Google is removing four attribution models from Google Ads in two phases during July and September 2026. First-click, linear, time-decay, and position-based attribution are being deprecated — leaving only data-driven attribution (DDA) and last-click as the two models available for any conversion action. For most B2B SaaS advertisers this is a background event they can navigate with a single afternoon of preparation. For accounts that still use the deprecated models, or for teams that do not document their baseline before the migration, the September auto-migration will produce reporting changes that look like performance shifts but are entirely caused by measurement changes.
The previous round of attribution model removals — which consolidated Google Ads from six models to two — happened in earlier years and is covered in our general guide to Google Ads attribution models. The July 2026 removal is a second phase targeting any remaining use of the deprecated models for existing conversions. This post is specifically about what to do in the next 60 days: audit, document, decide, and migrate before Google does it for you.
The exact removal timeline
The deprecation happens in two distinct phases. Phase one began in mid-July 2026: from that date, any new conversion action created in Google Ads can only be configured with data-driven attribution or last-click. The four deprecated models — first-click, linear, time-decay, position-based — are no longer available as options for conversions created after mid-July. This phase affects new account setup and any new conversion actions you add to existing accounts, but does not change existing conversion configurations.
Phase two completes the removal in September 2026: all existing conversions still using the four deprecated models will be automatically migrated to data-driven attribution, regardless of whether you have taken any action. The September deadline is the one that requires active preparation: if your account has conversions on deprecated models, and you do nothing, Google will switch them to DDA. Whether DDA is the right choice for those conversion actions — or whether last-click would serve your account better — is a decision worth making deliberately rather than having made for you.
The two models that remain: DDA vs last-click
Data-driven attribution uses Google's machine learning to analyze all the clicks in a converting path and assign credit proportionally based on each touchpoint's estimated contribution to the conversion. It considers the order of ad exposures, the time between exposures, the ad format, and historical patterns from converting and non-converting paths. When it works well, DDA provides the most accurate credit distribution available in the Google Ads ecosystem — more credit goes to earlier touches that genuinely influence buyers, and less credit goes to last-touch that often only serves buyers who were already decided.
Last-click attribution gives 100% of conversion credit to the last ad click before the conversion. It is simple, transparent, and easy to interpret: every conversion is attributed to exactly one campaign, one keyword, one ad. For accounts with low conversion volume, where DDA does not have enough data to make reliable credit assignments, last-click often produces more stable and actionable reporting than a DDA model that is starved for signal. For accounts with higher volume, last-click systematically overvalues bottom-of-funnel campaigns and undervalues awareness and consideration touchpoints — a distortion that affects budget allocation when campaign performance is evaluated based on attributed conversion counts.
Why this change is more complex for B2B SaaS
The standard guidance on attribution model selection assumes conversion volumes that most B2B SaaS accounts do not have. Google's recommendation that DDA requires at least 3,000 ad interactions and 300 conversions per month per conversion action to produce reliable credit assignments is based on statistical requirements for its machine learning model. A B2B SaaS company generating 25 demo requests per month from Google Ads — a respectable volume for a mid-market product — falls well short of the threshold where DDA produces stable credit distributions. The result, as Boostify noted in their July 2026 analysis, is that "a B2B SMB generating 20 or 30 leads a month does not give the model enough signal, and the result can be an erratic credit distribution that shifts month to month."
The complexity compounds for accounts using offline conversion imports from a CRM. If your primary conversion action is a downstream event — SQL created, opportunity opened, deal closed — the conversion volume is typically even lower than the form-fill count. DDA on a conversion action that fires 15 times per month is not meaningfully different from DDA on a conversion action with no data: the model defaults to a near-last-click distribution when it cannot identify meaningful patterns. The practical question for B2B SaaS accounts is therefore not which model is theoretically superior, but which model produces a credit distribution stable enough to make reliable optimization decisions on — and that answer varies by account volume. Our guide to offline conversion imports from a CRM covers how downstream conversion events affect data volume and model selection.
Three steps to take before the September deadline
The first step is a conversion action audit. In Google Ads, navigate to Goals → Conversions and add the attribution model column to the view. List every conversion action in your account along with its current attribution model. Any action using first-click, linear, time-decay, or position-based needs to be either manually migrated before September or left to Google's automatic migration. Pay particular attention to micro-conversion actions — page views, video watches, chatbot engagements — that may have been configured years ago and forgotten: these will also be auto-migrated and can produce unexpected reporting changes if not documented.
