Throughwork · Sales performance engineering

Sales performance is rarely lost evenly across the process.

Flow 7 works through your sales process one stage at a time. It finds where you're losing deals, works out what your strongest people do differently at that stage, and turns it into a standard your managers can actually hold the team to — so you convert more of the pipeline you already have, and grow revenue without growing the team.

Stage conversion — illustrative ◤ Constraint identified at stage 05
01
Targeting
31%
02
Engagement
24%
03
Qualification
46%
04
Discovery
61%
05
Proposal
19%
06
Negotiation
58%
07
Handover
88%

The point isn't to make everyone try harder everywhere. It's to find the one or two places where the process is actually costing you, and fix those. Figures illustrative — yours would come from your own data, where it's good enough to rely on.

In brief

What this is, plainly stated

Sales optimisation for the AI age. That's the short version. Before the argument — the facts, so you can decide in thirty seconds whether to keep reading.

What it is
A fixed-fee piece of consulting that finds where your sales process is losing deals, works out what your better people do differently there, and turns it into something you can measure.
Also available
Two shorter companion engagements: Buyer-Side AI Readiness (what AI systems tell your buyers about you before you know they exist) and Sales Leader Support (an independent read of the sales operation for an incoming sales lead). Both can be commissioned on their own.
Who it's for
B2B organisations with a defined sales team, recurring opportunity volume and complex, multi-stakeholder sales cycles — including platform, data, payments, content and compliance vendors selling into regulated markets. Analytical viability depends on opportunity volume, outcome balance, process stability and data quality, not a fixed seller-count threshold.
Who it's not for
B2C operators looking to improve player acquisition or retention. Businesses whose real constraint is product-market fit, pricing or demand generation — sales execution work will not fix those, and we'd decline the engagement.
Method
Seven steps, grouped into five phases. Work out what the data can support, map the stages, find the patterns, test whether they're real, write the standard, pilot it, then measure — keeping adoption, process and revenue separate so you can see what actually moved.
Evidence basis
One first-hand delivered result — 15%+ improvement, delivered manually before comparable tooling existed. One delivery, not a portfolio average and not a forecast. Every other material figure is externally attributable and cited with source, sample and date; illustrative examples are labelled where they appear.
Pricing
Fixed fee per phase, published below. Diagnostic from £4,000. Never a day rate.
Requirements
A CRM in daily use and lawful access to relevant sales records. Full scope may include recorded conversations where lawful, with any required transparency or consent established before analysis. Where recordings are unavailable, the method uses CRM and correspondence evidence only, and states the limitation.
Delivered by
Throughwork Ltd, Liverpool. Version 3.0, July 2026.
Why this exists

CRM gives you visibility. It doesn't give you a standard.

The software mostly does what it said it would. The question nobody can answer is whether any of it changed how selling actually happens.

4.8 hrs

Saved per seller, per week, by AI tools. The capacity gain is real and it is already happening.

Gartner, 2026 · 210 sales leaders
72%

Of sales organisations report low reinvestment of that reclaimed time into high-value sales activity.

Gartner, 2026 · same survey
5%

Of enterprise GenAI pilots achieved revenue acceleration. The other 95% showed no measurable impact on profit and loss.

MIT NANDA, 2025 · 300 deployments

AI is giving sales teams time back. Better capture, summarisation and automation mean sellers can spend less time on admin and more time selling. But saving time doesn't automatically improve sales performance. The missing piece is knowing what good execution actually looks like at each stage of your sales process — and making sure that extra capacity is spent on the things that genuinely improve outcomes.

MIT looked at 300 enterprise deployments and found the same thing. One detail is worth sitting with: most AI budget goes into sales and marketing, and that's where the measured return was lowest. Their explanation wasn't the technology — it was that organisations couldn't fit it into how they already worked.

Stated narrowly, because that's what the research supports. Gartner found 4.8 hours saved per seller per week and 72% of organisations reporting low reinvestment; that does not establish that AI fails to generate revenue, and we don't claim it does. MIT defined success as measurable P&L impact within six months, which excludes efficiency and pipeline effects, and the study has been publicly criticised on that basis — we'd concede the point if challenged. The direction of travel is the argument, not the decimal place.

The idea underneath it

Your best closer may be an average qualifier

Nobody is the best at everything. So the thing worth studying is the stage, not the salesperson.

