Google Smart Bidding optimises bids in every auction using signals available at the moment of the impression. The quality of the conversion value you pass determines the quality of what the algorithm learns to find. When that value is first-order checkout AOV, the algorithm learns to find efficient first-order spenders. When that value is predicted 12-month or 24-month customer lifetime value, it learns to find the customers who are still buying from you 18 months later.

This gap is the mechanism behind most DTC Smart Bidding underperformance. The value-based bidding framework for Google Search is well-built and capable of significant performance improvements: but only when the conversion value it receives accurately represents customer worth to the business. Passing AOV into a sophisticated ML system is like giving it a precise instrument calibrated to the wrong objective.

This guide covers the three technical paths for passing LTV signals into Google Smart Bidding, the correct setup for Conversion Value Rules, Customer Match, and Offline Conversion Import, the learning period mechanics that determine how quickly the system recalibrates, and the signal quality requirements that determine whether the recalibration is meaningful. All technical steps link directly to Google's official documentation for each component.

In This Article

Why Signal Quality Determines Smart Bidding Outcome

Smart Bidding uses hundreds of signals at auction time to predict conversion probability and value, then sets bids accordingly. The signals include user device, location, time of day, search query, browsing behaviour, and audience memberships. But all of these serve one purpose: predicting what value this click will produce based on historical data about what similar clicks produced. The conversion value you pass is the ground truth the algorithm trains on. Every other signal is in service of predicting that value.

If checkout AOV is the conversion value, the algorithm builds a model of which user profiles produce high first-order purchases. If 12-month predicted LTV is the conversion value, it builds a model of which user profiles produce customers who compound. Both models optimise accurately toward their respective targets. The business impact is entirely different. A supplement brand whose highest-LTV customers have a modest first order but a 24-month replenishment pattern will systematically be under-bid for by AOV-based Smart Bidding. The algorithm is not malfunctioning: it is functioning exactly as instructed, toward the wrong objective.

Google Smart Bidding is a prediction engine. It predicts what conversion value a click will produce, then bids based on that prediction. The prediction quality is bounded by the quality of the historical conversion values it trained on. Garbage in, garbage out: but with very sophisticated garbage processing. The fix is improving the signal, not the campaign structure.

The Three Paths to LTV Signals in Google

Google provides three mechanisms for passing LTV-informed values into Smart Bidding. They differ in technical complexity, signal fidelity, and how dynamically they update per customer. Google's value-based bidding documentation for Search outlines the requirements for each. The diagram below shows all three paths and where each connects to the Smart Bidding system.

Three paths to pass LTV signals into Google Smart Bidding Choose the path based on your technical infrastructure, data freshness requirements, and whether you have pLTV predictions ready PATH 1 Conversion Value Rules MECHANISM Multipliers on conversion events based on audience, device, geo EFFORT Low. Google Ads UI only. No dev work required. LIMITATION Segment-level, not customer-level. Approximate, not ML-predicted. PATH 2 Customer Match + OCI MECHANISM Upload segmented customer lists stratified by pLTV tier via OCI EFFORT Medium. CRM export + regular uploads. Basic data pipeline. LIMITATION Batch update latency. Not real-time per-customer scoring at auction. PATH 3 Real-Time pLTV via OCI MECHANISM ML model scores each customer at purchase; OCI passes pLTV as conversion value within hours EFFORT High. Requires ML pipeline, real-time inference, OCI setup. OUTCOME Best signal quality. 20-30% CAC reduction within 60 days. Google Smart Bidding Receives conversion value signal per auction. Optimises bids to find users predicted to generate the highest value. tROAS and Maximize Conversion Value both use this signal. Better signal quality = better customer quality acquired. adzeta.io

Path 1: Conversion Value Rules (Easiest)

Conversion Value Rules are the lowest-effort entry point for LTV-informed bidding. You create rules in Google Ads that multiply or add to conversion values based on audience segments, device type, or geographic location. A Customer Match list of known high-LTV customers can receive a 2x or 3x multiplier on their checkout conversion value, telling Smart Bidding that purchases by this segment are worth more to the business than their transaction amount suggests.

