A lead is lost in two ways, and the second is quieter than the first. The obvious loss is the one where nobody wrote back and nobody followed up: the thread ends with a proposal and a silence, and six weeks later the buyer has bought from somebody who did reply. The quiet loss runs the other way. The seller follows up a dozen times, each message carrying nothing, and the lead is lost because the buyer learned what the thread was worth — a channel that produces reminders, and never information.
Both failures come from the same place. A follow-up sequence is usually written as a schedule: day 3, day 7, day 14, day 30. The schedule is the easy half and it is not the half that matters. What a sequence needs is a reason for each touch that exists independently of the schedule — a question the buyer asked and the seller has now answered, a comparison the buyer is going to run, a validity date in the offer that is about to pass, a premise of the buyer’s brief that has changed. Where the reason is real, the touch is correspondence. Where it is missing, no interval is short enough to make the message interesting.
This piece sets out that sequence: six touches, each with its reason and the document that carries it, the stop rule that has to be written before the first one is sent, a worked example of thirty-three days and eleven calls that adds up, the limits of what a sequence can do, and the decision that ends it properly.
What is happening on the buyer’s side
A buyer who has received a proposal is not waiting to be reminded. They are doing one of a small number of things, none of which the seller can see: reading the proposal against a brief, circulating it to somebody who has to approve the spend, building the comparison with the other quotations they asked for, checking a delivery date against a contract that is ending, or doing nothing at all because the requirement has moved down a list.
That list matters because it determines what a useful touch is. The seller cannot know which of those is happening. What the seller can do is make each touch useful under more than one of them: a message that closes an open question helps a buyer who is reading, and also a buyer whose approver asked it. A message that restates the price helps neither, because both of those readers already have the price.
The distinction is not about warmth or persistence. It is about whether the touch changes what the buyer knows. A follow-up that carries a fact, an answer or a document leaves the buyer better able to decide; one that carries a nudge leaves them exactly where they were, with an extra message to file.
A touch without a reason is a reminder
The archetype is one sentence — just checking in — and it fails for a reason that is structural rather than stylistic. It asks the buyer to supply the reason for the exchange. The seller has no new information, so the buyer is invited to produce one: either a status, or a decision, or an apology for not replying. Each of those is work for the buyer and none of them advances the sale. The seller’s own record then shows a touch that happened, which is the only thing that improved.
There is a second cost, and it compounds. A thread in which two of five messages were reminders teaches the reader that the thread can be skimmed. The third message — the one that carried the answer to the approver’s question — arrives in the same envelope as the reminders, and it gets the same treatment. The sequence is judged by its weakest touch, not its average.
This is why the reason for each touch has to be written before the sequence starts, and why the reasons have to be things the seller can actually produce. “We will have the site survey by the 15th” is a reason, and it is a commitment to produce a document. “A gentle reminder” is not a reason; it is an admission that there was nothing to send and the message went anyway.
The sequence
Table 1 is the sequence: six touches over thirty-three days, the reason each exists, the document that carries it, and what the buyer can do as a result. The reason column is the test. If a touch cannot be given a reason from that column by the day it is due, it does not go — the schedule bends to the reasons, never the other way.
Table 1. A follow-up sequence of six touches, each with the reason it exists, the artefact that carries it, and what the buyer can do with it. Illustrative sequence for this piece.
| Touch | Day | The reason it exists | The artefact | What the buyer can do |
|---|---|---|---|---|
| 1 | 0 | The proposal is being submitted, against a brief with a stated requirement | The proposal, with the requirement restated and the decision requested | Read it against the brief; accept, question or decline |
| 2 | 4 | The proposal carried a marked gap, and the answer exists in the account’s own material | The gap closed on the page it appears on, with the source named | Check the answer against their own requirement |
| 3 | 10 | The buyer will build a comparison, and the fields of that comparison are known | A comparison table with the buyer’s named alternatives, field by field | Use the table, or correct its fields |
| 4 | 17 | A validity date in the offer is about to pass | A restatement of the date and what acceptance requires | Accept inside the validity, or ask for an extension |
| 5 | 25 | A premise of the buyer’s brief has changed, and the change is in the buyer’s favour or against it | A revised line — a date, a term, a quantity — traced to the document that changed | Re-plan against the new premise |
| 6 | 33 | The decision was requested for a date, and that date has arrived | The decision, restated as a question with an acceptance route and a name | Decide; a refusal is a decision |
The intervals shorten and lengthen around the dates rather than the calendar. Touch 4 is not on day 17 because seventeen is a good number; it is set from the validity date, which in the worked example below is four days from expiring. That is why the sequence’s dates are derived from the proposal’s own dates — the validity period, the requested decision date, the buyer’s stated deadline. A sequence with fixed intervals and no dates attached is a schedule. A sequence whose dates come from the documents is a record.
