Your average is hiding the problem

I worked with a sales business where the leadership team was trying to improve conversion. The overall number wasn't disastrous, but it had been stubbornly below where they wanted it for long enough that people had started suggesting the usual collection of remedies: more sales training, better accountability, changes to commission and, inevitably, some fairly robust opinions about the quality of recent hires.

The lightbulb moment came when we stopped looking at the overall conversion rate.

There were several teams apparently selling the same service, under the same brand, using the same CRM and ostensibly following the same sales process. They had access to the same training and were operating in broadly comparable parts of the market. Their results, however, were anything but comparable. One team was converting consistently well. Another was some distance behind it. New starters in one part of the business seemed to become productive considerably faster than new starters elsewhere, and there were meaningful differences in retention too.

The company average wasn't wrong. It just wasn't telling us very much.

That's a distinction I think matters in growing businesses. We become understandably interested in whether performance is going up or down, whether the number is above or below target and how we compare with the market. But averages have an unfortunate habit of smoothing away exactly the variation that might tell us where to look next.

Sometimes the most useful question isn't “How are we performing?” It's “Why are these two apparently similar parts of our business performing so differently?”

Same company, different company

This isn't particularly unusual.

Gallup has spent decades examining differences between workgroups and has repeatedly found substantial variation inside the same organisation. Its research attributes around 70% of the variance in team-level employee engagement to the manager, based on factors including manager behaviour, manager engagement and employees' perceptions of their manager.1 Gallup's current research continues to report the same finding across its work on engagement and performance.

We need to be precise about what that does and doesn't tell us. Engagement isn't the same thing as sales performance, and the finding certainly doesn't mean that managers explain 70% of the difference in revenue between teams. It does, however, illustrate something commercially important: people working for the same organisation can experience meaningfully different working environments depending on the team they happen to join.

There is stronger evidence that this can extend to productivity itself. Metcalfe, Sollaci and Syverson (2023) analysed store-level data from two multibillion-dollar retail businesses where managers moved between locations while many of the organisation's formal management practices remained fixed. That made it possible to separate, to some extent, the effect of the individual manager from the wider company system. They found that managers explained a substantial share of variation in store productivity.2

More recent experimental research reaches a similar conclusion from a different direction. In a study published in The Quarterly Journal of Economics, researchers repeatedly assigned managers to different three-person teams, allowing them to observe whether particular managers consistently improved team output rather than simply benefiting from stronger people. A one-standard-deviation increase in managerial skill improved team performance by 0.22 standard deviations, even after accounting for the productive skills of the people in the team.3

The important point isn't that every performance gap is secretly a management problem. It's that “they work for the same company” tells us considerably less about people's working conditions than we tend to assume.

Two salespeople can have the same job title, compensation plan, CRM and sales methodology while working in quite different performance environments. One manager coaches before the forecast becomes a problem; another intervenes afterwards. One team shares information freely; another hoards it. One new starter gets regular opportunities to practise difficult conversations; another is protected from them until somebody decides they're “ready”. One team challenges weak opportunities early; another carries them hopefully through the pipeline for six weeks.

On the organisation chart, those teams look almost identical. In practice, they may barely be doing the same job.

Variation is diagnostic information

This is where I think leadership teams can make a fairly simple mistake. They see an overall performance problem and respond with an overall intervention.

Conversion is down, so everybody gets sales training. Engagement is poor, so the whole management population attends the same leadership programme. New starters are taking too long to become productive, so onboarding gets redesigned centrally.

Sometimes that's exactly the right answer. If the problem is genuinely widespread and the underlying cause is common, a common intervention makes sense.

But suppose the company-wide conversion rate is 24%, while comparable teams are converting at 34%, 29%, 21% and 14%.

The average is still 24%. I want to know what the hell is happening between 34% and 14%.

Not because the highest-performing team necessarily has the answer. There are plenty of reasons apparently comparable teams produce different results that have nothing to do with capability. Lead quality might differ. Territories may not be as comparable as everybody thinks. One team may have inherited better accounts, more experienced people or a temporary market advantage. Small sample sizes can produce exciting-looking differences that disappear once enough data arrives.

But once you've tested those explanations, persistent variation becomes incredibly useful.

