Google eliminated all attribution models except two: Last-Click and Data-Driven Attribution (DDA). Having analyzed attribution data across 60+ service accounts, the answer to "which one should I use?" is more nuanced than Google suggests. DDA can deliver 6-30% more conversions and 20-30% cost reduction — but only if your account qualifies. For most service businesses, it doesn't.
Attribution directly controls where Smart Bidding sends your budget. Choose wrong, and you optimize toward the wrong outcomes.
What Does an Attribution Model Actually Do?
When a customer clicks your ads multiple times before converting, which click gets credit? Consider a typical journey: Day 1 searches "best plumbers near me," Day 5 searches "plumber reviews [city]," Day 8 searches "[company name]" and converts. Three clicks. One conversion.
The answer directly controls Smart Bidding — if only the last click gets credit, budget flows to branded campaigns. If credit is distributed, Smart Bidding recognizes the value of discovery keywords. The wrong model can systematically starve your top-of-funnel keywords of budget.
The Two Remaining Models
Google eliminated Linear, Time Decay, Position-Based, and First-Click in September 2023. Two options remain.
Last-Click Attribution
100% credit goes to the final ad click. In the plumber example, the branded search on Day 8 gets all credit; the discovery clicks get zero.
Works well when: Data volume is low, conversion paths are single-touch, or you're in the first 3-6 months of a new account.
Fails when: Multi-touch journeys are common — it over-credits branded keywords and starves awareness campaigns.
Data-Driven Attribution (DDA)
Google's ML distributes credit based on each touchpoint's incremental contribution. In the plumber example, DDA might assign 40% to the initial discovery search, 25% to the review search, and 35% to the branded click.
Works well when: You have sufficient data volume, multi-touch paths, and professional/high-consideration services.
Fails when: Data is insufficient (see thresholds below) or iOS privacy blocks intermediate touchpoints.
Does My Account Have Enough Data for DDA?
DDA is the default for new conversion actions. But "default" doesn't mean "right for every account."
Minimum Thresholds
| Requirement | Threshold | Time Period |
|---|---|---|
| Ad interactions (clicks) | ~3,000 | 30 days |
| Conversions | ~300 | 30 days |
If your account falls below these thresholds, DDA does not have enough data to model attribution accurately. It reverts to a simplified model internally, which may be less accurate than simple Last-Click.
Threshold Reality Check for Service Businesses
Here are the numbers most people don't want to see:
| Monthly Spend | Average CPC | Monthly Clicks | Monthly Conversions (7.5% CVR) | DDA Qualified? |
|---|---|---|---|---|
| $1,000 | $5.26 | ~190 | ~14 | No |
| $2,500 | $5.26 | ~475 | ~36 | No |
| $5,000 | $5.26 | ~950 | ~71 | No |
| $10,000 | $5.26 | ~1,900 | ~143 | No |
| $15,000 | $5.26 | ~2,850 | ~214 | Borderline |
| $20,000 | $5.26 | ~3,800 | ~285 | Close |
| $25,000 | $5.26 | ~4,750 | ~356 | Yes |
The uncomfortable truth: Most service businesses spending under $20,000/month on Google Ads do not generate enough data for DDA to work properly.
Using the overall average CPC of $5.26 and average conversion rate of 7.52%, you need roughly $20,000-$25,000/month in spend to reliably hit DDA thresholds.
Higher-CPC industries (legal at $8.58, dental at $7.85) need even more spend. Lower-CPC industries might qualify at lower spend levels.
How to Check If Your Account Qualifies
- Go to Google Ads > Goals > Conversions
- Click on your conversion action
- Under Attribution model, check if DDA is available
- If Google shows a warning about insufficient data, your account does not qualify
Alternatively, check your Conversion Paths report:
- Tools & Settings > Attribution > Conversion Paths
- If most paths show only 1 touchpoint, DDA provides minimal benefit anyway
How Much Better Is DDA When It Works?
For qualifying accounts: 6-30% more conversions and 20-30% cost-per-conversion reduction (multiple studies).
Why: The Budget Reallocation Effect
Under Last-Click, a law firm's branded campaigns might show 30 conversions and get 33% of budget. Under DDA, those same branded campaigns show 18 (partial credit), while "best lawyer near me" jumps from 8 to 15 conversions as it gets discovery credit. Smart Bidding then redistributes budget from branded to non-branded campaigns that actually initiate customer journeys.