Setting up Conversion Value Rules requires no developer work and no external data pipeline. The setup is done entirely within Google Ads under Goals. The core limitation is that rules operate at segment level, not individual customer level. A Customer Match list labelled "high-LTV customers" applies a uniform multiplier to all members regardless of individual predicted value. It is a meaningful improvement over pure AOV signals, but it is an approximation rather than a per-customer prediction.

According to Google's documentation on the impact of Conversion Value Rules on Smart Bidding, rules are applied in real time at auction time when tROAS or Maximize Conversion Value is active. The system already accounts for signals like device and location in its baseline predictions: adding explicit value rules on top of these means the algorithm has both its internal model and your explicit business-value weighting to work from simultaneously.

When Conversion Value Rules work best

Path 1 is the right starting point for accounts that have not yet built LTV prediction infrastructure. It produces measurable signal improvement within the first learning cycle and requires no data engineering. It works particularly well when you have clear high-LTV audience segments already identified in your CRM: known subscribers, regimen purchasers, or multi-SKU buyers: that can be uploaded as Customer Match lists and used as the rule condition.

Path 2: Customer Match With LTV Tiers

Customer Match allows you to upload first-party customer data and create audience segments that Smart Bidding uses as signals. For LTV-informed bidding, the approach is to stratify your customer base into LTV tiers based on historical cohort data, upload each tier as a separate Customer Match list, and apply differentiated Conversion Value Rules or bid modifiers to each tier.

A practical implementation for a supplement or skincare brand: Tier 1 (predicted top quartile LTV customers, based on 12-month cohort data) receives a 3x value multiplier. Tier 2 (mid-LTV) receives 1.5x. Tier 3 (one-time buyers from historical data) receives no adjustment or a downward multiplier. Customer Match best practices from Google confirm that when these lists are used with Smart Bidding, the algorithm automatically incorporates them as signals for bid optimisation.

Your guide to Customer Match notes that Customer Match lists must contain at least 1,000 matched users to be active for targeting and bidding. They require regular refreshes (every 30 to 90 days is recommended) to remain accurate as customer LTV scores update with new purchase behaviour. Lists go stale after 540 days without a refresh. For brands using automated CRM exports, this refresh can be handled via a scheduled job rather than manual uploads.

Path 3: Offline Conversion Import With pLTV Values

Offline Conversion Import (OCI) is the highest-fidelity path for passing LTV signals into Google Smart Bidding. Instead of applying segment-level multipliers, OCI sends a specific predicted value for each individual conversion event, matched back to the original ad click via the Google Click ID (GCLID). This means every purchase receives its own unique pLTV score as the conversion value: the algorithm trains on individual customer-level predictions, not segment averages.

The mechanics: when a customer clicks an ad, the GCLID is captured and stored alongside their profile in the CRM or data warehouse. When they purchase, an ML model scores their predicted 12-month LTV based on early behavioural signals. That score is paired with the original GCLID and uploaded to Google via OCI, where it replaces or supplements the standard transaction value as the conversion signal for Smart Bidding.

The OCI FAQ documentation is explicit about the requirements for Smart Bidding to learn effectively from OCI data: uploads must happen at least daily, the conversion value field must contain varied (non-uniform) numbers, and the upload must occur within 90 days of the original click or the GCLID expires. The pLTV bidding on Google Ads guide covers the full technical implementation including GCLID capture, ML model requirements, and OCI file format.

The OCI Data Flow: From Purchase to Bid Adjustment

The diagram below shows the exact path a pLTV value takes from customer click to Smart Bidding bid adjustment, including where each technical step happens and what data flows between them.

