Reading a franchise network through its WhatsApp groups

To analyse an operations chat group, read a month of it by hand and sort what you find into four piles: the questions people ask, the routes their answers take, the questions that die without a decision and the places where someone quietly worked around the standard. Look for patterns and never for culprits. Tell the network you are reading. Done that way, a franchise network's WhatsApp groups will show you how it actually runs, which is seldom how the manual says it runs.

Where the work actually lives

Every network I have worked in has two operating systems. One is written down. It has version numbers, a sign-off page and a training module attached. The other lives in a handful of chat groups, and it is the one people use at eleven in the morning with a patient waiting.

Nobody designed the second one. It grew because a message to a group gets an answer faster than a ticket does, and because the person who knows the answer is in the group and not in the ticket queue. Head office tends to treat this as a compliance problem to be tidied away. I think that misreads what the groups are. They hold the only complete record of what the network asks, decides and improvises, written by the people doing the work, with a timestamp on every line.

Most organisations never read that record. They read the dashboard built on the formal system, which sees only the fraction of the work that was routed through it.

What the surgical teams showed

Medicine got here first, and the published work is useful because it shows what a transcript can reveal that a survey cannot.

A team at a London hospital, writing in the American Journal of Surgery, moved an emergency surgical team of forty people onto WhatsApp for nineteen weeks and then analysed who said what to whom. The study logged 1,495 communication events. Speed was the headline. The more useful finding was the shape of the team. The most senior surgeons sent most of the instructions. The most junior doctors asked most of the clinical questions. The middle grade answered fastest. Participants also said the group made the hierarchy feel flatter, which is a finding about culture drawn from a pile of short messages.

An orthopaedic team in Dublin reported in Cureus on six months of their own group. They counted 5,492 messages, and 5,090 of them concerned patient care. That matters for anyone tempted to dismiss chat groups as social noise. In a working clinical team, the group is mostly work.

A franchise network is the same thing at a greater distance. Swap the junior doctor for a new clinic coordinator and the senior surgeon for a regional lead, and the transcript will tell you who really carries the network's knowledge.

The four reads

I read a month of one group with four questions in mind, one pass for each. It is slower than trying to do all four at once and far more accurate.

**Questions.** Mark every message that asks something, then group them by what they ask about. The recurring ones are the interesting ones. A question asked once is a question. The same question asked by four clinics in a month is a missing paragraph in the standard, and every answer given in the group instead of in the manual guarantees that the fifth clinic will ask again.

**Routes.** For each question, note who answered, by role. In most networks I have looked at, a very small number of people answer nearly everything. That feels like good service until one of them goes on leave. It also shows you where authority really sits, which is often somewhere other than the organisation chart. If clinical questions are being settled by someone with no clinical remit, you have found a governance gap that no audit of the written system would ever reveal.

**Silences.** Look for the questions that got no reply, or got sympathy without a decision. Then look for the clinics that have stopped posting. A quiet clinic reads as a well run clinic on a dashboard. In a transcript it often reads as a clinic that has decided asking is not worth the wait.

**Workarounds.** Search for the phrases people use when they improvise. We did it like last time. Just for today. Until the stock comes in. Each of those is a place where the lived process has already moved away from the written one. Most are harmless and a few are sensible enough to adopt across the network. You cannot tell which until you have found them.

Patterns, not people

This is the part that decides whether the exercise helps or poisons the groups for good.

The UK Information Commissioner's Office published guidance on monitoring workers in October 2023, and its logic travels well beyond Britain. Be open that you are doing it. Name a specific purpose before you start. Do not let information gathered for one purpose drift into another. Ask the people affected for their views first, and consider a formal impact assessment even where one is not strictly required.

In operating language, that gives me three rules. I tell the network, in the group itself, that I am reading a month of messages to find gaps in the standard. I strip names and replace them with roles before any analysis leaves my desk. And nothing found in the reading goes into anyone's appraisal, including the message that made me wince.

That last rule has a cost. Now and then you will read something that shows a person doing what they should not. Fix the process that allowed it. If it was genuinely unsafe, it goes through the clinical safety process like any other report, openly. What it never becomes is a performance note built on a quoted message. Use the transcript against one person and you lose the record for everyone.

The first time a chat message turns up in a performance review, the real conversation moves somewhere you cannot read.

What the reading should produce

A month of reading should end in a short list, and every item on it should be one of three kinds of change. A paragraph added to the standard, for each question that recurs. A routing change, where one person is carrying answers that belong to a role. Or a named owner, for each type of question that currently dissolves into agreement.

If an item cannot be written as one of those, it is an observation, and observations are how these exercises end up forgotten in a slide deck.

I repeat the reading every quarter. The second pass on the same group is the one that earns its time, because it tells you whether the questions you wrote into the standard have stopped being asked.

Where this leaves you

You will not stop a network running on chat groups, and you should not try. The groups exist because they are faster than anything head office has built. Ban them and the traffic moves to private messages, where nobody can learn from it at all.

What you do not control is what gets said in them, who says it or how many questions a growing network produces. What you do control is whether anyone reads the record, whether the people in it know why, whether a recurring question becomes a written answer and whether the reading stays about the process. Hold those four and the groups become the most honest audit you have. Ignore them and the manual will go on describing a network that stopped existing some time ago.

Questions people ask

How do you analyse a chat group for operations problems?

Read a month of it by hand and make four passes. Mark the questions and group them by topic. Note who answered each one, by role. Find the questions that never got a decision and the sites that stopped posting. Then search for the language of improvisation. Recurring questions point to gaps in the written standard, and concentrated answers point to knowledge held by too few people.

Is it ethical to read staff WhatsApp groups for operational analysis?

It can be, if the reading is open and limited. Tell the group you are reading it and why before you start. State one purpose, such as finding gaps in the standard, and keep to it. Replace names with roles before analysis. Never let a message found this way reach a performance review. The UK Information Commissioner's Office guidance on monitoring workers follows the same logic.

What does it mean when a clinic goes quiet in a network chat group?

A clinic going quiet rarely means everything is fine. A site that has stopped asking questions has usually concluded that asking is slower than deciding locally, so its practice is drifting away from the written standard without anyone at head office seeing it. Treat a sudden drop in questions from one site as a reason to call or visit, and never as evidence of good performance.