GA4 ships with seven attribution models for crediting conversions across the touchpoints that led to them. The default is ‘Data-driven attribution’ (DDA), which uses machine learning to weight each touchpoint by its contribution. DDA is good in theory; it’s unreliable for small SMB accounts where the model doesn’t have enough data to be confident. Here’s when each model is the right reference for SMB decisions.
The seven models
- Data-driven (DDA) — ML-weighted credit across all touchpoints. Default.
- Last click — 100% credit to the last paid/organic touchpoint before conversion. The pre-2023 GA default.
- First click — 100% credit to the first touchpoint. Useful for measuring awareness channels.
- Linear — equal credit to every touchpoint.
- Time decay — recent touchpoints get more credit; older touchpoints get less.
- Position-based — 40% to first touch, 40% to last touch, 20% to everything in between.
- Last non-direct click — last touchpoint that isn’t Direct traffic. The default for older GA versions.
Why DDA is unreliable for SMBs
DDA needs ~300 conversions per month to produce stable attribution weights. Most SMBs get 30-100 conversions per month. At that volume, DDA is using sparse data — its weights swing widely month to month based on small samples. You can’t make confident channel-investment decisions from numbers that move 20% on noise.
When to use last-click instead
When you have under 300 conversions/month, last-click is more honest. It’s deterministic — the touchpoint right before the conversion gets credit. It under-credits awareness channels but it’s not noisy. We default to last-click reporting for SMB clients while watching DDA secondarily for the channels DDA gives meaningful credit to.
When first-click matters
When you’re investing in upper-funnel channels (display ads, video ads, brand campaigns) and want to know if they contribute to eventual conversions. First-click tells you ‘did this channel ever bring this user in.’ Different from last-click ‘did this channel close the conversion.’ Both are real signals.
When time-decay matters
When your sales cycle is short and recent touches matter more than distant ones. A 2-week sales cycle SMB benefits from time-decay; a 6-month enterprise SaaS sale benefits less because the early touches still matter.
Reading the same conversion across models
GA4 → Advertising → Attribution → Model comparison. Pick two models, see how credit shifts between channels. If a channel earns substantially more credit under first-click than last-click, it’s an awareness channel. If a channel earns more under last-click, it’s a closing channel. Both have value; the bias of which one you measure changes your decisions.
The cross-channel reality
Most SMB conversions involve 2-4 touchpoints across 1-2 channels. A buyer searches your name, lands on the homepage, leaves, sees a retargeting ad later, clicks back, fills out the form. Last-click credits the retargeting ad. First-click credits the organic search. Both are honest readings of the same buyer journey.
Smart Bidding and attribution
If you’re running Google Ads with Smart Bidding (Maximize Conversions, Target CPA, Target ROAS), the bidding algorithm uses Google’s own attribution. That attribution is closer to DDA but specific to ads. Your reporting attribution doesn’t affect bidding; bidding uses its own model. Keep them separate in your head.
The 90-day Lookback window
GA4’s default lookback is 90 days for paid and 30 days for organic. Touchpoints older than the lookback don’t get credit. Long-cycle B2B sales need a longer lookback (GA4 caps at 90 days for paid, which can miss real attribution). Workaround: use BigQuery export and run attribution off-platform when sales cycles exceed 90 days.
Branded vs non-branded segmentation in attribution
When your largest channel is branded search (people Googling your business name), it dominates last-click attribution. Always segment: ‘what brought the buyer in originally’ usually matters more than ‘which channel they were on when they typed your name into Google.’ Use first-click attribution on the branded-vs-non-branded split to see the real story.
Reporting attribution to SMB clients
We report two numbers per channel: last-click conversions (what closed) and first-click conversions (what started). Together they tell the real story. A single ‘conversions by channel’ number under any model is over-simplified.
The bigger reality check
Attribution at SMB volumes is always partly wrong. The data is sparse, the models are approximations, the user’s actual journey crosses devices and offline touchpoints you can’t see. Make decisions on directional signals, not precise weights. If a channel is consistently producing conversions under multiple attribution models, it’s working. If it only earns credit under one model, look harder before you scale spend.
How this lands across FH client work
Across the FH client book, we use last-click + first-click as primary, watch DDA as a secondary check, and run lead-score-weighted attribution for the clients with enough volume to make it stable. The combination has been the reliable foundation for ad-budget rebalancing on multi-service clients. If your attribution is whatever GA4 shows by default and you’re not sure what it means, book a consultation — the reporting refresh is a one-week engagement that ships a setup you can actually trust.
