> ## Documentation Index
> Fetch the complete documentation index at: https://docs.innkeepr.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Set Up Seed Signals on Performance Max

## Overview

Google's **audience signals** field in Performance Max provides its algorithm with targeting guidance. By default, PMax optimizes broadly across all users. Innkeepr **Seed Signals** narrow that guidance using incremental prediction models — they identify users who are more likely to convert *because* of an ad, rather than users who would convert regardless.

This guide walks you through setting up a test-vs-control campaign structure with Innkeepr Seed Signals. After completing all steps, you will have two comparable PMax campaigns running: one with Innkeepr Seed Signals applied and one without, which lets you measure the difference.

## Prerequisites

**Technical requirements:**

* An active connection to your Google Ads account
* Sufficient data history for signal generation (30 days minimum recommended)

**Timeline at a glance:**

| Phase               | Timeframe   | What happens                                                             |
| ------------------- | ----------- | ------------------------------------------------------------------------ |
| Signal sync         | 24–48 hours | The signal is synced via GA4 and becomes available in Google Ads         |
| Learning phase      | Weeks 1–4   | Google's algorithm adapts to the new signals — no changes, no evaluation |
| Stabilization phase | Weeks 4–8   | Spend distribution and bidding settle; results are directional only      |
| Evaluation phase    | Weeks 8–12  | Reliable data for the test-vs-control comparison                         |

<Warning>
  Do not evaluate performance before the 60-day mark. PMax needs time to collect data, optimize bid strategies, and stabilize across inventory types. Early results are unreliable.
</Warning>

## How Seed Signals Work

A Seed Signal provides Google with a reference group of users. Google then finds additional users with similar characteristics to target. Innkeepr builds Seed Signals from users with high predicted incremental conversion probability, so the lookalike expansion is based on incrementality rather than general conversion likelihood.

Before configuring your Seed Signals, it helps to understand what each parameter controls:

| Parameter                | What it does                                                                                                                                            | Example                                                                                                      |
| ------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------ |
| **Lookback window**      | How far back Innkeepr analyzes user behavior to build the signal. Shorter windows reflect recent behavior; longer windows include more historical data. | 30 days = recent activity. 90 days = longer-term behavioral patterns.                                        |
| **Percentile threshold** | Which share of users to include, ranked by predicted incremental conversion probability. Lower percentages are more selective.                          | Top 50% = broader signal. Top 30% = more selective signal.                                                   |
| **Conversion lag**       | Accounts for products where purchases happen days or weeks after the initial interaction. Without it, the model may undercount users who convert later. | No lag = fast-converting products (fashion, FMCG). 180 days = longer purchase cycles (insurance, furniture). |

## Setup Instructions

**Step 1: Design Your Campaign Structure**

**Why a campaign split and not an asset group split?**

As of February 2026, Google does not provide access to click-based performance data at the asset group level through the API. Only campaign-level data includes full-funnel metrics (clicks, CTR, CPC, conversions, ROAS). Testing in separate campaigns is the only way to compare performance accurately.

**Budget allocation:**

| Campaign                          | Budget share | Purpose                 |
| --------------------------------- | ------------ | ----------------------- |
| Test (with Innkeepr Seed Signals) | 20–30%       | Measures signal impact  |
| Control (without Seed Signals)    | 70–80%       | Baseline for comparison |

<Note>
  **Why not 50/50?** A smaller test budget reduces risk during the learning phase while still generating enough conversion volume for comparison. Increase the test share once you have enough data to evaluate results.
</Note>

**Step 2: Choose Your Signal Configurations**

Start with one configuration per test campaign. Because click-based metrics are not available at the asset group level (see Step 1), running several configurations side by side inside one test campaign cannot be evaluated reliably — each configuration you want to compare needs its own campaign.

**Recommended starting configurations:**

| Configuration | Lookback | Percentile | Conversion lag | Best for                                 |
| ------------- | -------- | ---------- | -------------- | ---------------------------------------- |
| A             | 30 days  | Top 50%    | None           | Fast-converting products (fashion, FMCG) |
| B             | 30 days  | Top 50%    | 180 days       | Products with delayed purchase decisions |
| C             | 90 days  | Top 30%    | 180 days       | High-value, long-consideration products  |

<Tip>
  **Not sure which configuration fits your use case?** Contact our team — we can help determine the right parameters based on your conversion data and product categories.
</Tip>

**Step 3: Create and Apply Seed Signals in Innkeepr**

1. **Create the signal** in Innkeepr based on your selected configuration
2. **Select "Google Analytics"** as the connection — the signal is synced via GA4 and then becomes available in your linked Google Ads account
3. **Set the signal type** to **Seed**
4. **Configure the parameters** — set the lookback window, percentile threshold, and conversion lag according to your chosen configuration
5. **Link the signal** to the asset groups in your test campaign in Google Ads

**Step 4: Configure Audience Signals in Google Ads**

In the **Audience signals** field on your asset groups, include only the following:

**Keep:**

* **Innkeepr Seed Signals** — your primary targeting guidance
* **Search terms** — custom search themes that are relevant to your products

**Remove:**

* **Demographics** (age, gender, household income, parental status) — these restrict the signal unnecessarily and overlap with what Google already optimizes for
* **Interests and detailed demographics** (in-market and affinity segments) — these dilute the Innkeepr signal with broad Google segments
* **Other first-party user lists** (CRM uploads, website visitors, customer match lists) — mixing other sources with Innkeepr Seed Signals makes it impossible to attribute performance to a specific signal source

The goal is to isolate the Innkeepr Seed Signal as the primary targeting guidance. Additional inputs introduce noise and make it difficult to measure the effect of the signal.

