Sales reps spend 78% of the week on non-selling work. Here's where all 31 hours go - and why they're so hard to win back.

A sales rep spends about 22% of the working week in conversation with customers and 78% on everything else. In a 40-hour week that is 9 hours with customers and 31 without.

The 31 is not one job. It is five. Prospecting takes 7.2 hours, quote-building 6.8, planning 6.4, CRM data entry 5.2, and training 4.4.

This is usually described as a sales productivity problem. We think that diagnosis is incomplete.

Those proportions come from Salesforce's 2026 State of Sales survey of 4,050 sales professionals across 22 countries, and they sit close to most other measurements of the last fifteen years. Ebsta and Pavilion, working from observed pipeline and call data across 387 companies rather than from a survey, put customer time at 21%. No credible study we found puts the average above a third.

So the numbers are not in dispute, and there is not much point publishing them again.

The question nobody answers is the interesting one: why does each of those blocks cost what it does? Not why sellers are busy - why this particular work, which everyone agrees is necessary, resists every attempt to make it smaller. Fifteen years of tooling has been aimed squarely at it, and the share of the week that reaches a customer has stayed where it was.

Our answer, and the argument of this piece: the 31 hours look like five separate problems. They are not. They are five places where a seller builds, rebuilds, translates or hands over context - and most of that work would not exist if what the seller already knew survived the trip from one task to the next.

The five sections below are the evidence for that.

The published figures are shares of the week. The hours below convert them against a 40-hour week, which is the conservative reading: most B2B sellers work more than 40, and if they do, the shares stay the same while the hours grow.

Prospecting: 18% of the week, 7.2 hours

Almost none of this is talking to prospects. It is deciding who is worth approaching, finding a credible reason to make contact, locating the right person, writing something that isn't generic, and chasing the ones who didn't reply. This is context assembly, done from scratch, one account at a time.

It costs 7.2 hours because the work is priced per attempt, not per conversation. A considered approach might take twenty minutes, and most approaches go nowhere. The hours are set by how many attempts it takes to get one conversation, and the market decides that ratio, not the rep.

Which makes prospecting the only work in the week with no natural stopping point. A quote is finished when it is sent; preparation is finished when the meeting starts. Prospecting is finished when the rep runs out of week.

It also explains why faster research has never shrunk this block - it buys more attempts in the same 7.2 hours. Better targeting is the only lever that moves it, because it changes the ratio: fewer approaches wasted on accounts that were never going to answer means fewer attempts per conversation. And targeting is a context problem, not a speed one - knowing which accounts resemble the ones that closed, and who inside them actually engaged last time. That is what fifteen years of tooling still hasn't given the rep.

Creating quotes: 17% of the week, 6.8 hours

The name of this category is doing a lot of quiet work. In earlier editions of the same survey it was written out in full: generating quotes and proposals, and gaining approvals.

That is two jobs, and only one is the rep's. This is context transfer.

Assembling a quote - line items, price list, document - is the part the rep controls. The rest happens after they have finished: discount authority, legal review of a non-standard clause, finance signing off payment terms, a decision-maker who is travelling. Much of this block is queue time.

Queue time is not free, and that is what gets missed. Each handoff generates chasing: a nudge, a Slack message, a re-explanation of the deal to somebody who has no context on it and needs the background before they can approve anything. The rep is not idle while the quote sits in an inbox; they are doing the work of moving it.

Which is why quoting software has not shrunk this block. CPQ tools speed up assembly; they do not speed up a VP of Sales with forty unread emails. The bottleneck is other people's attention - the one input a seller cannot buy their way out of.

No published study splits this block into assembly and approval time, so we cannot give you the ratio. Our hypothesis, from conversations with sales teams, is that much of the 6.8 hours sits downstream of the rep finishing the quote. It is a hypothesis, not a measurement.

Planning: 16% of the week, 6.4 hours

This is preparing for calls, building account plans, working out what to do about a stalled deal, and getting ready for pipeline reviews. Very little of that time is spent deciding anything.

Most of it goes on rebuilding context that already existed - reassembling the state of an account before any thinking can start. What was said on the last call. What was promised, and by whom. Why it went quiet in June.

