The Challenge
Meridian Advisory Group had grown from a boutique strategy consultancy of four partners to a twelve-person firm with a diversified client base across financial services, infrastructure, and technology. The growth was real and sustained — but it had outpaced their operational infrastructure entirely.
The core problem was data fragmentation. Client engagement data lived in Salesforce. Time tracking was in Harvest. Financial reporting ran through Xero. Project milestones were in Asana. Client communications history was spread across email threads and a shared drive. Analyst utilisation — a critical metric for a people-led business — required a manual cross-reference of three of these systems to calculate.
Every Monday morning, the firm's two senior analysts spent the first half of the day assembling the weekly management report. The process was the same every week: export from Salesforce, export from Harvest, pull the Xero report, merge and reconcile in Excel, fix the formatting, send. The whole process took between ten and fourteen hours across both analysts — time that Meridian was paying £180–£220 per hour for.
The deeper problem was reliability. The manual assembly process introduced errors. Not catastrophic ones, but persistent low-level inconsistencies: a client listed as active in one system and pending in another, a timesheet entry that pre-dated a project start, a revenue figure that didn't reconcile against the Xero invoices. Leadership had quietly stopped trusting the numbers. The weekly report was produced and circulated, but the strategic conversations it was supposed to inform were increasingly happening off the back of gut instinct rather than data.
The Approach
We began with a two-week diagnostic before writing a single line of code. This phase had three components: a process audit to map exactly what data was being collected and how it moved, a data quality assessment to understand what was reliable and what was not, and a requirements alignment session with the partners and both analysts to agree on what the dashboard actually needed to answer.
The diagnostic produced three conclusions that shaped the entire build. First, 80% of the report content came from two systems — Salesforce and Harvest — with Xero contributing the remaining 20%. The other three systems were referenced rarely and could be addressed in a later phase. Second, the data quality problems were addressable with transformation rules, not with data re-entry. Third, the analysts' most important use case was not the weekly report itself, but the ability to answer unstructured questions quickly — "how much time have we spent on client X this quarter?" — without building a new query each time.
This reframed the build. We were not building a reporting tool. We were building a query layer with a pre-built weekly report as its primary output.
The technical implementation had three components. An integration layer that pulled from Salesforce, Harvest, and Xero via their APIs on a four-hour refresh cycle, normalised the data into a unified schema, and applied the reconciliation rules identified in the diagnostic. A data warehouse layer — a structured PostgreSQL database — that stored the normalised data with full history, enabling trend analysis and period comparisons. And a dashboard layer built in Retool, configured to Meridian's specific reporting requirements, with the weekly report auto-generated every Monday at 07:00 and delivered by email to the partners.
The build phase took six weeks from specification sign-off to go-live. The first two weeks were integration development and testing. Weeks three and four were the warehouse schema and reconciliation logic. The final two weeks were the dashboard configuration and user acceptance testing with both analysts.
The Outcome
The weekly report now generates automatically. Both analysts reviewed the first three auto-generated reports against their manual equivalents and found the numbers to be more accurate — the reconciliation logic caught several systematic errors in the manual process that had been running undetected for months.
Report assembly time dropped from twelve hours per week across both analysts to under forty-five minutes for review and distribution. At Meridian's blended analyst rate, this represents approximately £180,000 in annual time savings — time that has been reallocated to client deliverables and business development.
The unstructured query capability has been used more than anyone expected. Partners now answer client questions in real time during calls rather than following up the next day. One partner described it as "the difference between having an analyst in the room and having to call one."
The data quality improvement had a secondary effect that Meridian did not anticipate: the partners started using the management report again. Trust, once lost, returned when the numbers began to reconcile reliably.