I saw this in another business where two managers had teams that looked broadly similar on paper. One manager was regarded as particularly demanding; the other was generally seen as more supportive. It would have been easy to turn that into a slightly tedious argument about management styles.

When we looked at what actually happened, the distinction was much more practical. One manager had developed a rhythm around performance. Expectations were clearer, coaching happened closer to the event, difficult conversations happened earlier and people knew when they were expected to solve something themselves rather than escalate it. The other manager was working just as hard, probably harder, but much more of that work was reactive.

The useful difference wasn't personality as the exec team had assumed; it was what happened repeatedly in one team that mapped directly to their stronger outcomes.

That matters because repeatable differences give you somewhere to intervene.

And this is where internal variation becomes much more valuable than simply identifying your “best team”. If one part of the business consistently gets new hires productive faster, don't congratulate them and move on. Find out what happens during those first 90 days that doesn't happen elsewhere. If one manager's team retains people longer, look at the experience people are having before deciding the answer is another retention initiative. If one office converts similar opportunities at a materially higher rate, examine the work before assuming the explanation is better salespeople.

The organisation may already contain a working solution to its own problem.

Before benchmarking outside, benchmark inside

Growing businesses understandably spend a lot of time looking outwards.

What are competitors doing? What does “best practice” look like? What conversion rate should we achieve? How much should we spend on learning? Which sales methodology is performing well elsewhere? What are high-performing companies doing with AI?

Those can all be useful questions.

But before I benchmarked externally, I'd want to know how much unexplained variation already exists internally.

If your Birmingham team is consistently outperforming Manchester under genuinely comparable conditions, an industry benchmark is less immediately interesting than understanding Birmingham. If one manager routinely develops competent new hires in four months while another needs seven, I want to understand that difference before buying a new onboarding platform. If one team uses the same CRM as everyone else but somehow maintains considerably better data and forecasting discipline, perhaps we don't have a technology problem.

This is particularly important in businesses that have grown quickly. As organisations expand, local practices emerge. That's inevitable and often useful. Good managers adapt. Teams discover shortcuts. People solve problems the central organisation hasn't noticed yet. Over time, however, that means what leadership thinks is one operating model can become several subtly different versions of it.

• Some of those differences will be improvements.
• Some will be harmless.
• Some will be absolutely murdering performance.

Averages make all three harder to see.

Find the useful difference

I wouldn't turn this into a six-month analytics project.

Take one commercially important outcome where performance varies: conversion, margin, retention, productivity, time-to-competence, customer satisfaction, whatever genuinely matters in the business.

Break it down by a meaningful operational unit. Team is often the obvious place to start, but it might be manager, office, cohort, region or tenure. Then look for differences large enough and persistent enough to be interesting.

The next stage is the important one. Don't immediately try to fix the bottom. Compare.

Are the inputs genuinely similar? Is opportunity quality comparable? Are we measuring the same thing consistently? Are there structural reasons for the difference?

If those explanations don't account for it, get closer to the work.

What happens differently? How are managers spending their time? How frequently are people coached? What happens when somebody makes a mistake? How are opportunities reviewed? How does information move around the team? Which parts of the formal process are actually followed? Where have teams developed their own practices? What does the stronger team do routinely that the weaker one does occasionally, or not at all?

Somewhere in that comparison may be a training need. There may be a coaching problem, a management capability gap, a broken process or a piece of technology making good performance unnecessarily difficult. Equally, you may discover that the difference you thought was behavioural is actually being created by lead allocation, incentives, workload or something else entirely.

That's why the diagnosis comes first.

If four comparable teams are producing four materially different outcomes, sending all four through the same intervention may create consistency in the training calendar while doing absolutely nothing to create consistency in performance.

The goal isn't to eliminate variation. People are different, markets are messy and no sensible organisation should expect every team to produce identical results. The goal is to become curious about unexplained variation.

Because a company average tells you what happened across the business. The differences underneath it can start to tell you why.

And in a growing organisation, one of the cheapest places to look for the next performance improvement may not be a consultancy report, a new piece of technology or another external benchmark. It may already be happening somewhere inside your own business.

You just averaged it away.

RB