The usual approach is to take whoever's billing most and hold them up as the model. The trouble is you get their weak stages along with their strong ones — which is why "go and shadow Dave for a week" so often doesn't work. New hires end up copying a personality instead of a method.

So we look at each stage on its own terms. What does winning look like here? Who's actually good at it? What do they do that others don't? We're looking for the small details that genuinely work — then turning those into something your whole sales team can repeat.

What you end up with is built from the best of what your own team already does, stage by stage — rather than from an assumption that one person has it all figured out.

Why small gains matter

Get 5% better at four consecutive stages and you finish roughly 22% ahead. That's multiplication rather than a forecast, and you can check it yourself in ten seconds. It's also why a handful of small, evidenced fixes tends to beat one large promise.

What's actually new

None of this is a new idea — good sales leaders have always known their best people do things differently. The problem was cost. Listening to enough calls to spot the pattern took weeks, so most teams did it once, from a small sample, and never went back. The playbook quietly went out of date. AI makes it cheap enough to look at every call instead of a handful, and cheap enough to look again next year.

Who this is for

Five situations where this tends to be worth doing

Not an exhaustive list, but if you recognise your business here the conversation is usually a short one.

Founder-led sales, building the first real team

The founder sold everything up to this point, mostly on instinct and relationships. None of that is written down, and it doesn't transfer cleanly to the first three hires. Worth getting right before you scale it, because whatever you set up now is what the next twenty people will learn.

Post-raise, with targets attached

You've taken the investment and the numbers that came with it. There isn't time to work the process out by trial and error across four quarters, and the board will be asking about conversion long before that.

Growing the team quickly

Going from eight sellers to twenty-five multiplies whatever inconsistency you already have. Onboarding by shadowing works at eight and stops working somewhere around fifteen, usually without anyone noticing until the ramp times slip.

Merged teams after an acquisition

Two sales teams, two processes, two different definitions of a qualified opportunity — and a combined forecast that nobody quite believes. Deciding whose process wins is a political question until somebody makes it an evidence question.

Entering the UK market

Your methodology works at home, so the assumption is that it travels. It usually does, apart from in one or two specific places — UK buying groups, procurement behaviour and how much directness a first meeting will take are the common ones. Better to find out which than to rebuild the whole thing.

And when it isn't

If your sales process is genuinely well documented, consistently followed and already measured stage by stage, you've done this work. Tell us that on the call and we'll say so rather than sell you something.

Scope

Use what you already own

Your CRM and whatever conversation tooling sits alongside it already do most of the capturing, summarising and automating. We're not here to replace any of it.

Worth checking your platform first

Churn prediction. Anomaly and abuse detection. CRM lifecycle automation. Support deflection. Compliance monitoring. Content localisation.

What your platform covers depends on which one you're on, which edition, and how it's set up — and it's changing fast. Before paying anyone else to build these, it's worth asking your account manager what's already there or coming.

We don't sell these, and we'd rather say so than find out three weeks into an engagement.

What Flow 7 adds

What none of them give you is a definition of what good looks like in your business, drawn from your own results.

That's what Flow 7 builds — clear stage definitions, findings you can trust, specific behaviours to start and stop, all of it dropped into the tools your team already opens every morning.

If your existing platform can deliver that standard, use it. Building anything separate is optional, and only worth it when the numbers say so.

The seven-stage diagram on this page is illustrative. Flow 7 maps your actual sales process during the diagnostic, and those stages replace the illustrative model.
The method

Seven steps, grouped into five phases

Flow 7 identifies behaviours associated with stronger and weaker outcomes at each stage, grades the evidence, and turns what survives scrutiny into observable operating standards. Each phase produces a usable output and ends at a decision gate.

Phase 1 · Diagnostic 2–3 weeks · £4,000 – £6,000 indicative
1
Readiness

What the available data can and cannot support. Sometimes the honest answer is that it can't yet — you'd get that verdict and a remediation plan rather than findings built on unreliable inputs.

2
Define stages

The process as it is actually executed, with entry, exit and measurement criteria, and a baseline view by stage where the data supports it.

Phase 2 · Extraction & validation 3–5 weeks · £7,000 – £12,000 indicative
3
Extract

Candidate signals from calls, CRM, correspondence, wins, losses and stalls — lost and stalled opportunities can be particularly instructive.

4
Validate

Test alternative explanations against the obvious confounders, and grade every finding for evidence strength.

