Data-Driven Attribution (DDA) — Google Ads' machine learning attribution model — became the default for new conversions in 2023 and the recommended model for most accounts by 2026. Despite this, many accounts still run on legacy last-click attribution, often because they migrated to Google Ads before 2023 or because someone manually overrode the default.
This guide covers the 2026 attribution landscape: when DDA outperforms last-click, the hidden gotchas (especially the 300 monthly conversion threshold), and a migration playbook for accounts still on last-click.
Attribution model choice directly affects Smart Bidding optimization. Smart Bidding optimizes to whatever attribution model is configured — choose poorly and Smart Bidding optimizes to flawed signal. The difference between DDA and last-click for the same account can be 10-20 % difference in Smart Bidding efficiency.
Attribution models in Google Ads 2026: the current options
Google Ads 2026 supports five attribution models (down from 7 pre-2023):
Position-based attribution (40/20/40 first/middle/last) was deprecated in 2023. Linear and Time Decay still exist but are deprecated and being phased out in favor of DDA for non-eligible accounts.
For 2026, the practical choice is DDA vs Last Click. The other models exist for legacy/specific use cases but aren't recommended starting points.
How Data-Driven Attribution actually works
DDA uses a counterfactual machine learning model to assign fractional credit to each touchpoint in a conversion path. The simplified mechanism:
- Training: Google trains a model on your account's conversion paths over rolling 90 days. Path = sequence of clicks (and possibly impressions) leading to conversion.
- Counterfactual analysis: model estimates "if this click had not happened, would the conversion still have occurred?" The marginal contribution of each click determines its credit share.
- Credit allocation: credit summed across all touchpoints equals 100 % of conversion. Each touchpoint gets fractional credit (e.g. 30 % brand + 50 % generic + 20 % display).
- Continuous retraining: model retrains regularly as new data arrives.
DDA quality depends on:
- Conversion volume (more = better statistical confidence)
- Path diversity (multiple campaign types contribute = better learning)
- Conversion path length (multi-touch paths are where DDA shines)
For single-touch paths (user clicks one ad, converts), DDA gives 100 % to that touch — same as last-click. The lift comes from multi-touch paths where DDA recognizes earlier touches' contribution.
When DDA outperforms last-click (and when it doesn't)
DDA outperforms last-click in:
- B2B SaaS with multi-session sales cycles (multi-touch paths common)
- Considered-purchase ecommerce (€200+ AOV, comparison shopping)
- Multi-channel accounts (search + display + video)
- Branded + non-branded mixed strategies (DDA credits non-branded discovery clicks)
- Accounts above 1,000 monthly conversions (strong DDA training signal)
Last-click outperforms DDA (or ties) in:
- Single-session impulse purchases (low AOV ecommerce, fashion, food delivery)
- Pure brand traffic (single-touch by definition)
- Accounts below 300 conversions/month (DDA silently reverts)
- Strict cross-channel MMM modeling where stable comparison baseline matters
For 2026 default decision: if your account is above the threshold and has any multi-channel or multi-touch behavior, switch to DDA. Below threshold, accept last-click (it's what's running anyway).
Hidden gotchas: minimum thresholds, model decay, cross-channel
Gotcha 1 — 300 monthly conversion threshold. Below this per conversion type, DDA silently reverts to a rules-based model. Reporting still says "DDA" but math is closer to last-click. Solution: consolidate conversion definitions to clear threshold.
Gotcha 2 — Model decay. If account spend drops significantly or path patterns change, DDA model accuracy decays. Re-stabilization takes 30-60 days. Don't make major attribution changes during seasonal volatility.
Gotcha 3 — Cross-channel mismatch. Google Ads DDA only sees Google clicks. Cross-channel attribution (GA4, MMM) sees more touchpoints. Discrepancy is normal but disorienting. Treat Google Ads DDA as Google's internal attribution, GA4 DDA as cross-channel.
Gotcha 4 — View-through credit. DDA credits view-through impressions but weighted lower than clicks. Display/Video campaigns may show lower credit than under "last-click + 30-day view-through" legacy reporting. This is more accurate, not a bug.
Gotcha 5 — Brand campaign credit shifts. DDA often reallocates credit from brand campaigns to upper-funnel campaigns. If you measure "brand search ROAS" in isolation, expect it to drop under DDA. The conversions didn't disappear — they're credited elsewhere now.
Migration playbook: last-click to DDA
The HowTo schema details day-by-day execution. Strategic framing:
Pre-migration audit: document current performance baseline, identify which conversions are DDA-eligible, run DDA vs last-click comparison.
Phased migration: switch top 3-5 conversions first, monitor for 14 days, then migrate remainder. Avoid bulk switching all conversions at once — adds Smart Bidding instability.
Stabilization window: expect 7-14 days of CPA/conversion volume volatility post-switch. Don't revert during this window unless catastrophic (>50 % CPA degradation persisting >7 days).