That information exists. It is in an email thread, a meeting transcript, a Slack channel, someone's memory, and partially in the CRM. Assembling it takes twenty minutes before a call that lasts thirty.

Here is the part worth sitting with: planning is where an organisation pays for thin data entry. They are not two problems on two rows of the table; they are the same problem, observed at two different moments.

Capturing what happened costs a few minutes. Reconstructing it later costs far more, and it is not paid once - it is paid by the rep before every call, by the manager before every review, and all at once by whoever inherits the account, which is why territory changes cost more than the org chart suggests.

It persists because the two costs land in different places. Skipping the update saves the rep five minutes today, visibly. The reconstruction is paid later, by somebody else, in an hour nobody counts.

Manually entering data: 13% of the week, 5.2 hours

5.2 hours a week is a strange amount of time to spend typing, and the reason is not typing speed. Writing a paragraph about a call takes two minutes.

What takes the time is translating context into a shape somebody else designed. A sales conversation is unstructured: forty minutes, three people, digressions, a hint about budget timing, one half-commitment and one unresolved objection. The CRM wants that as a stage, a next step, a close date, an amount, a competitor field and eleven custom properties. Converting the first into the second is cognitive work, and it is slow in a way that has nothing to do with the interface.

The translation is also lossy in one direction. Those fields exist because someone needs to report on them - they are shaped around what management sees in aggregate, not around what the rep needs to remember before the next call. So the rep spends 5.2 hours producing something whose value accrues mostly to somebody else, and predictably does it last and does it thinly.

There is a timing effect on top. The natural moment to record is immediately afterwards, which is exactly when a rep cannot - so it moves to Friday, by which point recording has become recall, and recall is both slower and less accurate.

That is why "reps don't update the CRM" has survived every generation of CRM interface improvement. It is not a usability problem. It is a problem about who the work is for.

Training: 11% of the week, 4.4 hours

This block contains two different things and the published data does not separate them. Some of it is real enablement - onboarding, product knowledge, coaching - and it is not obvious a company should want that number lower. The rest is tool overhead: the same 2026 survey found teams without a single platform run an average of eight standalone tools, and that 42% of reps feel overwhelmed by how many they use. That half is context too, just one level up - learning where the information lives rather than using it. We are flagging this block rather than explaining it.

The pattern: all 31 hours are a context cost

Look at the five blocks together and they stop being five different problems.

None of them is stupid work. Every one is necessary and a well-run sales team does all of it. But read what each block actually consists of, underneath the label, and it is the same activity wearing five different uniforms:

• Prospecting is assembling context - who this account is, why now, who to approach, what to say
• Planning is reassembling context that already existed but wasn't kept
• Data entry is translating context into a structure somebody else designed
• Quoting is transferring context to whoever has to approve it, repeatedly, by hand
• Training is acquiring context about the tools that hold all the other context

Not one of those is selling. All of them are the overhead of keeping information available to the people who need it - including the rep's own future self.

That reframes the seams, too. The handoffs are expensive not because handoffs are inherently slow, but because context does not survive them. A quote crosses to an approver and arrives stripped of the reasoning behind it, so the rep re-explains it. A call ends and the understanding in the rep's head does not reach the record, so someone rebuilds it next week. An account moves to a new rep and eighteen months of accumulated judgement evaporates, because almost none of it was ever written down in a form anyone else could use.

This is, we think, why fifteen years of sales software has made the individual tasks faster without moving the share of the week that reaches a customer. Almost every sales productivity tool has been aimed at making individual blocks faster - better quoting, better prospecting databases, better CRM interfaces, better enablement platforms - and speed was never the constraint. Each of those tools holds one slice of context and hands nothing useful across its own edges, so the rebuilding cost is untouched and simply reappears at the next boundary.

The instinctive fix is to consolidate the tools, and that is only half right. Buying five modules from one vendor gets you five tools with one logo on them. It does not, by itself, mean anything a rep learned in one shows up in another. Consolidating vendors is not the same as consolidating context, and the second is the one that pays.

What follows from this

If the 78% is a context cost, then the lever is not automating tasks one at a time. It is making context persist across all of them - captured once, carried everywhere, available to whoever needs it next. That is a different design goal from "make this screen faster," and it changes what each block looks like.

