Beneath MarketingSeason 3Episode 1
Why does branch 3’s coffee taste better than branch 30’s?
In short: The recipe isn’t the card; it’s the whole branch. The Geographic Replication Index compares your latest wave with your reference branch at the same age.
The recipe isn’t the card on the kitchen wall. It’s the whole branch
Abstract
Branch 3’s coffee is better not because the beans are better, but because branch 3 was built with you standing over it. The fix isn’t a clearer recipe card. It’s one rule: copy exactly first, and improve only after the numbers match.
Before you sign branch thirty, make sure the copies’ numbers match; replication is the fourth test of the opening gate.
When to come back to this episode: Before every wave of openings, and when someone proposes to improve the recipe in the new branch.
From the riyal to the machine
Season 1 of Beneath Marketing was about the customer; Season 2 was about the riyal. This season is about the machine that branch 30 runs on: the recipe, the people, the menu, the incentives, the calendar, the borrowed ideas and head office.
Eight episodes ask one question: will branch 30 work the way branch 3 works? Expansion adds nothing new. It repeats what you already have, success or mistake, thirty times.
A cup on the way
Composite example, not a specific client
Thursday night. On his way to the istiraha, the group’s owner stops at branch 30 and orders the same Saudi coffee and dates his guests praise at branch 3.
The coffee is thinner. The dates are a different grade: “the supplier ran out of the usual ones.” The barista has been at the branch five weeks. The recipe card on the wall is identical, word for word.
At the istiraha a friend says: “Your new branch isn’t like the old one.” His first thought: the barista.
But the barista didn’t break the card. The card just didn’t say everything.
The dalla
Every Saudi house has a dalla, the coffee pot. Same beans, same cardamom, and yet the coffee is different in the grandmother’s hands. Because what she does isn’t written down: how high the fire is, how long the pot rests, the ratio she adjusts by smell rather than by spoon.
That unwritten part is what doesn’t travel. Branch 3 had the grandmother: you, standing over it, tasting, correcting, choosing the supplier. Branch 30 got the card.
The idea beneath the dalla
In 2001 Sidney Winter and Gabriel Szulanski argued that replication, the “McDonald’s approach” of running many similar outlets, is a strategy in its own right. It needs two things: a template, a working example you copy from, and the knowledge of which features actually produce the result, which they called the “Arrow core”.1
The problem is that you don’t know which details are the core. Maybe the cup temperature, maybe the grade of dates. As long as you don’t know, any “improvement” made while copying is a gamble.
In 2002 they wrote “Getting It Right the Second Time” in Harvard Business Review: attempts to copy a best practice usually fail when people improvise, and managers are overconfident about what makes the original work. Their advice: pick a template that works, copy it exactly, including the details that seem silly, and adapt only once the copy produces the same results. One of their examples is Intel’s “Copy Exactly” policy.2
That doesn’t mean never adapting to Jeddah or Abha. It means adapting after the numbers match.
The one number
How do you know the dalla travelled? Does the new branch earn what your reference branch earned at the same age?
Geographic Replication Index = median of (new branch’s contribution margin ÷ reference branch’s margin at the same age), at months 6 and 12, for the latest opening wave
An index of 1.0 means branch 30 earns like branch 3.3
Illustrative example: at month 6, branch 3 made a 20% contribution margin. At month 6, the latest wave of four branches made 17%, 15%, 19% and 13%. The ratios are 0.85, 0.75, 0.95 and 0.65. The median is 0.80.
The median says the system isn’t travelling as it should; the ratios say where to start: at the 0.65 branch, not the 0.95 one.
ChartIllustrative example
The latest wave against branch 3 at the same age
Source: Illustrative example from the episode. Bands: Playbook Gate 5. (6)
Data as a table
| Branch | Margin at month 6 | Ratio to reference | Band |
|---|---|---|---|
| New branch 1 | 17% | 0.85 | amber |
| New branch 2 | 15% | 0.75 | amber |
| New branch 3 | 19% | 0.95 | green |
| New branch 4 | 13% | 0.65 | red |
| Reference 1.0 | 20% | 1.00 | — |
One honest line: I found no published Saudi benchmark for replication across restaurant branches. Your only reference is your own branch 3, at the same age.
From my own practice: I worked in a group that went from 3 to 33 branches over two to three years, as reported by the group.4 I’m not saying an index did that. I’m saying that at that speed you are copying from a copy, and any error in the template multiplies.
Where the money leaks
Every point below 1.0 is margin lost on every riyal a new branch sells. If each of the four branches sells SAR 200,000 a month, their combined gap to the reference is 16 margin points: SAR 32,000 a month, SAR 384,000 a year, from one wave.
ChartIllustrative example
What one wave loses every month
Source: Illustrative example from the episode.