The second step is to export model comparison data while it is still available. Google Ads provides a model comparison report that shows how conversion credit would be distributed differently under different attribution models. Export this report for your primary conversion actions over the last 90 days and save it. Once the deprecated models are removed in September, this comparison data will no longer be available for the period before migration. Having it gives you a reference point for explaining reporting discontinuities to stakeholders, and for auditing whether the auto-migration changed which campaigns appeared to be the top performers.
The third step is to document current performance baselines per campaign under your current attribution model: conversion counts, cost per conversion, conversion rate, and any ROAS or target CPA figures used in your bidding strategy. When the attribution model changes — whether you change it manually or Google does it automatically — these numbers will shift even if nothing about your campaigns or your business changes. Without a pre-migration baseline, it is impossible to distinguish a real performance change from a measurement artifact. This is the single most neglected step in attribution migrations and the one that causes the most confusion in post-migration analysis.
When to choose last-click over data-driven
Choose data-driven attribution when: your account generates at least 300 conversions per month for the conversion action in question; your Smart Bidding strategy is tCPA or tROAS (rather than maximize conversions without a target); and you have stable offline conversion import pipelines that give the DDA model consistent signal. DDA genuinely improves Smart Bidding performance when it has enough data to learn from, because it gives the algorithm a more accurate picture of which query patterns and audience signals are contributing to conversions throughout the path rather than only at the final touch.
Choose last-click attribution when: your primary conversion action generates fewer than 100 conversions per month in the account; you are using offline conversion imports for a downstream event with even lower volume; or you are in an account recovery situation where measurement stability is more important than measurement accuracy. Last-click on a low-volume account gives Smart Bidding a consistent, interpretable signal. DDA on the same account gives the algorithm a noisy, unstable signal that can produce counterproductive bidding decisions — raising bids on campaigns that happened to capture a last-touch conversion in a given month, then reversing course the next month when the distribution shifts. For B2B SaaS accounts with long sales cycles and low conversion volumes, deliberate last-click is often better than accidental DDA. This connects directly to the broader question of conversion tracking strategy for SaaS businesses, where signal quality and volume thresholds determine which measurement approaches are actually viable.
How the model change flows into Smart Bidding
Attribution model changes are not just reporting changes — they directly affect how Smart Bidding allocates budget and sets bids. Smart Bidding learns from the conversion data available in the account, and the attribution model determines which campaigns, ad groups, keywords, and audiences get credit for each conversion. If your account switches from position-based (which gave 40% credit to the first touch and 40% to the last touch) to data-driven (which distributes credit based on estimated influence), the bidding algorithm immediately starts receiving different credit signals. Campaigns that appeared strong under position-based may look weaker under DDA, and vice versa.
The practical implication is that attribution model migrations should be treated as bidding strategy events, not just reporting configuration changes. After migrating to a new attribution model, expect Smart Bidding to need a re-learning period of typically 2-4 weeks before performance stabilizes under the new signal set. During this period, avoid making significant bid strategy changes or campaign structure modifications — isolate the attribution migration as the single variable. If you are using target CPA or target ROAS bidding, watch your actual CPA and ROAS closely in the two weeks following migration and compare them to your pre-migration baseline. Changes larger than 15-20% from baseline in either direction are worth investigating as potential measurement artifacts rather than real performance changes. For more on how bidding strategies interact with conversion data quality, see our guide to Google Ads bidding strategies for B2B SaaS.
What to monitor after the migration
In the first two weeks after an attribution model migration — voluntary or automatic — monitor three specific metrics daily: total conversions per campaign (to catch unexpected volume shifts caused by credit redistribution), cost per conversion per campaign (to identify bidding strategy responses to the new signal), and campaign budget utilization (to catch cases where Smart Bidding has significantly changed how it allocates spend across campaigns with the new credit distribution). Compare each of these metrics to the pre-migration baselines you documented in step three of the audit process above.
The most common post-migration surprise in B2B SaaS accounts is a shift in which campaign appears to have the best cost per SQL. Under last-click or position-based attribution, brand campaigns often dominate — they capture conversions at the final touch, when a buyer who has already been through an awareness and consideration journey searches your brand name and converts. Under data-driven attribution, credit flows back to the earlier campaigns that introduced those buyers to your product. This makes non-brand campaigns appear more valuable and brand campaigns appear less efficient. Neither view is wrong — they reflect different parts of the same conversion path. Understanding this redistribution helps you avoid over-reacting to apparent performance changes in the weeks after migration, and it reinforces the value of measuring conversion windows that match your actual B2B SaaS sales cycle so the attribution model has the full path to work with.