03 · Advertiser Managers
Stop Tabbing Between Dashboards: Affise MCP for Advertiser Managers
Ask questions about your advertisers in plain English — performance, payouts, trafficback, and offer health, without opening a single CSV.
Managing advertisers on Affise means handling many things at the same time. You monitor offer performance across geos. You check payout discrepancies before they turn into disputes. You watch trafficback rates that point to traffic quality problems. And you prepare accurate updates for brand partners. Meanwhile, the platform keeps producing data faster than any dashboard can show you what really matters.
The usual workflow: click into each advertiser, filter stats by date, export the data, open it in a spreadsheet, and do the analysis by hand. By the time you have the answer, a new problem has appeared somewhere else.
The Affise MCP changes this loop. Instead of clicking through the platform UI, you type a plain-English question into an AI assistant (Claude, Cursor, or Continue.dev) and get an answer backed by live Affise data. No exports, no pivot tables. The AI picks the right Affise tool, runs the query, and returns a structured answer you can act on right away.
Meet the Affise MCP
The Affise MCP (Model Context Protocol) is a bridge between your Affise account and an AI assistant. It gives the AI a set of purpose-built tools it can call for you. You stay in a conversation; the AI handles the API calls. The server runs at https://mcp.affise.com and connects through OAuth. Setup takes about two minutes (see the quickstart guide).
What You Can Do
1. Advertiser and Offer Performance at a Glance
When an advertiser asks how their campaigns are performing, you should not have to build a report by hand. Ask directly:
›"List all advertisers with active offers."
›"Performance by offer for advertiser \"Acme Mobile\" last 30 days."
You do not need to look up an ID first — just name the advertiser. The MCP matches the name against your advertiser list, resolves it to the internal ID, and runs the query. If the name matches several advertisers, it shows you the options so you can pick one. If you prefer to be exact, pass the 24-character MongoId string from the platform URL, and it is used directly. Either way, you get a full performance picture: revenue, conversions, and trends, grouped by offer. You can break down any slice you need without writing a single filter.
The result shows which offers drive results and which fall behind — before the advertiser brings it up.
›[screenshot: chat showing offer performance breakdown by CR and revenue for a single advertiser]
2. Payout and Conversion Reconciliation
Payout discrepancies create the most tension in the advertiser relationship. Catching them early, before invoices go out, is the difference between a quick correction and a difficult call.
›"Pull confirmed conversions for offer 95119 last 7 days with payouts."
›"/analyze_conversions on the last result with analysis_type: payouts."
The affise_conversions_raw tool pulls per-event conversion data — status, payout, affiliate, geo, and device — for up to 365 days. The analyze_conversions prompt with the payouts lens then flags where payouts look inconsistent: unusual rates, affiliates with strange payout patterns, or conversions that were approved but look odd next to similar ones. Run this weekly before reconciliation, and the gap between what you track and what the advertiser expects gets smaller.
3. Geo and Device Breakdown Per Offer
Traffic quality varies a lot by geography and device type. An offer can be profitable in one market and lose money fast in another — and the total numbers hide it.
›"Revenue and conversions by country last 30 days for offer 95119."
›"Show me device and OS breakdown for offer 95119 — last 14 days."
The stats tools understand natural-language slice requests. Ask for any combination of geo, device, OS, or affiliate, and you get back a structured table without touching the report builder. The 6-month maximum date range covers any standard review period. Then pair the output with the analyze_conversions prompt using the geo_tech lens to see which geos or device segments produce the most rejections or low-quality signals.
4. Brand Protection: Trafficback and Compliance Review
Trafficback is an early warning sign of traffic quality problems. A spike in redirected traffic — especially in specific geos or from particular affiliates — is worth checking before it affects advertiser trust.
›"Show trafficback for the last 30 days, top reasons and affected geos."
›"/analyze_trafficback on the last result with analysis_type: advertiser, format: actionable."
The affise_trafficback tool returns a breakdown of rejected traffic by reason code, geo, and affiliate. The analyze_trafficback prompt with the advertiser lens looks at the output from the advertiser's point of view: which offers take the most redirected traffic, and what the reason distribution says about compliance or targeting issues. The actionable format produces a prioritized list of what to fix, not just a summary of what happened.
5. Offer Health and Spotting Early Decline
Offers do not fail overnight. The CR drops step by step, the geo mix shifts, partners slowly lose interest. By the time it shows up in a monthly report, the advertiser has already noticed. Catching the decline early gives you time to respond.
›"Get offer 95119 detail."
›"/analyze_offers with analysis_type: performance, format: actionable."
The affise_get_offer tool returns the full offer configuration: targeting, caps, payout structure, status. Running analyze_offers with the performance lens against recent stats finds offers with early decline signals — conversion rate moving in the wrong direction, geo concentration growing, or cap usage that suggests volume is dropping. Use this as part of a weekly sweep across your active offers.
Going From Reactive to Proactive
The default mode for an advertiser manager is reactive: something looks wrong, you investigate, you fix it. The Affise MCP makes proactive checks practical, at a frequency manual workflows never allow.
A daily question like "Show trafficback for yesterday, top reasons" costs ten seconds. A weekly payout reconciliation pass with analyze_conversions and the payouts lens takes a few minutes instead of an hour. Geo breakdown checks become something you run before advertiser calls, not after them.
This shift matters because advertisers notice problems before they say anything. If you run these checks regularly, you spot the same signals they see — payout anomalies, geo quality drops, offer fatigue — and you raise them first, with context and a plan. That is what turns account management from a support role into a strategic one.
Breakdowns That Render as Tables
In clients that support MCP UI apps, an advertiser's offer, geo, and device breakdowns render as an interactive, sortable stats table in the chat instead of a block of text. Offer lookups come back as easy-to-scan cards. It is the same live Affise data either way — but where the client supports UI, you also get a visual you can screenshot straight into an advertiser update.
Getting Connected
Connect your AI assistant to Affise in about two minutes. The Affise MCP server runs at https://mcp.affise.com and supports one-click OAuth: open your AI client's settings, add a custom connector, paste the URL, and authorize. Full step-by-step instructions are in the client quickstart guide, covering Claude, Cursor, and Continue.dev. There are no API keys to manage by hand and no local setup.
The questions you ask every day — how are this advertiser's offers performing, are these payouts clean, where is traffic quality slipping — now take seconds instead of most of an hour. Start with the prompts in this article and build from there.