Data as a table
| Margin points | SAR a month | |
|---|---|---|
| New branch 1 | 3 | 6,000 |
| New branch 2 | 5 | 10,000 |
| New branch 3 | 1 | 2,000 |
| New branch 4 | 7 | 14,000 |
| Total a month | 16 | 32,000 |
| Total a year | — | 384,000 |
The bigger leak is signing branch 31 on branch 3’s numbers. The payback plan you wrote for the savings circle5 rests on a margin that won’t arrive, and the gap only shows after you’ve signed.
That’s why the Doctrine calls a falling index the “earliest warning” that the system isn’t travelling.3 It shows up before complaints and ratings, and before your friend’s remark at the istiraha.
SimulatorIllustrative example
Copy the grandmother, or the last copy?
Move how many steps drift per new branch, then switch what you copy from: thirty branches slide from green to red, and the latest wave's index gives the Sunday verdict.
Result at the default values (turn on JavaScript to move the numbers).
- GRI, latest wave
- 0.84
- Green / amber / red branches
- 12 / 14 / 1
- Margin lost against the reference
- 123,300 a month · 1,479,600 a year
Watch 0.84: fix what isn’t travelling before the next wave. Start at branch 28 (0.69).
Rule source: Playbook default to calibrate, Gate 5
Illustrative model; the 0.9 and 0.75 bands are Playbook defaults to calibrate, not a market benchmark. Idea: Winter & Szulanski 2001, 2002.
Data as a table
| Branch | Ratio to reference | Band |
|---|---|---|
| branch 28 | 0.69 | red |
| branch 29 | 0.90 | amber |
| branch 30 | 0.84 | amber |
| Band | Branches |
|---|---|
| green | 12 |
| amber | 14 |
| red | 1 |
How it’s calculated
- Branches 4 to 30 open in waves; each new branch drifts from the template by the steps × its multiplier (0.5 to 1.5)
- Copy of a copy: the drift multiplies by the wave number. Copy from the reference: it doesn’t
- Branch ratio = 1 − (loss per step ÷ 100) × drift, floored at zero
- GRI = the median ratio of the latest wave; 0.9 or above green, from 0.75 amber, below 0.75 red
- Margin lost a month = Σ (1 − ratio) × reference margin × branch sales
The Sunday-morning question
One question, one metric, one rule
| Question | “Does our latest wave of branches earn like our reference branch at the same age?” |
|---|---|
| Metric | Geographic Replication Index: median of (new-branch margin ÷ reference-branch margin at the same age), at months 6 and 12 |
| Decision rule | A starting point your group sets in writing:6 0.9 or above: replicate. 0.75 to 0.9: fix what isn’t travelling before the next wave. Below 0.75: no new branch until the template is rebuilt. |
This week
- Name your reference branch in writing. Not necessarily your best branch, but the one you’re willing to copy exactly.
- Compute the index for your latest opening wave at month 6, and write each branch’s ratio next to the median.
- List ten “small” details at the reference branch that aren’t on any card: the supplier grade, the resting time, who opens, the cup temperature. Copy them before you improve anything.
Copy the grandmother, not the card
A dalla doesn’t travel on a sheet of paper. It travels when you move the whole house: the fire, the time, the dates, the hand. Then you improve.
Credit where it’s due: Winter and Szulanski gave replication its name and its discipline, and the dalla comes from our homes. I only put the two on the Sunday-morning table.
But the detail nobody writes down lives in an employee’s head. What happens when that employee leaves? That’s the next episode: “Your cheapest employee is your most expensive one.”
Sources and notes
Sidney G. Winter & Gabriel Szulanski, “Replication as Strategy”, Organization Science 12(6), 2001, pp. 730–743. doi.org/10.1287/orsc.12.6.730.10084 (opens in a new tab)
Gabriel Szulanski & Sidney G. Winter, “Getting It Right the Second Time”, Harvard Business Review, January 2002. hbr.org/2002/01/getting-it-right-the-second-time (opens in a new tab)
Revenue Engineering Doctrine: Thesis 6, replication is the asset, measuring replication and the GRI definition; lexicon entry Geographic Replication Index. 12
From the author’s own practice, as reported by the group for the period stated; no public source exists. Revenue Engineering Doctrine: evidence from practice.
Beneath Marketing, Season 2, Episode 1: Your branch is a savings circle: you pay first and collect last.
The 0.9 and 0.75 bands are the defaults in the Saudi Scale Playbook: Gate 5, replication to a second city, meant for calibration, not a market benchmark. The Doctrine sets no number; it requires the format to meet “the group’s GRI floor”. Compute yours with the Replication Index Calculator.
Cite this
English
Alhajj, S. (2026, October 3). Why does branch 3’s coffee taste better than branch 30’s?. Beneath Marketing, season 3, episode 1 (BM S3.E1). seiflden.com. https://seiflden.com/insights/branch-thirty-coffee-pot/
Arabic
الحاج، سيف الدين (3 أكتوبر 2026). «ليش دلّة فرعك الثالث أطيب من دلّة الفرع الثلاثين؟». «تحت التسويق»، الموسم 3، الحلقة 1 (BM S3.E1). seiflden.com. https://seiflden.com/ar/insights/branch-thirty-coffee-pot/