place 27. Nothing reports themissing three, and the sustainable rate drops by roughly a tenth because of aspacing setting nobody associates with volume. ## What this deliberately does not do - It does not know your provider's real limits. Those vary and they change.  Ask rather than assume, and say in the output that the answer is only as good  as the limit you were given.- It does not model deliverability, reply rates or bounce rates. Capacity is  what you are allowed to send. Whether it lands, and whether anyone answers,  are different questions needing different evidence.- It does not tell you which people on the list may lawfully be emailed.  In the UK that depends on each recipient's legal form, and getting it wrong  is a legal problem rather than an operational one. Use a compliance check for  that, before the list is loaded rather than after. ## Not for you if - You send fewer than a few hundred emails a month. The arithmetic still holds,  but capacity will not be the thing limiting you.- You want a deliverability audit. This is a volume model and it will tell you  nothing about whether your mail reaches an inbox.- You want one recommended number of mailboxes to buy. The honest answer  depends on your sequence length and your list size, which is why both are  inputs rather than assumptions. --- Written by TheCold.Ai, a B2B outbound agency in Middlesbrough, England.Not affiliated with Anthropic.

Want it run for you instead?

These files are the judgement. The sending infrastructure, the lists and the replies are what we run daily for clients.

Book a 15-minute call

Questions about this skill

Straight answers. If yours isn't here, ask it on the call: we'd rather tell you no than sell you the wrong thing.

Does this work with any sending tool?
Yes, because it models arithmetic rather than a product. It asks you for your provider's limits instead of assuming them, and it says in its own output that the answer is only as good as the limit you gave it. No tool is named anywhere in the file.
Do I need a Claude subscription to use it?
You need somewhere to run a Claude skill. If you would rather not, the arithmetic is written out in full on this page and in the file: capacity divided by sequence length, with the window and the gap setting the physical ceiling. You can do it on paper.
Are the numbers in the worked example yours?
No, and the file says so in the line above them. They are illustrative figures chosen to show the two ceilings changing places. We do publish our own measured numbers, but not inside a skill: a skill gets copied onto other people's machines, where a figure cannot be corrected later.
What is it not going to tell me?
Whether your mail lands, whether anyone replies, and which people on the list may lawfully be emailed. The last of those is a real question in the UK and it has its own checker.

The rest of measuring, capacity and compliance: uk-pecr-cold-email-checker and outbound-analyst. Everything sits on the library page, next to The compliance checker, which is the one people usually want after this. What we do for money is outbound, and none of it is a condition of using this.

Claude is a product of Anthropic. TheCold.Ai is not affiliated with, endorsed by or connected to Anthropic.

Want the version with your numbers in it?

The skill works from whatever you type into it. If you would rather someone looked at the actual fleet, book fifteen minutes and we will go through it.

Book a 15-minute call15 minutes, no obligation, no decksales@thecold.ai

You approve every call · no charge for no-shows · based in England