How to Track Commitments Made on Sales Calls

Last updated 20 September 2026.

What a commitment is, and why it is different from a note

A commitment is anything your company promises a customer in a conversation that someone else then has to deliver. "SSO ships in Q3." "We can turn off the retention cap for you." "Onboarding takes under a week." "We'll get you a CSV export on that screen."

It is not a note, an action item, or a next step. Those belong to the person who wrote them. A commitment belongs to a team that was not in the room — usually product, engineering or customer success — and most of them are made by a rep trying to close a deal.

Almost every company records the call the commitment was made on. Almost none of them can produce a list of outstanding commitments by customer. That gap is what this guide is about.

Why this breaks, in four steps

The failure is boringly consistent, and no individual step is anyone's fault.

1. The commitment is made verbally, under time pressure, near the end of a call, often as a conditional: "if that's a blocker, we can look at it for Q3."
2. The rep writes a clean CRM note. The CRM has a field for stage, amount and close date. It has no field for "things we said we would do."
3. Someone summarises the call into a ticket. The ticket says "improve bulk actions." The customer's name comes off at exactly this moment, because a tidy ticket does not carry who asked.
4. Six weeks later the work ships — or does not — and nobody can answer "who was waiting for this?" There is no list, because a list was never created.

The tell that this is happening to you: when a customer raises something in a QBR, your team debates whether it was ever promised. If that debate is possible, you do not have commitment tracking.

The method: five steps, no tooling required

You can do all of this with a spreadsheet and a discipline. Tools make it survive scale; they do not make it work.

1. Define what counts, in one sentence

Write it down and share it: a commitment is anything we said we would do, for a named customer, that someone outside this call has to deliver. Without a shared definition, half your team logs feature requests and the other half logs nothing.

Exclude two things explicitly, or your list will drown: things the customer asked for that you did not agree to (those are requests, track them separately), and things the rep will do themselves (those are follow-ups).

2. Capture at the moment, in the customer's words

The single highest-value habit: at the end of every call, the rep writes the commitment as a quote, not a summary. "Priya said the deal needs SSO before their security review in March" beats "customer wants SSO."

The quote survives handoffs. The summary does not, because the summary is already an interpretation and the next person will interpret it again.

3. Attach it to both a customer and an owner

Every commitment needs two names: which customer it is owed to, and which person or team owes it. A commitment with only one of those is a wish.

Owner is a person, not a department. "Product" does not read your list.

4. Give it a state, and review it weekly

Four states are enough: open, committed to a date, shipped, declined. Declined matters as much as shipped — a commitment you have decided not to honour is information the account owner needs before the customer discovers it.

Review takes ten minutes in an existing meeting. Read the open list by customer, not by team. Reading by team hides the account that is owed four things.

5. Close the loop out loud

When it ships, tell the person who asked, quoting them back. "In March you said your security review needed SSO. It's live — here's how to turn it on." That message lands completely differently from a changelog entry, because the customer recognises their own words.

This step is the entire return on the previous four. If you never tell anyone, you have built an archive.

A worked example

A rep closes a $40,000 account in March, committing to three things: SAML "in the next couple of quarters," a CSV export on the reporting screen, and onboarding in under a week.

Without this table, here is what happens instead. In April, product ships a roadmap containing none of the three, because product was not on the call. In August the customer raises SAML in a QBR and the CSM, who also never heard the call, treats it as a new request. In November the renewal is at risk over a promise made in March that the company kept no usable record of.

The recording existed the whole time. Nobody could search it for promises.

Where commitments actually get made

Before you can track them you have to know where they happen. In most B2B companies there are five, and teams usually watch only the first.

The last two are where the expensive ones happen, because they are made by senior people, under pressure, to keep an account — and they are the least likely to be written anywhere. If you only fix one moment, fix the QBR.

How to tell whether you have a problem

Four checks, none of which require new tooling. Do them this week.

1. Pick your three largest accounts and ask: what are we currently on the hook for? If assembling the answer takes more than ten minutes, or produces disagreement, you do not have tracking.
2. Read the last two churn or downgrade post-mortems. Count how many mention an expectation that was set earlier and not met. In our experience this is the single most common line in a churn write-up.
3. Ask a PM to name one thing sales promised this quarter. If they cannot, the pipeline from call to roadmap is not running, whatever your process document says.
4. Look at one closed-won deal and one recording from it. Compare what was said to what was written down. The delta is your capture rate.