Two rows are missing from Table 1 on purpose. There is no touch on day 1 to confirm receipt, because a confirmation of delivery is not information for the buyer. And there is no touch after the sixth, because the sixth is the decision and the sequence ends there — either with an answer, or with a stop under the rule in the next section, and the second of those is an outcome rather than a failure.
The stop rule
A sequence without a stop rule is not a sequence; it is a habit. Five conditions end it, and all five are written before the first touch goes out, so that the decision to stop is not taken on the day the seller is most tempted to continue.
A decision is received. Accepted or declined, the sequence ends and the record says what was decided and on what date. A declined lead is a closed record, not a paused one.
The buyer names another supplier. Where the buyer says they have selected somebody else, that is a decision and the record closes. Where they say they are still deciding and name the alternatives, that is information, and touch 3 exists for it.
The buyer says not now, with a date. A pause with a date is a scheduled restart, and a restart has a reason of its own: the date. A pause without a date is a decline with better manners, and it closes the record.
Two consecutive touches have no reason available. This is the stop rule that does the work. If the seller reaches day 17 and the only thing that can be sent is a reminder, the sequence has run out of information and the correct action is to stop, record why, and leave the door open in one sentence. The touches that were scheduled after it do not happen.
The premise of the brief has gone. Where the requirement was a contract ending in January and that contract has been renewed, or the site has been sold, or the budget has been withdrawn, the lead is not stalled; it is over, and the record should say so.
Written in advance, the rule is a relief rather than a loss. It converts an open-ended obligation to keep writing into a bounded piece of work with an end, and it makes the ended record usable: a lead that stopped because the premise disappeared is a different fact from a lead that stopped because nobody answered, and only the first is worth revisiting next year.
Worked example: thirty-three days, eleven calls
The example is constructed for this piece. Northgate Facilities sends a brief on 3 September 2026 for a cleaning contract at two sites. The seller’s price list carries the units, the terms sheet carries a twelve-month term and thirty days’ payment, and the delivery lead time is ten weeks from acceptance. A proposal goes out on 5 September with a validity of twenty-one days, to 26 September, and a decision requested by 8 October.
Six touches follow the sequence, and Table 2 sets them out with the calls each consumed. The proposal at touch 1 takes six drafting calls at the rate the sales role specifies — six calls a proposal. Each of the five follow-ups takes one call, because each is a single document drafted against material the account already holds.
Table 2. The six touches, their dates, the calls each consumed and the cost at the measured Base rate. Constructed sequence; call counts follow the role’s specification.
| Touch | Date | What it carried | Calls | Cost |
|---|---|---|---|---|
| 1 | 5 September | The proposal, requirement restated, decision requested for 8 October | 6 | $0.04200 |
| 2 | 9 September | The installation question answered from the site schedule, cited to the document | 1 | $0.00700 |
| 3 | 15 September | Comparison table against the buyer’s named incumbent | 1 | $0.00700 |
| 4 | 22 September | Validity restated, four days before expiry, with the acceptance route | 1 | $0.00700 |
| 5 | 30 September | The buyer’s contract-end premise moved to 31 January, delivery date revised against it | 1 | $0.00700 |
| 6 | 8 October | The decision, dated, addressed to the named approver, with a refusal as an acceptable answer | 1 | $0.00700 |
| Total | 33 days | 11 | $0.07700 |
The arithmetic is short. Six proposal calls and five follow-up calls are eleven calls, and eleven calls at $0.00700 are $0.07700 for the whole sequence. At the metered ceiling of $0.00955 a call the same eleven calls are $0.10505, which is $0.02805, or 36 per cent, more, because a ceiling is what an agent authorises rather than what it spends.
Two touches make a commercial claim — the validity restatement and the decision — and those are the two worth a second reader. Re-reading them on Qwen flash at its measured $0.00875 adds $0.01750, taking the sequence to $0.09450, which is 22.7 per cent more than $0.07700 for a check on the two messages that commit the seller to something.
The scale arithmetic is linear and slightly dull, which is the point. Thirty leads through the same six-touch sequence are 330 calls, $2.31000 a month at the measured rate and $3.15150 at the ceiling. What that buys is not a number of replies; it is thirty sequences that can each be read back, each with its reasons, each ending in a decision or a documented stop.
Where a rate is genuinely needed, it has to be supplied by the reader rather than invented here. The cost of one order produced by this method is $0.07700 divided by the proportion of leads that reach an order. At one order in twenty the figure is $1.54000 — 0.077 per cent of a $2,000 order — and at one in ten it is $0.77000. The publication has measured neither proportion, and the arithmetic is given as a formula precisely so that a reader can put their own figure into it instead of accepting one from here.
What a sequence cannot do
The first limit is the hardest: a sequence cannot manufacture a reason. When the seller has nothing new, the honest touch is none, and the schedule contracts. This means the six-touch sequence is a ceiling rather than a plan — many leads will end at touch 3 or 4 because by then the reasons have run out, and an ended sequence with a reason recorded is a better outcome than a complete sequence of reminders.
