How to Run Weekly SEO Reporting for Agency Clients With Google Search Console

Weekly SEO reporting stops being weekly the moment an agency’s client roster crosses about eight accounts. Nobody has the time to pull Google Search Console for a dozen clients by hand every Monday, so weekly reporting quietly collapses into monthly reporting, which arrives too late for the response to matter. The data still exists. What’s missing is the layer that pulls it every Monday, at scale, for every client, without human labor scaling alongside.

This walkthrough gives you the manual method for one client, the scale point at which the manual method stops paying off, and the n8n workflow (free, available at the Databox Skills Marketplace) that runs the pull-and-analysis for every client every Monday and posts the results to Slack.

TL;DR

• Weekly SEO reporting at agency scale is a linear-labor problem. Fifteen minutes per client is fine for three clients and a bottleneck for fifteen.
• The manual method uses Search Console for two 30-day windows, filters out pages with fewer than 5 impressions, and looks for movers in both impression share and average position. The three signals worth acting on each week are: pages gaining share, pages losing share, and pages where position drifts while impressions hold flat.
• The judgment layer: deciding which movers matter for which client is what no automation replaces. The pull, the join, the noise filter, and the initial analysis pass all can be.
• When the client roster crosses eight to ten accounts, the manual pull stops running consistently. Weekly reporting collapses to monthly. The Monday brief you promised the client becomes a monthly PDF three weeks later.
• The Weekly SEO Performance Monitor n8n workflow closes the gap. It loops through each configured client, pulls Search Console via Databox MCP, runs per-client AI analysis, and posts individual accountability reports plus an agency-level ops summary to Slack every Monday morning.

The manual method for one client

The manual weekly SEO report for a single client is a defined, walkable workflow. It’s the thing agencies do well when they have three or four clients, and the thing that quietly stops running when they have fifteen. Before you automate anything, build it once by hand; the manual pass is what teaches you which movers matter for which kind of client, and no automation will help you calibrate that judgment.

Step 1: Pull Search Console impressions and average position for two windows

Open Google Search Console for the client property. Go to Performance → Search results. Set the date range to the last 7 days, and note the total impressions and average position. Then change the range to the prior 7 days (day 8 through day 14) and note the same two numbers.

Now switch the report view from Queries to Pages. For each of the two 7-day windows, export the impressions and average position by URL. You’ll end up with two CSVs, each listing every URL that received at least one impression in the window.

Seven-day windows are the practical minimum for weekly reporting. Daily windows are too noisy to read. Rolling 30-day windows smooth over the fast movers you want to catch. Seven and seven is the cadence that surfaces week-over-week change without the volatility of shorter windows.

Step 2: Join the two windows and filter the noise

In a spreadsheet, join the two exports on URL. You’ll want four columns per page: impressions current, impressions prior, average position current, average position prior. Add two more columns for the deltas: impressions % change, average position change (in ranking positions).

Filter out any page with fewer than 5 impressions in either window. Below that threshold, the movement is statistical noise; a page going from 2 impressions to 4 is technically a 100% increase and tells you nothing. A page going from 47 impressions to 94 is the same 100% increase and tells you something is happening. The 5-impression floor cuts the report from a spreadsheet with 400 rows to one with 60 or 80, which is the version anyone can actually read.

Step 3: Read the movers

Every page in the filtered set falls into one of four patterns. Add a Pattern column and classify each row using this rubric:

• Gaining share: impressions up more than 25%, average position stable or improving. Something is working. Either the query volume is growing or the page is winning more of an existing share.
• Losing share: impressions down more than 25%, average position stable or worsening. Something is shifting. Either a competitor ate share or a SERP feature displaced the page.
• Position drift, impressions flat: impressions within plus or minus 15%, average position moved 3 or more positions in either direction. The page is finding a new equilibrium. Worth watching, not yet worth acting on.
• Flat: everything within noise bands. Ignore for this cycle.

The three patterns worth reporting to the client are the first three. Flat pages don’t earn a mention. Sort your filtered spreadsheet by pattern, then within each pattern by impression volume. That’s your weekly brief in raw form.

The judgment layer: which movers matter

Not every mover deserves the client’s attention. The judgment step is where you decide which of the flagged pages actually make it into the Monday brief.

Three tests, applied in order:

Test 1: Does the page tie to a client priority? A gaining-share page on a topic the client considers table stakes gets less real estate than a losing-share page on the client’s flagship feature. Traffic movement on a low-priority page is noise dressed as signal. Traffic movement on a high-priority page is the whole point of the report.

Test 2: Is the pattern consistent across weeks, or a one-off? A page losing 40% of impressions in a single week could be an algorithm nudge, a SERP feature launch, or noise. Losing 40% for three consecutive weeks is a diagnosis. Never fire the alarm on a one-week movement unless the page is on the client’s priority list and the movement is severe.