Offline Conversion Import: how pLTV values reach Google OCI is the only channel for passing customer-level predicted LTV into Google Smart Bidding. The upload must happen within 90 days of the original click. 1 Customer Clicks Ad GCLID captured + stored in CRM 2 Customer Purchases Order + GCLID linked in CRM 3 ML Model Scores pLTV 12-mo revenue predicted in hours 4 OCI Upload Sends Value GCLID + pLTV value sent daily to Google 5 Smart Bidding Re-Optimises Bids towards high pLTV profiles OCI requirements for Smart Bidding to learn from pLTV values Upload within 90 days of click GCLID expires after 90 days Daily uploads strongly preferred Delays slow Smart Bidding learning Values must be varied, not uniform Identical values = no pLTV signal OCI CSV format (minimum required columns) Google Click ID (GCLID) | Conversion Name | Conversion Time | Conversion Value | Conversion Currency AbCdEfGhIj12345 | pLTV Purchase | 2026-04-14 10:30:00 | 247.50 | USD The value column (247.50) is your pLTV prediction: not the transaction amount. Smart Bidding uses this to learn which profiles produce high-value customers.

The most common OCI configuration failure is GCLID capture. If GCLID is not stored at the point of the ad click, the pLTV value has no identifier to attach to. The GCLID must be captured at landing and stored in the CRM or order management system before any downstream matching is possible. A secondary failure mode is conversion delay: Google's documentation states that Smart Bidding performance is impacted when OCI uploads take more than 7 days post-click. For pLTV signals, which require the ML model to score the customer, the target is to complete the scoring and upload within 24 to 48 hours of purchase.

Smart Bidding Learning Period: What to Expect

Every time the conversion signal changes: whether by switching from AOV to pLTV values or by making a significant change to Conversion Value Rules: Smart Bidding enters a learning period. According to Google's documentation, the learning period requires up to 50 conversion events or 3 full conversion cycles for the algorithm to calibrate to the new signal. Campaigns with lower conversion volume take longer.

During the learning period, performance fluctuations are expected and documented. The algorithm is recalibrating its internal model of which user profiles are associated with high-value conversions. ROAS may dip, CPA may increase, and conversion volume may shift. These are symptoms of recalibration, not malfunction. The correct response is to maintain the new signal consistently and evaluate performance only after the learning period completes: typically 4 to 6 weeks after the signal change.

The most damaging behaviour during the learning period is reverting the signal change in response to short-term performance dips. Each reversion resets the learning period. Brands that switch between AOV and LTV signals multiple times during a quarter never complete a single learning cycle: the algorithm perpetually re-learns without accumulating enough signal history to optimise effectively. Set the signal, define a review date at week 6, and do not change the conversion action or value structure during that window. Google's VBB best practices documentation reinforces this explicitly.

LEARNING PERIOD
50
Conversion events or 3 full conversion cycles needed for Smart Bidding to calibrate to a new LTV signal. Accounts below this threshold need longer: up to several months for very low-volume accounts.
Source: Google Ads 2026 Help Centre

The Full Setup Sequence

  1. Verify GCLID capture is active on your landing pages

    GCLID is the identifier that links a Google Ads click to downstream conversion data. Confirm your landing page URL parameters include gclid= and that it is being stored in your CRM or order management system against each lead or order record. Without reliable GCLID capture, OCI cannot function and Customer Match refreshes will not reflect recent purchase behaviour. Test by clicking a Google Ad, checking the landing page URL for gclid=, and confirming it appears in the CRM record. The offline conversion imports guide covers verification steps.

  2. Set up dynamic conversion values in conversion tracking

    Navigate to Goals in Google Ads, select Conversions, and confirm your Purchase conversion action uses transaction-specific (dynamic) values rather than a static amount. Every purchase event should fire with the actual transaction value in the value field. Google requires at least two unique conversion values for Maximize Conversion Value to differentiate between high and low-value purchases. If all conversions show the same value, the tracking is not passing dynamic data.