Rolling it out without a mandate

Commitment tracking usually fails as a company initiative and succeeds as a habit somebody started. If you cannot mandate it:

• Start with one segment. Enterprise or top-20 accounts. The volume is manageable and the stakes justify the effort.
• Make logging cost under thirty seconds. One line, in a place the rep is already looking. Any form longer than that will be skipped under deadline, which is exactly when commitments get made.
• Show the list back weekly, by account. Visibility is the only enforcement that works. A list nobody reads stops being maintained within two weeks.
• Publicise the first close. The first time someone tells a customer "you asked for this in March and it's live," share the customer's reply internally. That reply does more for adoption than any process doc.

Expect the first month to be incomplete. A list covering sixty percent of commitments on your biggest accounts is dramatically better than nothing, and the gaps tell you where capture breaks.

What good looks like after a quarter

Three things should be true, and all three are observable rather than felt:

• Anyone can answer "what is this customer owed?" in under a minute, without asking the rep who sold it.
• Roadmap conversations cite customers by name. Not "customers want bulk actions" but "four accounts, here are the quotes, two are up for renewal in Q1." Prioritisation arguments get noticeably shorter.
• Customers hear back. The loop closes on things that ship, and — the harder discipline — on things you have decided not to build.

Common mistakes

• Logging requests as commitments. "They asked for SSO" and "we said we'd build SSO" are different obligations. Mixing them produces a list nobody trusts, and an untrusted list is not read.
• Tracking by team instead of by customer. Engineering wants a backlog. The account owner needs to know what Acme is owed. Same data, and only one view prevents a bad renewal conversation.
• Letting the summary replace the quote. Once the customer's wording is gone you cannot tell whether you satisfied the commitment or something adjacent to it.
• No declined state. Teams avoid marking things declined because it feels like failure. An unmarked commitment silently stays open forever and surfaces at renewal.
• Relying on the rep who made it. Reps change accounts and leave. A commitment stored in one person's memory or notebook has a shelf life measured in months.

When a spreadsheet stops working

The manual version works well up to roughly fifty calls a quarter, or one seller. Past that, three things break, in this order:

1. Capture becomes inconsistent. Some reps log, some do not, and you cannot tell which — so the list is incomplete in an unknown way, which is worse than empty.
2. Nobody links back to the call. The quote gets typed once and the recording is never referenced again, so disputes cannot be settled.
3. Nobody closes the loop. Shipping happens in one system, the commitment lives in another, and connecting them is manual work that gets skipped first.

What tooling should do about it

If you reach that point, the thing to buy is not another recorder. Every meeting tool records well. What matters is whether a commitment becomes a durable object: extracted from the conversation automatically, linked to the customer and the quote, routed to the team that owes it, and closed when the work ships.

That is what BuildBetter does. It records the call, extracts commitments alongside objections and requests, keeps the customer's exact words attached, and connects them to the project that satisfies them — so "what is Acme owed?" is a filter rather than an archaeology exercise. It also imports existing call libraries, so commitments buried in calls you already recorded are not lost.

Two honest limits. If you are a single founder doing four calls a month, a spreadsheet is genuinely fine and you should not buy anything. And if what you actually want is rep coaching and scorecards, a conversation-intelligence tool like Gong is purpose-built for that and more complete — it is simply priced per sales seat, which means the product and CS people who need to read the commitments never get a login.

Frequently asked questions

Who should own commitment tracking? The account owner should own the list. The individual commitments need individual owners. Do not give the whole thing to sales ops as a reporting exercise — it becomes a report nobody acts on.

Isn't this what the CRM is for? A CRM tracks the deal, not the promises made to win it. You can bolt a custom object on, and some teams do. It works if you also solve capture and linking back to the call.

What about commitments made over email or Slack? Same rules. The reason calls get the attention is that calls are where the unrecorded promises happen — email at least leaves a searchable trail.

How far back should we go? Start today, then backfill your largest accounts only. A complete history of small accounts is not worth the effort; the open commitments on your top ten accounts are.

What if we decide not to honour one? Mark it declined and tell the account owner immediately. The damage from a declined commitment is almost entirely in how late the customer finds out.

Sources and further reading

• Gartner Peer Insights — Revenue Intelligence Platforms
• Atlassian — Product requirements documents, downsized
• Harvard Business Review — The One Number You Need to Grow