Result: More total conversions at lower cost because budget flows to truly valuable keywords instead of just last-touch branded terms.
What Attribution Model Should I Use?
Use Last-Click when: Monthly spend under $15K, fewer than 100 monthly conversions, single-interaction paths (emergency services), or new accounts (first 3-6 months).
Use DDA when: Monthly spend $20K+, 300+ monthly conversions, multi-touch paths (professional services, high-value home improvement), running branded + non-branded campaigns with Smart Bidding.
The Transition Path
Months 1-6: Start with Last-Click. Build data, monitor conversion path lengths. Month 6: Evaluate — do you meet 3,000 interactions / 300 conversions thresholds? Are paths multi-touch? When ready: Switch DDA on your primary conversion action. Expect 7-14 day learning phase. Compare 30-day pre/post metrics. Ongoing: Check Model Comparison report monthly — if DDA shows more conversions for non-branded campaigns, it is working.
How Does Attribution Change Smart Bidding Behavior?
Attribution isn't just reporting — it fundamentally changes how Smart Bidding allocates budget. Under Last-Click, Smart Bidding sees branded terms converting best and concentrates budget there. Under DDA, non-branded discovery keywords get partial credit, so Smart Bidding distributes budget across the full funnel.
If you switch from Last-Click to DDA, expect: branded conversions to decrease (credit redistributed), non-branded to increase, a 7-14 day learning phase, and net total conversions to increase (typically 6-30%).
Warning: Switching when your account doesn't meet DDA thresholds causes the opposite — Smart Bidding gets confused by unreliable signals and performance degrades.
Is Last-Click Really That Bad?
For most service businesses? No. It's simple, deterministic, and works with any data volume. Across 60+ accounts, the businesses that suffered weren't using the "wrong" model — they were using DDA without enough data to support it.
Where Last-Click genuinely fails is multi-touch journeys. A law firm client was pouring budget into branded campaigns because Last-Click made them look like conversion machines. Switching to DDA revealed non-branded discovery campaigns were initiating 40% of those journeys.
Match your model to your data volume and journey complexity. Don't let anyone tell you one model is universally better.
iOS Privacy and Cross-Device Challenges
iOS Data Loss
iOS Safari causes an 18-32% reduction in observable conversion data. DDA is more affected than Last-Click because it needs to observe full multi-touch paths.
Mitigation: Enhanced Conversions (low effort, high effectiveness) + server-side tracking (13-27% accuracy improvement, medium-high effort) recovers most data loss. For $5K+/month accounts, this combination is essential.
Related reading: Conversion Tracking for Service Businesses covers the complete setup.
Cross-Device Journeys
Mobile starts 60-70% of service business conversion paths, but desktop/phone completes 40-50% of conversions. Google connects these via signed-in user data, but coverage is incomplete. DDA handles cross-device better than Last-Click by assigning partial credit to the mobile click even when conversion happens on desktop.
View cross-device data at Tools & Settings > Attribution > Cross-Device Activity.
How Do I Compare Models Before Switching?
Tools & Settings > Attribution > Model Comparison shows your conversions attributed under both models side by side.
How to interpret: Campaigns gaining credit under DDA are contributing more value than Last-Click reveals. If DDA shows 20%+ fewer branded conversions, those branded campaigns are over-credited under Last-Click — consider switching. If numbers are nearly identical, most conversions are single-touch and Last-Click is fine. If DDA shows fewer total conversions, audit your tracking before switching.
Which Model Fits My Type of Business?
| Business Type | Recommended Model | Typical Path Length | Rationale |
|---|---|---|---|
| Emergency Services (plumbing, HVAC, locksmith) | Last-Click | 1 interaction (90%+) | Single-touch — search, click, call |
| Professional Services (law, accounting, financial) | DDA (if thresholds met) | 3-7 interactions | Extensive research over weeks |
| Home Improvement (remodeling, roofing, landscaping) | DDA (if thresholds met) | 2-5 interactions | Multiple provider comparisons |
| Health & Wellness (dentists, chiropractors, med spas) | Last-Click (most); DDA for large practices | 1-2 routine, 3-5 elective | Simple paths except elective procedures |
| Recurring Services (cleaning, pest control, lawn care) | Last-Click | 1-2 interactions | Low-risk, low-commitment first purchase |
The Decision Framework
- Check data volume: 3,000+ clicks and 300+ conversions/month? If no, use Last-Click.