Test 3: Does the pattern have a hypothesis? If a page dropped and you can name the likely cause, like “competitor published a comparison article ranking in position 3 that we didn’t have before”, you have a report worth sending. If a page dropped and the best you can offer is “impressions declined,” you have a spreadsheet, not a report.

A completed weekly brief for one client is roughly three to five flagged pages, each with a pattern label, a hypothesis, and one recommended action. Fewer than three and you’re padding. More than five and the client stops reading past page two.

Two things the manual method does poorly at agency scale

The manual method works. It works well enough that any SEO lead running a single client should keep doing it by hand. The failure mode is not quality. The failure mode is scale.

It scales linearly with the client roster. Fifteen minutes of Search Console work per client, plus another fifteen minutes of judgment and writeup, comes to roughly thirty minutes per client per week. For three clients, ninety minutes on a Monday morning. For fifteen clients, seven and a half hours: an entire working day. Nobody has an entire working day for weekly reporting, so the practical outcome is that the reports get done for the first three or four clients on Monday morning and the rest slip. By the end of the month, the clients furthest down the list haven’t received a weekly brief in three weeks. The word “weekly” quietly stops applying.

It doesn’t produce consistent judgment across leads. When one SEO lead handles all clients, the pattern classifications and the priority tests get applied the same way each week. When a team of three or four leads splits the client roster, the client with the senior lead gets a real Monday brief and the client with the junior lead gets a numbers export. That inconsistency is the thing agencies lose accounts over: the client compares notes with a peer at another agency client and realizes they’ve been getting a lower tier of work for the same fee.

Both failure modes converge on the same conclusion. Below a certain client count (roughly: eight) the manual method holds. Above it, weekly reporting quietly becomes monthly reporting whether you meant it to or not.

The workflow: what closes the gap

The Weekly SEO Performance Monitor is an n8n workflow built by Databox and shipped to the Skills Marketplace. It automates the parts of the manual method that scale poorly and leaves the judgment layer where it belongs, with the SEO lead reading the output.

Here’s what it does. Every Monday at 9 AM ( or whatever cadence you configure in n8n) the workflow loops through each client you’ve registered in a single configuration node. For each client, it pulls Search Console impressions and average position by URL through Databox MCP, joins the two datasets, filters out pages with fewer than 5 impressions, and passes the combined page-level data to an AI agent for the initial analysis pass. The agent produces a structured per-client accountability report covering which pages gained or lost impression share, position drift patterns, and a ranked list of pages to prioritize. After every client’s report has run, a second AI step synthesizes them into an agency-level ops summary: one paragraph per client, plus a top-line view across the roster. All of it lands in Slack.

The mechanical steps the workflow replaces are exactly the ones that scale linearly with client count: the pull, the join, the noise filter, and the pattern classification. The step it doesn’t replace is the judgment layer; which of the flagged pages actually goes into the client conversation, whether the pattern is a one-off or a diagnosis, and what to do about it. That call still belongs to the human reading the Slack output on Monday morning.

Data source is Search Console only, which is a deliberate scope choice. The workflow doesn’t join GA4 traffic data, engagement metrics, or conversion events into the report. If you want organic traffic and engagement in the picture, that’s a different workflow; this one is scoped tightly to the SEO performance signal that Search Console gives you and nothing else. The tighter scope is what makes the workflow useful across every client in the roster without per-client customization work.

How to install and configure it

The workflow is a JSON file you import into n8n. Total setup is about five minutes per agency, plus a few minutes per client for the configuration node.

1. Download the workflow. Click Get it free on the workflow page. The JSON file downloads to your machine.
2. Import into n8n. Open your n8n instance, go to Workflows → Import from File, and select the downloaded JSON. The workflow appears in your n8n workspace, with all nodes and connections intact.
3. Connect Databox via the MCP node. The workflow uses n8n’s MCP connector to pull data from your Databox account. The setup guide included with the download walks through the connection. You’ll need a Databox account with Search Console connected for each client property. Databox’s free plan includes the integrations required.
4. Configure the client roster. Open the client configuration code node. For each client, add three things: a client name, the Search Console data source ID from Databox, and the two metric keys for impressions and average position. Databox Genie can provide the data source IDs and metric keys on request; ask it something like “give me the Search Console data source ID for [client name].”
5. Set the schedule and activate. The default cron trigger is Monday at 9 AM. Change it in the trigger node if you need a different day or time. Activate the workflow. From that point on it runs on the cron without further intervention.

The first scheduled run produces a Slack post per client plus the agency ops summary. Verify against one client’s Search Console data manually the first time. Once the outputs match what you’d have pulled by hand, the workflow is running against the same signal you were, just without the human labor scaling alongside the roster.

Run the manual method for one client before you install the workflow. The pattern classifications, the priority tests, and the “does this have a hypothesis” filter are what let you read the workflow’s Slack output as a report rather than as a raw data dump. Automation carries the pull-and-analysis at scale. The judgment layer only exists in your head, and only gets there by doing the manual pass at least once.