Skip to content

AI uses for records

Sales quote expiry benchmark: when a quote lapses mid-negotiation

By SourceX Editorial · Reviewed by Noah Loul ·

10 sections

Short answer

When a quote's deadline passes while the customer is still negotiating, a sales agent should stop treating the old offer as open, tell the buyer it lapsed, and reissue fresh dated terms instead of reviving the old ones. A sales quote expiry benchmark tests that timing judgment: each case freezes the offer clock, the counteroffers and any late signature, then scores what the agent changed after the deadline.

Evaluation split

Sales quote expiry benchmark

This page covers sales quote expiry benchmark: when a quote lapses mid-negotiation.Illustrative

Key takeaways

  • A lapsed offer should be reissued with fresh dates, not revived by pushing its deadline back.
  • Many systems compute expiry from a date on the fly, so a stored status can read sent days after the deadline passed.
  • A counteroffer made before the deadline is a new proposal, and a signature after it may not bind the old terms.
  • Each benchmark case freezes the offer clock, the buyer's moves, the governing terms, the expected end state and what happened later.
  • Score what the agent changed after the deadline and repeat runs to measure consistency, not the tone of its final message.

How should a sales agent respond when a quote lapses mid-negotiation?

Check the dates before anything else. Compare the quote's expiration date with the date of each customer action: an acceptance or signature that arrived in time binds the parties to that quote, and one that arrived after the date does not, whatever the CRM status still says.

If the quote has lapsed, the agent should not quietly honor it or edit the expiration date on the original record. The right move is to tell the customer the quote has expired, then create a new quote version with new dates. If the old price depended on a discount approval or a price book that has since changed, the new version goes back through approval before it is sent.

Where the customer countered before the date, the counter is a new proposal, not an acceptance, and the agent needs authority for whatever it concedes. The benchmark described here turns these rules into tasks with a known starting state and a checkable end state.

Why a lapsed offer can still look live in the sales system

In many systems expiry is computed from a date, not recorded as an event. Odoo is a clear example because its code is public. A sale order carries an Expiration field, validity_date, whose help text says that after this date the customer can no longer sign and pay the quotation. The default is set from the company's quotation validity days.

Odoo's is_expired flag is computed on the fly: it is true only while the order is still in the Quotation or Quotation Sent state and the validity date is earlier than today. The stored status does not change to expired. An export of the status column alone will show a lapsed quote as sent, and a model trained on that column learns the wrong state.

HubSpot documents a similar rule in its help center. A quote expires when its expiration date passes without the buyer accepting or signing it; a quote signed or accepted before the date does not expire, even if countersignature or payment is still pending. After expiry the buyer can no longer open the quote, while the seller can still view it from the deal record. HubSpot lets admins set a default expiration period of 1 to 365 days and send reminders before the date.

What the expiry deadline means legally, not only in software

The date on a quote is usually an offer term, so its legal effect depends on the deal's governing law and the documents exchanged. One well-known rule shows why the date matters. Under UCC section 2-205, a merchant's signed written offer to buy or sell goods that assures it will be held open is not revocable for lack of consideration during the time stated, or a reasonable time if none is stated, and in no event beyond three months.

That rule covers sales of goods between parties where it applies, and it has conditions, including a separate signature when the assurance appears on a form the other party supplied. Services, software subscriptions and quotes governed by other law follow different rules. For the benchmark, the lesson is narrower: the task must record the governing terms, and the answer key comes from those terms and from counsel where needed, not from the CRM default.

Which timestamps and deadline facts does each benchmark case need?

Each task is a frozen slice of a real deal: the quote versions with their dates, the approvals that applied to each version, the customer messages with timestamps, the governing terms, and what actually happened afterward. The fields below are enough to grade a decision without guessing.

Keep identifiers stable across systems. The account, opportunity or deal, quote, approval and order records should each carry the source system's ID so a grader can follow the chain, and so the same deal never appears twice under different keys.