**Step 5: Configure Campaign Settings**

Ensure parity between the test and control campaigns:

* **Identical target ROAS (tROAS)** across both campaigns — a mismatched tROAS invalidates the comparison (see Best Practices below)
* **Proportional daily budgets** — 20–30% for the test campaign, 70–80% for the control campaign
* **Identical asset groups** in both campaigns, except for the Seed Signal in the test campaign
* **Identical inventory types** enabled across both campaigns (Search, Display, Shopping, Video, Discover)

**Step 6: Pre-Launch Checklist**

Before activating your campaigns, verify that:

* The test and control campaigns have **identical tROAS** values
* The budget split is **20–30% test / 70–80% control**
* Seed Signals are applied **only** to the test campaign's asset groups
* Asset groups (products, creatives, ad copy) are **identical** across both campaigns
* **Inventory types match** across both campaigns
* **Conversion tracking** is verified and capturing all relevant conversions
* The Innkeepr **signal size** is **sufficiently large** (check in the Innkeepr dashboard)

## Monitor & Optimize

**Weeks 1–4: Learning phase (hands off)**

* Do **not** adjust tROAS, budgets, or signals during the initial learning phase
* Review the **PMax placement report** weekly to understand spend distribution
* Expect inventory shifts — campaigns with Seed Signals may allocate more spend to Display and Video early on. This is expected behavior.

**Weeks 4–8: Stabilization phase (monitor, no conclusions)**

* Check **ROAS, CPA, and conversion volume** at the campaign level (test vs. control), but treat the numbers as directional only — the campaign is still learning
* If performance is significantly off target, make **small adjustments** (a 10–20% tROAS change) — but always adjust tROAS on **both** the test and the control campaign equally, otherwise the campaign with the higher tROAS will lose spend
* Allow **7–14 days** after any adjustment before re-evaluating

**Weeks 8–12: Evaluation phase**

* Compare **ROAS, CPA, and conversion volume** between the test and control campaign
* Aim for **50 or more conversions per campaign** before drawing conclusions
* If the test campaign outperforms the control campaign, increase its budget share and allow one to two weeks for the campaigns to stabilize

## Best Practices

**Target ROAS alignment**

Set identical tROAS values across the test and control campaigns. Google Ads gives more bidding flexibility to campaigns with a lower tROAS, which skews results and makes it difficult to isolate the effect of the Seed Signal. If you need to adjust tROAS, always adjust both campaigns equally — otherwise the campaign with the higher tROAS will receive less spend.

* Use your standard tROAS as the baseline for both campaigns
* Only adjust tROAS if **both** campaigns deviate from target, and always by the same margin
* Align the tROAS value itself with your business objectives and historical performance

**Signal isolation**

The test is only valid if the Seed Signal is the single difference between the two campaigns. Do not add other audience signal inputs, change creatives in only one campaign, or enable different inventory types mid-test.

## Troubleshooting

**ROAS below target after 60 or more days**

| Check                  | Action                                                                                      |
| ---------------------- | ------------------------------------------------------------------------------------------- |
| Inventory distribution | Review the placement report — is spend concentrated in low-performing channels?             |
| Signal configuration   | Confirm that the Seed Signal configuration matches your campaign objective and product type |
| tROAS setting          | Increase it by 10–20% to push Google toward higher-quality inventory                        |
| Conversion tracking    | Verify that all conversions are captured correctly                                          |

**Spend concentrated in Display and Video**

This is common in the first 30–45 days. Before making changes, review actual ROAS — Display and Video placements may be converting efficiently despite the inventory shift.

If ROAS is genuinely below target, increase tROAS, temporarily reduce the budget, and wait 7–14 days for PMax to rebalance.

**Campaign still in the learning phase after 60 or more days**

| Possible cause        | Solution                                                                  |
| --------------------- | ------------------------------------------------------------------------- |
| Budget too low        | Increase the daily budget to accelerate data collection                   |
| tROAS too restrictive | Lower the tROAS to allow broader bid participation                        |
| Insufficient demand   | Verify that the products have adequate search volume                      |
| Too many campaigns    | Consolidate — small budgets spread across many campaigns prevent learning |

**Low conversion volume in the test campaign**

| Possible cause         | Solution                                                                                         |
| ---------------------- | ------------------------------------------------------------------------------------------------ |
| Signal too restrictive | Broaden the percentile from Top 30% to Top 50%, or extend the lookback window from 30 to 90 days |
| Signal size too small  | Check the signal size in the Innkeepr dashboard                                                  |
| Signal not applied     | Verify that the Seed Signal is linked to the correct asset groups                                |
| Insufficient time      | Some configurations need 60 days or more — wait before drawing conclusions                       |

<Info>
  Learn more

  * [Audience Signals](/guides/signals-and-activation/audience-signals)
  * [Set Up Seed Signals on Meta](/guides/how-to-guides/meta)
  * [Set Up Exclusion Signals on Meta](/guides/how-to-guides/meta-exclusions)
  * [Set Up Conversion Signals on Google](/guides/how-to-guides/google-ads-conversions)
  * [Spec: Identify](/connections/spec-identify)
</Info>