Phase 3 · Codification 2–3 weeks · £4,000 – £7,000 indicative
5
Codify

Defensible findings become observable standards, scoreable rubrics and an anti-pattern set — specific behaviours to stop, which may deliver value faster than adding advanced ones.

Phase 4 · Pilot installation 4–6 weeks · £5,000 – £9,000 indicative
6
Install

Pilot the standard inside the tools your team already uses — pre-call briefs, post-call review against the rubric, alerts on known failure patterns, and a manager view that replaces generic one-to-ones.

Phase 5 · Retained improvement quarterly · £1,500 – £3,000 per month indicative
7
Measure

Stage metrics tracked against the pre-engagement baseline where comparison is credible, with adoption, process movement and commercial outcome reported separately. Scheduled re-extraction keeps the standard current.

Automated analysis generates candidate signals. Human validation determines what is defensible enough to become a finding — and a finding only acts with the authority its evidence supports. Fees are indicative, not quotations. Fixed fee per phase, never a day rate. Each phase is separately scoped and no phase requires the next to have been worth doing.
The deliverable

What a finding actually looks like

It's hard to buy something you can't picture. So here's the shape of what you'd actually get — the structure is real, the numbers are made up.

Stage 03 — Qualification Evidence grade: strong
Naming the economic buyer before the second conversation is associated with materially stronger stage progression.

Evidence. Opportunities where the economic buyer was named before call two progressed to proposal at 34% against 19% where they weren't. The difference held after segmenting for account size and lead source.

Illustrative operating change. Named economic buyer becomes a mandatory exit criterion for Stage 03 — a gate, not a target.

Stage 05 — Proposal · anti-pattern Evidence grade: strong
Proposals issued without a prior verbal walkthrough are associated with a materially higher rate of late-stage loss.

Evidence. 62% of Stage 05 losses had no recorded conversation between proposal issue and client response. Won opportunities included a walkthrough in 81% of cases.

Illustrative operating change. Proposals are not issued cold. The walkthrough is booked before the document is sent, not after.

Stage 04 — Discovery Evidence grade: indicative
The strongest qualifier asks fewer questions than the team average, but a higher proportion are open and consequence-framed.

Evidence. Question count roughly 30% below team mean; proportion of open questions roughly double. Sample size limits confidence — the pattern is consistent but not statistically established.

Illustrative operating change. Framework issued and coached, flagged for re-testing at the next cycle before being treated as settled.

You're not buying insight. Insight is what a consultant leaves behind on a slide. You're buying a change to how the work gets done, that someone can be held to.

Note the third example. Weak findings are labelled weak, including where the honest conclusion is "consistent, but not yet established". A finding only acts with the authority its evidence supports — and client policy, not the grade alone, determines whether that authority is granted.
The obvious question

Why not just do this yourselves?

It's a fair question, and for some teams the answer is that you should. Here's what an outside pair of hands actually buys you.

People tell an outsider things they won't tell you

A seller who thinks the discovery framework is a waste of time isn't going to say so to the person who wrote it. Saying it to us costs them nothing. That's not a comment on your culture — it's true almost everywhere, and the most useful thing we hear is often the thing nobody has said out loud internally.

It's hard to see a process you're inside

Everyone on your team learned the process the same way, so the assumptions baked into it are invisible from where you're sitting. And the person who designed it is rarely the right person to judge whether it works — not through any lack of capability, just because nobody can be neutral about their own thinking.

It's not that you can't. It's that you won't get to it

Most sales ops teams could do a version of this. They also have a quarter to close. Analysis loses to pipeline every single time, which is exactly why the playbook gets written once and never revisited. A fair amount of what you're buying is simply the fact that it happens.

The same finding lands differently

"The CRO thinks we should qualify harder" is an opinion from someone with authority over your career. "Here's what happened across four hundred opportunities" is evidence. Same conclusion, very different reception — and only one of them tends to change what people actually do on Monday.

When you genuinely don't need us. Picture a RevOps function that's completely on top of this. Stage definitions are written down and every seller uses them the same way. Conversion is measured at each stage, and the numbers are trusted enough that people argue with them. Somebody listens to calls systematically rather than dipping in when a deal goes wrong. They can tell you what the best qualifier does differently from the average one, and it's documented. New starters learn from that document rather than by sitting next to whoever's free that week. Someone has worked out which behaviours consistently precede a loss, and managers are actively coaching them out. The whole thing gets revisited annually rather than written once in 2023. And none of it quietly gets shelved when the quarter gets tight.