- Check path length: Go to Attribution > Conversion Paths. 30%+ multi-touch? If no, Last-Click is fine.
- Check model comparison: Does DDA redistribute 20%+ credit to non-branded? If yes, switch.
- Monitor: Did total conversions increase within 30 days? If not, revert to Last-Click.
The Most Common Attribution Mistakes
| Mistake | Fix |
|---|---|
| DDA with insufficient data — model unreliable, Smart Bidding confused | Check thresholds: 3K interactions + 300 conversions/30 days. Under? Use Last-Click. |
| Ignoring attribution entirely — accepting Google's default blindly | Actively choose. Review Model Comparison quarterly. |
| Switching without a baseline — can't tell if performance degraded | Record CPA, volume, ROAS for 30 days before switching. Compare at 14, 30, 60 days. |
| Over-crediting branded — Last-Click makes branded look like conversion machines | Check Model Comparison — if branded loses 20%+ credit under DDA, non-branded is undervalued. |
| Ignoring iOS data loss — DDA distorted by 18-32% invisible conversions | Implement Enhanced Conversions + server-side tracking. |
| Set-and-forget — model that was right 6 months ago may not be today | Quarterly review of Model Comparison, thresholds, and path patterns. |
Monthly Reports to Monitor
| Report | Location | Key Action |
|---|---|---|
| Conversion Paths | Attribution > Conversion Paths | If path length grows, adjust conversion windows |
| Model Comparison | Attribution > Model Comparison | If non-branded gains 20%+ credit under DDA, those campaigns are undervalued |
| Time Lag | Conversion Paths > Time Lag | If 40%+ convert after 7 days, set 30+ day conversion window |
| Assisted Conversions | GA4 > Advertising > Attribution | Keywords with high assists but low last-click = hidden value |
Monthly dashboard metrics: Average path length, single-touch vs multi-touch percentage, DDA vs Last-Click variance by campaign, median time to conversion, cross-device conversion percentage, iOS vs Android conversion rates.
Attribution in the Bigger Measurement Stack
Attribution can only credit what it can see. It works best when combined with: complete conversion tracking (especially phone calls — see Conversion Tracking for Service Businesses), offline conversion data (signed contracts uploaded back to Google Ads), server-side tracking (13-27% accuracy improvement), and correct Primary/Secondary conversion configuration (attribution only applies to Primary actions).
Last-Click vs DDA: Full Comparison
| Factor | Last-Click | Data-Driven Attribution |
|---|---|---|
| How it works | 100% credit to final click | ML distributes credit across touchpoints |
| Data requirement | None | ~3,000 interactions + 300 conversions/month |
| Minimum spend (approx.) | Any | $20-25K/month |
| Accuracy for single-touch | Excellent | No advantage |
| Accuracy for multi-touch | Poor (ignores earlier touchpoints) | Good (when data sufficient) |
| Smart Bidding impact | Over-credits branded/bottom-funnel | Distributes credit across funnel |
| Performance improvement | Baseline | 6-30% more conversions, 20-30% cost reduction |
| Transparency | Full — deterministic | Black box — cannot see credit distribution |
| iOS privacy sensitivity | Lower | Higher (needs full path visibility) |
| Emergency services | Ideal (single-touch journeys) | No benefit |
| Professional services | Undervalues discovery | Properly credits research phase |
| Learning phase after switch | N/A | 7-14 days |
The Bottom Line
Most service businesses spending under $20K/month should use Last-Click and not think twice about it. If you're above that threshold with multi-touch customer journeys, DDA is worth testing — but document your baseline first and monitor for 30 days.
Regardless of which model you choose, the bigger wins come from feeding it complete data: track phone calls, upload offline conversions, implement Enhanced Conversions, and add server-side tracking for accounts spending $5K+/month. A perfect attribution model running on incomplete data is still making bad decisions.