The statistics you actually use

Chidi reads one number more than any other: "average basket size." It is also the single easiest number to accidentally misrepresent, in four specific ways worth knowing by name.

Outliers pull it	One ₦180,000 wholesale order among two hundred normal ₦4,000 baskets drags the average up enough to make every branch look busier than it is. The median wouldn't move.
It hides a bimodal split	An "average" basket size of ₦8,000 can mean everyone spends around ₦8,000 — or it can mean half the customers spend ₦2,000 and half spend ₦14,000, with almost nobody near the average itself.
It changes meaning across group sizes	Averaging nine branch averages together is not the same as averaging every individual transaction — a branch with fewer, bigger baskets gets equal weight to one with many small ones.
It erases the trend inside the period	A month's average can look flat while masking a real decline in the second half that a single number never shows.

"What's the average?" is rarely the real question. "What's typical, and how much does it vary?" usually is — and that needs at least two numbers, not one.

Two branches can post the exact same average daily revenue and be running completely different businesses — one steady day after day, the other swinging wildly between a great Saturday and a dead Tuesday. The average alone can't tell them apart; the spread can.

You don't need a formal standard deviation calculation to make this point to Chidi — the range (highest minus lowest) or simply eyeballing daily values against the average usually says enough. The habit that matters is remembering to look, not the specific formula.

Branches that run more promotions also tend to have higher footfall. It would be easy — and wrong — to conclude promotions drive footfall. The confounder here is obvious once you look for it: bigger, busier branches both run more promotions (they can afford to) and naturally get more footfall (they're in busier locations). The promotion isn't causing the footfall; branch size is driving both.

A three-question habit before claiming causation

1. Could a third factor explain both? Branch size, location, season — something driving both variables at once.
2. Which came first? If footfall was already rising before the promotion started, the promotion can't be the cause of that earlier rise.
3. Would the relationship survive controlling for the obvious confounder? Compare promotion-vs-no-promotion within branches of similar size, not across all nine branches mixed together.

"Yaba's numbers were down 20% this week" sounds alarming. Whether it should worry Chidi depends enormously on how much a branch's week-to-week numbers normally bounce around — which is a sample-size question in disguise.

If Yaba's weekly revenue normally varies by ±25% just from ordinary randomness — different weather, different days of the week the month lands on, one big or small order — then a 20% dip is comfortably inside normal noise and probably means nothing. If Yaba's weekly revenue is normally rock-steady within ±3%, a 20% dip is a real signal worth investigating immediately.

"Is this a real change, or just noise?" is answerable with one look at the branch's normal week-to-week variation — you rarely need a formal significance test to make a sound call.

Sahara Mart trialled a new receipt layout with a loyalty-points reminder at three branches for two weeks, against six branches running the old layout. The trial branches showed 4% higher loyalty sign-ups. Is the new receipt the reason?

What to check before answering yes

• Were the trial and control branches actually comparable beforehand? If the three trial branches were already trending upward before the test started, the receipt isn't the explanation.
• Is 4% bigger than the normal week-to-week gap between these branches anyway? Apply lesson 5.4's habit directly here.
• Two weeks — long enough? A change this small, over a short window, at only three branches, is thin evidence for a permanent rollout decision.

The honest answer to Chidi is rarely a clean "yes, roll it out" or "no, scrap it." It's closer to: "the early signal is positive and worth continuing to test, but two weeks at three branches isn't enough to be confident it'll hold at scale — recommend extending the trial before a full rollout."