If that's your business, you've already done this work and you should keep your money. Genuinely — say so on the call and we'll agree with you. Everyone else recognises about half of that list, which is usually where the conversation starts.

And whatever we do build, you own afterwards. The standard is yours. The retained phase exists to keep it current as your market moves, not to keep us in the building.
Companion engagement

What does AI tell your buyers about you?

Buyers now research, compare and shortlist before you know the opportunity exists. A vendor can lose without ever appearing in the pipeline.

On the 6sense evidence, buyers purchase from a Day One shortlist vendor 95% of the time, and 94% rank that shortlist by preference before contacting anyone. Meanwhile 67% of B2B buyers now prefer a rep-free buying experience.

Conventional pipeline, conversion and win-rate analysis begins once an opportunity has been recorded — so that loss does not appear in it.

The Buyer-Side AI Readiness audit is a separate engagement that tests what leading AI systems currently tell prospective buyers about you, using a controlled and dated question set — then identifies what to change, and re-tests.

What you get

  • A standard question set tested across multiple model families, graded against reality
  • A source map — which sources the models actually draw on
  • Objection register, separating genuine gaps from documentation gaps
  • Competitive simulation by buyer scenario, not "who's best"
  • Prioritised remediation, and a re-test date

£3,000 – £5,000 indicative. Faster and cheaper than the Flow 7 diagnostic, and can be commissioned independently of it.

Two honest limits. Model outputs vary by model, prompt and date, so this is a dated snapshot rather than a measurement — we record the test set so change is detectable. And improving how AI describes you doesn't guarantee shortlist inclusion: prior experience with a vendor remains the most decisive factor, and no amount of documentation overcomes a poor previous engagement.
Companion engagement

You've just hired a sales leader. Now they need six months.

Hiring a sales lead costs a six-figure salary plus fees. Then they spend their first quarter working out what they've inherited — largely by asking the people whose work they're assessing.

It's nobody's fault. A new leader has no independent view of the process they've taken on, so they build one from conversations, and those conversations are with people who have a stake in the answer. Meanwhile the clock on their targets is already running.

Some of the biggest decisions they'll make — who to keep, what to restructure, where to invest — get made in that window, on the least reliable information they'll ever have.

Sales Leader Support hands them an independent read of the sales operation, a roadmap of what to address in what order, and a timeline built around the targets they've actually been given.

What they get on day one

  • An independent report on the current state of the sales operation — process, stages, data quality, and where performance is being lost
  • What's working and worth protecting, stated as plainly as what isn't
  • A prioritised roadmap: what to fix first, what can wait, what to leave alone
  • A timeline mapped against their actual targets, not a generic ninety-day plan
  • The evidence behind each recommendation, so they can defend it upward

£6,000 – £9,000 indicative. Commissioned either before they start, so they land with a map, or in their first ninety days, when they know what they need to test.

Worth being clear about what this isn't: it's not an assessment of the incoming leader, and it's not a review of their predecessor. Leadership owns the strategy. This independently tests whether the current system supports it — which is a considerably easier thing for a new leader to commission, and a considerably easier thing for the existing team to be part of. The report is written for them and they get it first; we won't take the engagement on any other basis.
Evidence

Material claims trace to a named source, or to work we delivered

Illustrative figures and arithmetic examples are labelled where they appear. Four defensible figures beat twelve of mixed provenance.

15%+

Sales performance improvement, delivered first-hand. Seven-stage decomposition inside an enterprise AI platform business serving regulated industries — run manually, before comparable tooling existed.

Throughwork · founder-led
4.8 hrs

Saved per seller per week by AI tools, with 72% of organisations reporting low reinvestment of that reclaimed time. An association, not our outcome claim.

Gartner, 2026 · 210 sales leaders
5%

Of enterprise GenAI pilots achieved revenue acceleration. Diagnosis was workflow integration, not model quality.

MIT NANDA, 2025 · 300 deployments, 52 interviews, 153 surveys

What we won't claim. Flow 7 has one measured delivery behind it, not a portfolio. We don't quote a blended average improvement across clients, because there isn't one. Any figure specific to your business would come from your own baseline at the diagnostic — and from nowhere else.