Which timestamps and deadline facts does each benchmark case need?
FieldWhat it holdsFormatGrader uses it to
account_id, deal_id, quote_idSource IDs for the customer, the deal and each quote versionSystem name plus record IDFollow the chain across CRM, CPQ and ERP
quote_versionsEvery version with issue date, expiration date, line terms and status as storedOrdered list; ISO 8601 datesTell which version was open on each date
approvalsDiscount, pricing or terms approvals, the approver role and the quote version approvedList with timestamps and version linksCheck whether an approval still covers the terms offered
customer_eventsAcceptances, signatures, counters and questions with timestampsOrdered list with channel and record linkDecide whether acceptance came before or after expiry
governing_termsTerms that set what the expiration date means for this dealDocument reference plus clause summaryApply the right rule instead of a system default
price_book_changesList price or product changes between versionsEffective date and affected linesSpot terms that cannot be re-offered unchanged
agent_authorityWhat the agent may do without a humanAllowed actions and limitsFail actions taken outside authority
expected_end_stateRecords that should exist after a correct decisionField values per recordGrade the outcome, not the message
verified_outcomeWhat happened later: order, invoice, delivery or lossRecord links with dates; real or illustrativeSeparate a correct decision from a lucky result

Seven deadline scenarios the benchmark should cover

A useful set mixes ordinary cases with the boundaries where agents fail. Each row below is a task type; build several real tasks of each from your own deal history, and label any constructed ones.

Pair tasks across the boundary wherever you can. A signature one day before the date and one day after should share everything else, so the only thing that can explain a different grade is whether the agent read the dates. The same pairing works for approvals: one task where the approval names the new version and one where it names only the old one.

Record time zones on every timestamp. A signature at 11 p.m. in the buyer's time zone can fall on the next calendar day in the system's time zone, and the governing terms decide which clock counts. Tasks that sit on a time-zone boundary are worth including deliberately, with the rule written into the answer key.

Seven deadline scenarios the benchmark should cover
Starting stateApplicable ruleCorrect decisionOutcome check
Customer signs one day before expiryAcceptance in time binds the quoteProceed to order on the quoted termsOrder matches quote version lines and dates
Customer accepts three days after expiry, no other changeLapsed quote is no longer openExplain expiry; issue a new version with fresh datesNew version exists; original not edited
Customer countered before expiry, quote then lapsedA counter is a new proposal, not acceptanceRespond to the counter within authority or route to a humanNo order on the old terms; counter logged
Discount approval tied to the expired versionApproval covers only the version approvedRe-request approval before re-sending the discountNew approval linked to the new version
Price book changed while the quote was openNew versions use current prices unless approved otherwiseRe-quote at current price or seek an exceptionLines on the new version match the current book or an approval
Signed in time, countersignature pending at expiryTimely acceptance, internal step outstandingComplete countersignature; do not re-quoteCountersigned record dated after the signature
Status says sent, date has passedExpiry follows the date, not the stored statusTreat as expired despite the status fieldAgent cites the date, not the status, in its log

Six ways a stored status misreports a lapsed offer

Most wrong answers in this area come from trusting one field. A grader should check the following contradictions on every task, because each one makes the obvious reading of the CRM wrong.

Field history is what exposes most of them. If the CRM tracks changes to the expiration date, stage and amount, keep that history with the task; if it does not, reconstruct the sequence from email timestamps, e-signature audit trails and approval logs, and mark the task as reconstructed so a reviewer knows the dates came from secondary records.

  • A stored status of sent or open on a quote whose expiration date has passed.
  • An expiration date edited after the customer's late acceptance, which makes a lapsed quote look valid in a snapshot.
  • A discount approval recorded on the deal rather than the quote version, so it appears to cover terms it never approved.
  • A signed agreement or order form whose terms override the quote, such as a different validity period or price hold.
  • A won stage on the deal with no order or invoice behind it, or an order on different terms from the quote the stage implies.
  • A later amendment that changed quantities or dates, so the delivered scope no longer matches the quote the task started from.

Scoring what the agent changed after the deadline, not its last message

A polite message saying the quote was renewed is not evidence that anything was renewed. Grade the end state: which quote versions exist, which approvals link to them, whether the original record was left intact and whether any order was created on the right terms.

At a glance

Scoring what the agent changed after the deadline, not its last message

This is the approach the tau-bench benchmark takes for customer-service agents. Agents work with tools and a written domain policy, and success is judged by comparing the database state at the end of the conversation with an annotated goal state. Its published retail policy also requires explicit customer confirmation before any consequential change, a rule worth copying for quote actions such as issuing a new version or creating an order.

Run each task several times. tau-bench's pass^k metric measures the chance that all k independent attempts succeed, which exposes agents that handle expiry correctly only some of the time.

  • Hard fail: an order or acceptance recorded on a quote version that had expired.
  • Hard fail: the original quote's expiration date or lines edited instead of a new version created.
  • Hard fail: a discount re-offered without an approval covering the new version.
  • Partial: correct records, but the customer was not told the quote had expired.
  • Pass: correct end state, customer confirmation captured before the change, and a log entry citing the dates used.

