4 MCP servers replaced 6 SaaS seats + an analyst per quarter.

StowFast's ops + marketing teams needed continuous visibility into Uberall (935 UK locations, citation health, review velocity), Google Search Console (organic traffic by location, query performance, indexation), GA4 (acquisition, on-site behaviour, conversion paths), and Serpstat (ranking position, competitor visibility, keyword opportunities).

Each platform was on a per-seat licence. The team had 6 seats split across 4 platforms — not enough for everyone, too expensive to scale. The de facto workflow: an ops or marketing person describes the question in Slack, the analyst translates it into 3-5 platform queries, exports CSVs, builds a small dataset, replies in Slack a few hours later. 30+ requests per week. The analyst was the bottleneck on every decision.

Off-the-shelf BI tools couldn't fix it. Connecting Uberall + GSC + GA4 + Serpstat into a single dashboard sounded fine in theory, but the team's questions weren't dashboard-shaped. They were investigative: 'which locations had review velocity drop this month and also lost top-3 rankings on their primary query?' That's not a chart — that's a conversation.

Each MCP server wraps a single platform's API. The Uberall MCP exposes location lookups, citation health, review streams, and insights. The GSC MCP exposes URL inspection, sitemap status, and search analytics queries. The GA4 MCP exposes property metadata, custom dimension lookups, real-time and historical reports. The Serpstat MCP exposes 65 SEO tools (keywords, rankings, backlinks, competitor analysis, audit data).

All four are read-only and OAuth-gated. Team members add the MCPs to their Claude.ai workspace as connectors; queries are scoped to their permission level. The MCPs run as systemd services on internal infrastructure (subdomains under andyslab.uk), no client data leaves the firm's stack.

Cross-platform queries — the hard ones — are handled by Claude. The team asks 'which UK locations had reviews drop below 4.0 this quarter and also lost their #1 ranking for the brand+city query?' Claude calls the Uberall MCP for review trends + the Serpstat MCP for ranking history, joins the data, and answers in 90 seconds with citations.

The previous SQL queue went to zero within two weeks. The 6 per-seat licences dropped to 2 (the analyst + a single marketing manager who still uses the native platforms for edge cases). The analyst's time was reinvested in proactive analysis rather than reactive query-building.

SQL request volume to the analyst dropped from 30+/week to effectively zero within two weeks of rollout. Marketing and ops teams now answer their own questions; the analyst spends time on the higher-judgement analysis that actually needed their craft.

SaaS seat count dropped from 6 to 2. The savings cover the MCP infrastructure many times over. More importantly, the team isn't artificially gated on who can ask data questions — every person who needs data has Claude-mediated access to it.

The same MCP pattern is now being extended to internal systems (CRM, billing, support tickets) — same architecture, same access model, same conversational interface. The pattern compounds.