  3. Create Customer Match lists segmented by LTV tier

    Export your customer database from your CRM segmented into LTV tiers based on 12-month cohort data. Minimum three tiers: top quartile (high predicted LTV), mid-range, and one-time buyers. Upload each as a separate Customer Match list via Audience Manager in Google Ads. Each list needs at least 1,000 matched members to be active. Set a calendar reminder to refresh these lists every 60 days. The match rate improves with hashed email, phone number, first name, last name, and ZIP code all included.

  4. Apply Conversion Value Rules using Customer Match tiers

    Under Goals in Google Ads, navigate to Conversion Value Rules and create a rule for your high-LTV Customer Match list. Apply a multiplier of 2x to 3x on the conversion value for members of this list. This tells Smart Bidding that a purchase by a customer in your high-LTV segment is worth 2 to 3 times more than the checkout amount suggests. This is Path 1: it can be running within hours of Customer Match lists being active.

  5. Build and connect the OCI pipeline for pLTV values

    This is Path 3 and requires data engineering. The pipeline: (1) capture GCLID at ad click and store in CRM, (2) on purchase, pass the customer profile to your pLTV ML model, (3) receive the scored 12-month prediction, (4) format as an OCI CSV with GCLID, conversion name, timestamp, and the pLTV value, (5) upload daily via Google Ads Data Manager or API. OCI FAQs cover the file format, upload frequency requirements, and how delays in uploading impact Smart Bidding learning. The AdZeta ValueBid™ platform automates steps 2 through 5.

  6. Switch bid strategy to Maximize Conversion Value or tROAS

    Once value data is flowing, switch from Target CPA or Maximize Conversions to Maximize Conversion Value (no ROAS target initially). Allow 4 to 6 weeks for the algorithm to recalibrate. Do not change the bid strategy, conversion action, or value structure during this period. Set your first review date at week 6. Evaluate using revenue LTV:CAC per cohort, not same-week ROAS, which will fluctuate during learning.

AOV as signal
2.1x LTV:CAC
pLTV via OCI
5.4x LTV:CAC
Median LTV:CAC improvement in AdZeta accounts within 90 days of switching to pLTV-based OCI signals

Key Takeaways

  • Google Smart Bidding trains on the conversion value you provide. AOV-based signals produce an algorithm optimised for first-order spenders. pLTV signals produce an algorithm optimised for customers who compound. The campaign structure is the same. The outcome is different.
  • There are three paths: Conversion Value Rules (no code, segment-level approximation), Customer Match with LTV tiers (medium effort, regular refreshes), and OCI with per-customer pLTV values (highest fidelity, requires ML pipeline and GCLID capture).
  • OCI requires GCLID capture at the landing page and storage in your CRM before anything else. Without reliable GCLID data, pLTV values have no identifier to match back to the original click.
  • The Smart Bidding learning period takes up to 50 conversion events or 3 conversion cycles. Do not change the signal, bid strategy, or conversion action during the learning window. Evaluate at week 6, not week 1.
  • Smart Bidding Exploration (available on tROAS campaigns) allows the algorithm to test new traffic segments at a flexible ROAS, producing an 18% increase in unique search query categories with conversions on average. Enable it once the base signal is stable.
  • The AdZeta ValueBid™ platform automates the OCI pipeline: scoring pLTV at the moment of purchase and uploading values to Google daily. Most accounts see LTV:CAC improvement within 90 days of switching from AOV to pLTV-based OCI signals.

Further Reading

pLTV Bidding on Google Ads: Step-by-Step Guide for DTC Brands: the complete technical implementation guide for building the ML model, OCI pipeline, and Customer Match architecture for pLTV-based bidding.

What Is Value-Based Bidding? A Complete Guide for DTC Brands: how value-based bidding works across Google and Meta, what conversion values to pass, and the two core Google bidding strategies.

First-Party Data Activation: How to Use Your Customer Data Across Google, Meta, and Programmatic: the broader first-party data infrastructure required for reliable LTV signal delivery across all ad platforms.