Illustrative client findings on this page are not evidence of a Throughwork outcome. The 22% compounding example is arithmetic, not a forecast. Any client-specific performance figure comes from that client's own measured baseline.

Governance

Four controls, designed in rather than added afterwards

The method analyses customer conversations and employee performance data. In regulated businesses, these controls can determine whether a project is permitted to proceed at all.

Human validation

Automated extraction generates candidates, not conclusions. Any material finding requires traceable source examples and human review before it is codified.

Read-only first

Where automation is used, the system assesses before it acts. Write access and stage gating are granted by client policy, not by the model.

Coaching, not discipline

Findings are framed at process level. Individual-level output is used for coaching, not for disciplinary or capability proceedings. Agreed in writing before extraction begins.

Challenge and correction

Sellers and managers have a defined route to contest a classification and supply missing context. Disagreement is logged as evidence about the analysis, not treated as resistance.

Lawful basis, data minimisation and any required transparency or consent are established before analysis begins — not discovered during a security review.
Boundaries

Where this doesn't work

You've probably been pitched AI a dozen times this year. Here's where ours doesn't help.

When the constraint isn't execution

If the product isn't landing, the pricing is off, or you're chasing the wrong customers, tightening the sales process just helps you lose faster. The diagnostic is partly there to catch that — and if it does, we'll tell you and stop.

Small teams

Below a certain deal volume the numbers won't support firm conclusions, and we won't pretend otherwise. It's not simply headcount — it's how many deals you close, how consistent the process is, and how good the records are. Where the maths doesn't hold, we compare cases openly instead, and price it as the lighter piece of work it is.

Observation changes behaviour

People who know they're being scored will play to the score. That's half the point and half the danger — a rubric rewards what you can count, and some of what makes a good seller good doesn't count easily. We treat it as a floor, never a ceiling, and we say that to the team out loud.

Where recording isn't possible

The full version needs lawful access to recorded calls, with whatever transparency or consent that requires sorted out first. Without recordings we work from CRM and email alone — that's genuinely weaker evidence, and we'd price and promise it that way rather than pretend otherwise.

Who delivers it

Twenty-five years carrying a number, before codifying how

Paul McNea — founder

Twenty-five years in enterprise B2B sales across regulated markets — iGaming, financial services and payments. Built £1.5m ARR from zero as a sole commercial hire. Sold an enterprise AI platform into tier-one financial services. Earlier career at KPMG, Barclays and Hargreaves Lansdown, with financial services qualifications.

Flow 7 came out of that work rather than out of a framework. The seven-stage decomposition was first delivered by hand inside an enterprise AI platform business, and produced a measured 15%+ improvement. AI didn't create the method — it made it affordable to run properly, and to run twice.

Regular chair and moderator on innovation and technology panels across the major industry conferences.

How we work

  • The person you meet does the work. Nobody hands you off to a junior team after the pitch.
  • One recommendation, not an eighty-page report. Start where the return is clearest.
  • Measured, then reported. Success metric agreed upfront, outcome report before any renewal decision.
  • Findings are graded. We tell you how confident we are, and we label the weak ones weak.
  • Your team keeps the judgement. Standards raise the floor. They don't lower anyone's ceiling.
Start here

Which stage is costing you most?

If you can't answer that with evidence, that's the conversation. Thirty minutes, no slides — and you'll come away knowing whether a diagnostic would tell you anything you don't already know.

Privacy

How we handle your data

Throughwork Ltd is the data controller for personal data collected through this site. We are based in Liverpool, United Kingdom. For any data protection query, contact paul@throughwork.uk.

When you email us we collect only what you choose to provide — typically your name, email address, company and the content of your message. We use it to respond, to discuss potential work and to maintain our relationship with you. We do not use it for automated decision-making or profiling, and we do not sell personal data.

We process enquiry data on the basis of our legitimate interest in responding to you, and, where we are discussing or entering an engagement, on the basis of taking steps at your request prior to entering a contract.

Where we are engaged by a client and process personal data on their behalf — for example, sales records analysed during a project — we act as processor under a written agreement with that client, who remains controller. This website collects no such data.

You have the right to access the personal data we hold about you, to have it corrected or erased, to restrict or object to its processing, and to data portability. To exercise any of these, email the address above. You may also complain to the Information Commissioner's Office at ico.org.uk, though we would welcome the chance to resolve any concern first.

Last updated July 2026.