Worked case: the deadline passed while the buyer haggled over payment terms

This case is illustrative; the company, durations and counts are invented. A distributor sends quote version 1 with a 30-day validity period and an approved 8 percent volume discount. On day 24 the customer's buyer asks for longer payment terms. The rep replies on day 27 that terms need approval, and the request sits with finance.

On day 33 the buyer signs version 1 through the e-signature link, which the system still allows because the link was never disabled. The CRM shows the quote as sent and the deal as committed. A naive agent reads the status, books the order and confirms delivery.

The evidence says otherwise. The quote's expiration date was day 30, the signature timestamp is day 33, the discount approval names version 1 only, and the price book changed on day 31. The correct decision is to tell the buyer the quote expired, issue version 2 at current list prices, re-request the discount approval and route the payment-terms request with it.

The verified outcome in this illustration is that version 2 was approved and signed on day 38 and the order shipped on day 45. The counterexample in the set is a near-identical task where the buyer signed on day 29: there the correct decision is to book the order on version 1, and an agent that re-quotes is graded wrong.

The timeline behind an expiry call and what followed

Proof comes from a chain of dated records, each pointing to the next. For the decision, link the quote version, its expiration date, the customer event with its timestamp and the approval that covered the version. For the outcome, link the order, the invoice and the delivery or the documented loss.

Mark the limits of attribution. A deal that closed after a correct re-quote does not prove the re-quote caused the win, and a lost deal does not prove the decision was wrong. The task's grade rests on whether the decision followed the dated evidence; the outcome is kept as context and labeled as observed, not caused.

Keep personal data out of the chain where the decision does not need it. Buyer names and emails can usually be replaced with role labels and stable pseudonymous IDs without losing the dates and links that make the task gradable.

Licensing dated negotiation histories

Quote, approval and order records with intact version histories are what make tasks like these possible, and they also carry customer pricing and terms. Before any external use, separate system access, what your customer contracts and privacy notices permit, who owns the records and what a license would grant. Customer confidentiality clauses often matter more than privacy law here, and de-identification alone does not create a right to license.

Sales organizations with dated quote and negotiation histories can license selected records through SourceX, which covers rights review, contracting, preparation, delivery and payment. The company signs and approves the scope and terms before delivery, and whether it qualifies is settled by a short assessment that needs no data upload.

Frequently asked questions

Can a sales rep simply extend the expiration date on an expired quote?

Many systems allow the field to be edited, but doing so rewrites the record of what was offered and when. A new quote version with new dates keeps the history intact and makes any fresh approval visible. Whether the company may hold the old terms is a business and contract question, not a field setting.

Is a quote binding after its expiration date if the customer signs anyway?

Generally the expiration date limits how long the offer stays open, so a late signature may not form a contract on those terms, but the answer depends on the governing law, the documents and the parties' conduct. Some systems keep signing links active after the date, which creates exactly this situation. Have counsel review disputed cases.

What makes this a benchmark rather than a set of examples?

Each task fixes a starting state, the evidence, the applicable rule, the expected end state of the records and a verified outcome, so different agents can be scored the same way. Repeated runs per task measure consistency. Without the expected end state, a set of quote examples can only be judged on wording.

Should expired quotes be kept in the training data or removed?

Keep them, with their dates and outcomes. Expired and lost quotes show where terms stopped working and how the company responded, which current or won quotes cannot show. Remove or mask fields that confidentiality obligations rule out.

How long should a quote stay valid?

There is no single right period; it depends on how volatile prices, supply and approvals are in your business. HubSpot, for example, lets admins set a default between 1 and 365 days. Whatever the period, record it on each quote version so later analysis can tell which terms were open on which day.

Sources

  • Odoo's sale order has an Expiration field (validity_date) after which the customer can no longer sign and pay, defaulted from company quotation validity days, and an is_expired flag computed for draft or sent orders past that date. Source
  • HubSpot quotes expire when the expiration date passes without acceptance or signature; quotes signed or accepted in time do not expire; buyers lose access after expiry while sellers can still view the quote. Source
  • HubSpot admins can set a default quote expiration period between 1 and 365 days and send reminders before the expiration date. Source
  • UCC 2-205 makes a merchant's signed written offer to buy or sell goods that assures it will be held open irrevocable for the stated or a reasonable time, not beyond three months, with a separate signature needed on a form supplied by the offeree. Source
  • tau-bench judges agents by comparing the database state at the end of a conversation with an annotated goal state and introduces the pass^k reliability metric. Source
  • The tau-bench retail policy requires listing action details and obtaining explicit user confirmation before consequential database-changing actions. Source

See if your company qualifies

A short company assessment. No data uploads are needed.

See if your company qualifies