Prospecting shrinks through aim, not speed. A system that knows which accounts closed, which signals preceded them and who actually engaged last time produces better-targeted approaches, and better targeting means fewer attempts per conversation. That is the only mechanism that moves this block, because the block is priced per attempt.

Planning approaches zero when there is nothing to rebuild. This is the cleanest win in the table. Twenty minutes of reconstruction before a thirty-minute call exists only because the last conversation was never properly captured. Capture it and the preparation becomes reading, not archaeology.

Data entry becomes describing a change, not translating one. If the system already holds the state of the account, the rep supplies the delta - they pushed to Q1, Anna is the buyer now, pricing is the open objection - and the structure is the system's problem rather than the rep's. Five hours of translation is not five hours of typing, and it is the translation that goes.

Quoting improves at the transfer, not the assembly. Most of the chasing exists because approvers receive a request without the reasoning behind it. Send the context with the request and more approvals happen without a conversation. Being honest: this is the block that moves least, because part of the queue is genuinely other people's calendars and no software fixes that.

Training falls with the number of places to learn. Fewer interfaces, and the ones that remain answering questions rather than requiring a search.

Three things worth being straight about.

None of these goes to zero. Deals still need quotes, prospects still need researching, and somebody still has to think. The claim is that a meaningful share of 31 hours is rebuilding information that already existed somewhere - not that the 31 hours are waste.

A single block is a capped prize. Remove CRM data entry entirely and perfectly and that is 13% of the week; 65 points of the 78 are untouched. This is exactly why the end-to-end version matters more than any individual fix: the blocks compound. Better capture makes planning cheaper, which makes the record better, which makes targeting better. Optimised separately, each of those is small. Connected, they feed each other.

Recovered time does not become selling time on its own. Gartner surveyed 210 sales leaders in early 2026 and found AI is already saving sellers 4.8 hours a week - and that 72% of organisations report low reinvestment of that time into higher-value work. Somebody has to decide, in advance, which customer conversations those hours become. No tool does that by itself, and any vendor implying otherwise is describing a saving, not an outcome.

And if you want to know whether any of this is working in your own team, stop watching the total. It will barely move and it will tell you nothing. Watch the places where context gets rebuilt: how many days from quote requested to quote approved, how many hours after a call before it is recorded, how many systems a rep opens to prepare for a meeting, how long a new rep takes to be useful on an inherited account. Those are the numbers that move first, and almost nobody tracks them.

Where Spiich fits

Spiich is built on the argument above, so we should be explicit about what it covers today and what it doesn't.

The premise is a single contextual layer over the systems a team already runs, rather than another tool holding another slice. The loop is three steps: capture the context automatically, resolve it into the CRM, then put it back in front of the rep at the moment they need it.

Capture is not a task we hand back to the seller. Spiich records across email, in-person meetings and digital meetings, so what happened reaches the system without anyone stopping to file it. A rep who wants to add a judgement or a correction can simply say it or type it.

That context then does the work rather than sitting in a record. It writes the structured CRM data, prepares the next meeting brief, drafts the follow-up, and researches and qualifies prospects in the background on a schedule or off a CRM trigger - putting what the rep needs in front of them before the call, rather than waiting to be asked.

So yes, this automates a good deal of the 31 hours. But the automation is not the reason it works. A follow-up drafted from one call is a template; a follow-up drafted from every interaction that account has ever had is a useful piece of work. End-to-end context is what makes the difference between the two.

On the table above that covers data entry, most of planning, and a real part of prospecting - three of the five blocks. It does not do quoting or approvals, and it will not make a VP of Sales answer their email faster. It is not a CRM and does not replace your system of record - it writes into the one you already run, whether that is HubSpot, Attio, Pipedrive or Salesforce. It is not an AI SDR and does not run sequences on your behalf - it does the research so the conversation is better aimed, rather than the outreach so there is more of it.

If you want to put currency on any of it, we don't publish a cost-per-hour here - put your own numbers in.

No single tool takes the whole 78% down. Claiming otherwise would be a roadmap dressed up as a product.

What we do believe, and what this whole piece argues: the 78% is not immovable, and it does not come down one tool at a time.