The situation

Five reports arrived every Monday from five different parts of the group, each in its own format and each prepared by a different person. By the time they had been read and reconciled it was Wednesday, and the numbers were already a week old.

Nothing in the pile answered the question the owner actually had, which was simple: what changed since last week, and does anything need a decision today.

What we built

• One screen showing live figures across every entity in the group
• A weekly movement view so changes stand out without reading the detail
• Market signals pulled from outside the business, sitting alongside internal numbers
• An activity feed of what actually changed, and who changed it

The part that mattered

Most of the work was not the dashboard. It was agreeing what each number meant, because three of the five reports had been calculating the same figure differently for years. That conversation took longer than the build.

Result

REPLACE — add the outcome in your own words once you are comfortable publishing it. Time saved, decisions made earlier, reports retired.

The situation

The community had grown past the point where any one person could hold every member in their head. Introductions were the whole product, and they were happening by accident.

What we built

• A member directory with structured profiles rather than free text
• Suggested introductions based on what members said they were looking for
• An admin console for approving members and moderating the directory
• A newsletter system that goes out without anyone assembling it by hand

The part that mattered

Matchmaking is easy to get wrong in a way that damages trust. We built it to suggest rather than to connect, so a person still decides whether an introduction is worth making. The system proposes; a human presses send.

Result

REPLACE — member growth, introductions made, or whatever the honest measure turned out to be.

The situation

Enquiries came in on WhatsApp at all hours. The team answered them the next morning, by which point a share of those customers had gone elsewhere. The questions themselves rarely varied.

What we built

• An agent answering from the company's own documents, not from general knowledge
• Replies calibrated against twenty real past messages, so it sounds like the business
• A hard stop on pricing, complaints and anything involving money
• Full conversation handed to a person the moment it goes past what it knows

The part that mattered

The first conversation was not about what the agent should say. It was about what it must refuse to answer. An agent that invents a refund policy costs more than the staff time it saves.

Result

REPLACE — response times, share of enquiries handled without a person, or after-hours conversion.

The situation

Three days of every month went into assembling a report by hand. It arrived accurate and late, and it displayed figures without explaining them, so the meeting it fed always started with the same question: so what changed?

What we built

• The pack assembled automatically from source, on schedule
• Movement explained in plain language, not just charted
• Exceptions surfaced first, with the routine detail below
• A weekly version, because monthly was too late to act on

The part that mattered

We shortened the reporting cycle before we made it clever. A weekly report that says less beats a monthly one that says more, because you can still do something about the week.

Result

REPLACE — days recovered, decisions brought forward, or the meeting that stopped being needed.