Post-migration optimization: review campaign-level credit shifts, reallocate budget if appropriate, update reporting dashboards.
Reading DDA reports correctly
DDA reporting in Google Ads → Tools → Conversions → Attribution:
- Campaigns / Ad Groups: see fractional credit allocated. Brand may show 0.4 conversions where last-click showed 1.0 — same conversion, partial credit.
- Top Paths: most common multi-touch paths. Useful for understanding actual customer journey.
- Time Lag: distribution of conversion times after first click. Helps set appropriate conversion windows.
- Path Length: distribution of touchpoints per conversion. >2 average suggests multi-touch reality, validates DDA choice.
Read fractional conversion counts as expected behavior — Smart Bidding handles fractional credit natively.
Cross-channel attribution beyond Google Ads
Google Ads DDA is internal to Google. For cross-channel (Meta, LinkedIn, organic, direct), you need:
GA4 attribution: GA4's own DDA looks across all GA4-tracked sources. Provides cross-channel view but limited to last-click attribution model in standard reporting until 2024.
Marketing Mix Modeling (MMM): top-down statistical models (Meridian, Robyn) for attribution across all paid + organic + offline. Best for accounts >€500k/month total spend.
Multi-touch attribution platforms: Triple Whale, Northbeam, Rockerbox for ecommerce. Cross-channel deduplication. €1-10k/month.
For SteerAds-managed accounts: trust Google Ads DDA for in-platform optimization, supplement with GA4 + MMM for strategic budget allocation.
30-day attribution model evaluation playbook
Week 1 — Audit + comparison. Document baseline, run DDA vs last-click comparison.
Week 2 — Phased switch. Top 3-5 conversions to DDA.
Week 3 — Stabilization monitoring. Smart Bidding rebalances, CPA volatility expected.
Week 4 — Optimization + remainder. Campaign-level budget reallocation, migrate remaining conversions.
For deeper context, see our GA4 setup guide, MMM vs Attribution guide, and incrementality testing guide.
If you'd like AI-driven optimization that fully leverages DDA signal, SteerAds runs a free 14-day audit on Google + Microsoft Ads.
Sources
- support.google.com/google-ads — Attribution models documentation
- support.google.com/google-ads — Data-Driven Attribution documentation
- thinkwithgoogle.com — Case studies and benchmarks
- support.google.com/analytics — GA4 attribution documentation
- optmyzr.com/blog — Industry attribution analyses
FAQ
Is Data-Driven Attribution default in 2026?
Yes — DDA became default for new conversions in 2023 and is automatically applied unless you explicitly switch. Existing conversions on legacy models (last-click, time-decay, position-based, linear) were not auto-migrated; accounts must manually switch. Per Google's 2024-2026 data, ~85 % of accounts above 300 monthly conversions are now on DDA.
What's the conversion volume threshold for DDA to work?
Google's documented minimum: 300 conversions per conversion type in 30 days for full DDA. Below 300, DDA silently falls back to a rules-based model (similar to last-click). Aggregate by conversion type (e.g. 'Purchase' alone needs 300, not all conversions combined). Accounts with low conversion volume should consolidate conversion definitions to clear the 300 threshold.
When does last-click still make sense in 2026?
Three cases: (1) Below DDA threshold (<300 conversions/month per type), where DDA silently reverts anyway, (2) Highly transactional, single-session purchases (low ACV ecommerce, impulse buys) where multi-touch is rare, (3) Cross-channel MMM modeling where you need a stable comparison baseline. For most B2B and considered-purchase ecommerce in 2026, DDA outperforms last-click.
How much performance lift does DDA provide vs last-click?
Per Google's published 2024-2026 case studies and operator-reported data: 5-15 % improvement in conversion volume at same CPA, or 5-15 % lower CPA at same conversion volume. Larger lift in: B2B with multi-session funnels, considered-purchase ecommerce (€200+ AOV), accounts with active video/display alongside search. Smaller lift in: pure search, single-session purchases, branded-heavy accounts.
Does DDA work for view-through conversions?
Partially. DDA credits view-through within attribution windows but weights view less than click. Display + Video impressions still contribute, just with smaller fractional credit. This is a feature, not a bug — view-through has historically over-credited Display/Video in last-click + view-through 30-day setups.
How does DDA interact with Smart Bidding?
Smart Bidding uses DDA-allocated conversion data as its optimization signal. The combination is multiplicative — DDA gives Smart Bidding better signal about which touchpoints drove which conversions, Smart Bidding optimizes bids to maximize those touchpoints. Most accounts see Smart Bidding efficiency improve when DDA replaces last-click.
Can I still see what last-click would have credited?
Yes — in Google Ads → Tools → Conversions → Attribution, you can view 'Compare Models' that shows side-by-side last-click vs DDA. Useful for understanding the redistribution. Don't use this to second-guess DDA — use it to understand the funnel reality.