The second is that a reason has to matter to the buyer. “We have extended the discount to the end of the month” is a reason for the seller. “Your brief asked for a 24-month warranty and our terms sheet gives twelve, so here is the price for the second year” is a reason for the buyer. The test is whether the sentence changes what the buyer would decide, not whether it changes what the seller wants.
The third is that silence is not a signal. A non-reply does not distinguish between a buyer who has not read it, a buyer whose approver is on leave, a buyer who has chosen somebody else and a buyer who is waiting for the seller to answer a question. A sequence should not try to decode silence; it should make one touch that can be answered in a sentence, and then act on the stop rule.
The fourth is that a sequence cannot be the relationship. The strongest touch in Table 1 is the one where the seller tells the buyer something inconvenient — the contract-end date has moved and the delivery date has moved with it. That touch is only credible from somebody the buyer has spoken to. Where the whole sequence runs by document and mailbox, it will be accurate and it will be cold, and that is a real cost rather than a technicality.
The fifth is the channel. The sales role drafts and sends under an allow-listed mailbox, and the WhatsApp connector is not built: where a buyer works over WhatsApp, the thread is prepared as a handover note for a person to carry across. A sequence that assumes a channel the seller cannot reach is a plan with a hole in it, and the hole should be named in the record rather than discovered at touch 4.
The sixth is that nothing here predicts an outcome. No reply rate, no response time, no close rate and no benchmark appears anywhere in this piece, because none has been measured by this publication and none can be borrowed honestly from another seller’s account. What the sequence does is make the work visible and bounded; whether the buyer answers is not a property of the sequence.
The failure at the other end
The opposite mistake is worth pricing, because it is cheap in money and expensive in the only currency that matters here — the buyer’s attention. Three extra touches after the decision date, each carrying a reminder, cost three calls, $0.02100. That is the whole of the model cost of the error, and it is negligible, which is why the error persists: it is invisible in every cost line the seller keeps.
What it costs instead is the thread. Nine messages in which six carried something and three carried nothing reads as a campaign rather than a correspondence, and the buyer’s next decision — the one in eighteen months, with a different buyer in the same organisation — starts from that impression. The arithmetic for that is not available and should not be invented. The observation is enough: the marginal cost of a reminder is a call, and the marginal cost of a thread that reads as a campaign is the account.
The remedy is structural rather than disciplinary. Where the stop rule is written before the first touch, a seller who reaches day 17 with nothing to say has a documented reason to stop, and stopping is the instruction rather than the omission. What gets recorded is a decision: the sequence ended because there was no reason left, on this date, with this sentence left open.
The decision
Write the reasons and the stop rule before the first touch goes out, and treat the reasons as commitments to produce documents: the answer to the open question, the comparison table, the revised date. A touch that cannot be given a reason from that list does not go, and the schedule bends.
Then date the sequence from the documents rather than from the calendar — the validity period, the decision date requested, the buyer’s own deadline — because those are the dates on which a touch can change something.
And end every sequence in writing, whichever way it ends. A record that says accepted on this date, or declined with this reason, or stopped because the premise disappeared, is a record that can be read in a year. A record that says nothing at all is a lead somebody will follow up again in six months without knowing that the last four messages went unanswered — and that is how a lost lead becomes a lost account.
Where this goes
- The pillar: x402.review/2026/10/31/commercial-operations/ — one record, two readers, and the approvals between them
- The role that runs the sequence: studio.x402.press/roles/sales — the proposal, the follow-up thread and the mailbox it sends from
- The gate most sequences are waiting behind: approval limits and segregation of duties — who decides inside the buyer’s organisation, and on what evidence
- The comparison the buyer will run: comparing quotes that are not like for like — the basis the buyer normalises to, and where an offer stops being comparable
- The service it buys: x402.press/services/deepseek/ — the canonical price, seven chains
- The review: x402.review/2026/10/31/deepseek-review/ — the measured handshake and what it does not promise
Method note
The price was read from the live 402 challenge on Base on 31 October 2026, a Saturday and therefore outside DeepSeek’s weekday peak windows: 7000 atomic units under exact, which is $0.00700, and 9550 under upto at the max_tokens tested. Qwen flash’s $0.00875 was read from its own challenge the same day. Call counts are planning assumptions taken from the sales role’s own specification: six calls for a proposal, one for a follow-up message.
Northgate Facilities, the dates, the buyer’s premise and the replies described in the worked example are constructed for this piece; no real account, thread or sale is described. The publication holds no measurement of reply rates, close rates, response times or the effect of any touch, and asserts none. The one rate used in the arithmetic is labelled as a placeholder with the division shown, so that a reader substitutes their own.
x402.review is written and edited by N.P. Vincent. The publication is connected to x402.press, which sells the services reviewed here; that interest is disclosed on every piece that touches a service, and it is stated here rather than left to inference.
