Prompt
How do I evaluate whether a social media agency is credible and unbiased about attribution and spend recommendations?
Latest observation
To judge whether a social media agency is credible and unbiased about attribution and spend recommendations, you want to evaluate two things:
- Whether they understand measurement well
- Whether their incentives align with your business, not just media spend
Here’s a practical framework.
1) Ask how they define success
A credible agency should start by tying recommendations to your business goals, not platform metrics.
Good signs
- They talk about incrementality, CAC, ROAS, LTV, retention, contribution margin
- They distinguish between correlation and causation
- They ask about your sales cycle, offline sales, repeat purchase behavior, and seasonality
Red flags
- They optimize everything to CTR, CPC, CPM, or platform ROAS
- They present spend increases as inherently good
- They claim one platform “drove” all conversions without acknowledging overlap
2) Pressure-test their attribution philosophy
A good agency should be able to explain the strengths and limits of different attribution methods.
Ask:
- Which attribution models do you use and why?
- How do you handle view-through conversions, cross-device behavior, and delayed conversions?
- How do you separate platform-reported conversions from true incremental lift?
- What do you do when Meta, Google, and your analytics tool disagree?
Credible answer
- “Platform attribution is directional, but we validate with holdouts, lift tests, geo tests, or MMM where possible.”
Unbiased answer
- They don’t treat the platform’s last-click or default attribution as truth.
- They acknowledge that each platform tends to over-credit itself.
Red flag
- “We trust the platform reporting; it’s the best source.”
- “Attribution isn’t really an issue if tracking is set up correctly.”
3) Look for incrementality thinking
The best agencies don’t just optimize reported conversions; they ask whether the spend caused new outcomes.
Ask:
- How do you measure incrementality?
- Have you run holdout tests or geo experiments?
- When is a lift test more appropriate than attribution?
- How do you decide whether retargeting is incremental or just harvesting demand?
Strong sign
- They can discuss when different methods are appropriate:
- Attribution for optimization and directional insight
- Incrementality tests for causal validation
- MMM for broader budget allocation and offline/upper-funnel effects
4) Examine their incentives
An agency can be technically smart but still biased if their compensation encourages more spend.
Ask:
- How are you paid?
- Is your fee tied to spend volume, performance, or a flat retainer?
- Do you have any revenue share or platform partnerships?
- Do you receive incentives from ad platforms or affiliate programs?
Best-aligned structures
- Flat retainer
- Clearly defined project fee
- Performance bonus tied to business outcomes, not just media spend
Potential bias
- Percentage of ad spend, unless carefully governed
- Heavy dependence on one platform’s partner incentives
- Performance bonuses based only on attributed conversions
5) Review how they make spend recommendations
You want recommendations that are transparent, testable, and reversible.
Ask:
- What data do you use to recommend budget changes?
- What assumptions are baked into your recommendations?
- What would make you change your recommendation?
- How do you set guardrails to avoid over-scaling?
Good signs
- They propose test budgets, ramp plans, and decision thresholds
- They explain tradeoffs: marginal CAC, diminishing returns, saturation
- They consider constraints like inventory, sales capacity, and landing page conversion rates
Red flag
- They recommend aggressive scaling without discussing marginal efficiency or saturation.
- They present spend increases as a universal solution.
6) Ask for examples of disagreement
A credible agency can describe times when data contradicted platform reporting.
Ask:
- Tell me about a time platform attribution looked strong but incrementality was weak. What did you do?
- Have you ever recommended reducing spend even when platform ROAS looked good?
- How do you handle conflicting signals between paid social and analytics/CRM data?
A good agency should be comfortable saying:
- “This channel gets too much credit in-platform.”
- “We held back spend because marginal returns were weakening.”
- “The channel supports assisted conversions, but not all reported conversions are incremental.”
If they’ve never had a disagreement with reported performance, that’s a concern.
7) Evaluate their reporting quality
Great agencies make measurement limitations visible.
Look for reports that include:
- Blended metrics across channels
- Confidence intervals or uncertainty ranges
- Platform vs analytics vs CRM comparisons
- Incrementality test results
- New vs returning customer split
- Spend efficiency by audience, creative, and geo
Red flag
- Only dashboard screenshots from ad platforms
- No notes on methodology changes
- No explanation for attribution shifts after tracking updates or privacy changes
8) Check whether they ask the right questions about your business
Unbiased spend guidance requires business context.
A strong agency asks about:
- Gross margin and contribution margin
- Customer lifetime value
- Repeat rate and subscription behavior
- Sales cycle length
- Offline conversion paths
- Geographic differences
- Promo calendar and seasonality
- Supply constraints
If they recommend spend without understanding these, their advice may be mechanically optimized but strategically wrong.
9) Look at their testing culture
Credible agencies are test-driven, not opinion-driven.
Ask:
- What experiments do you run by default?
- How often do you test creatives, audiences, landing pages, and budgets?
- How do you decide if a test is statistically and operationally valid?
Good sign
- They embrace uncertainty and iterative testing.
Bad sign
- Their recommendations are mostly based on “experience” and “best practices” with little experimentation.
10) Ask for the exact decision framework
A very effective question is:
“If you were recommending a 20% budget increase, what evidence would you want to see first?”
A credible agency should mention some combination of:
- Stable or improving marginal CAC / ROAS
- Incrementality evidence
- Enough conversion volume for reliable learning
- Creative and audience saturation checks
- Downstream business capacity to absorb growth
- Confidence that measurement is not overstating performance
If they answer vaguely, they may be too biased toward spending.
A simple scorecard you can use
Rate each item 1–5:
Credibility
- Understands attribution limitations
- Uses incrementality tests
- Can explain model choices
- Produces transparent reporting
- Has cross-channel measurement experience
Unbiasedness
- Compensation not tied to spend
- Will recommend cutting spend when needed
- Discusses marginal returns and saturation
- Acknowledges platform bias
- Aligns with business outcomes, not vanity metrics
A strong agency should score high on both.
Quick red flags list
Be cautious if they:
- Overstate platform-reported ROAS as if it’s causal
- Avoid discussing incrementality
- Push budget growth without marginal efficiency analysis
- Have financial incentives to increase spend
- Can’t explain discrepancies between platforms and analytics
- Focus only on one channel’s results in isolation
The best single vetting question
If you only ask one thing, ask:
“How do you know the spend you recommend is incremental, and what evidence would cause you to reduce it?”
A trustworthy agency will answer clearly and concretely.
If you want, I can turn this into:
- a vendor interview checklist,
- a scored evaluation spreadsheet, or
- a list of specific questions to ask agencies in a pitch meeting.
Observation history
Unlock observation history
Create a free Obsurfable account to unlock historical responses, model comparisons, and deeper brand intelligence.
How did Obsurfable measure this prompt?
Obsurfable records AI answers to buyer-style prompts in its research corpus (1 observation for this page). Metrics are distributions over observations, not a single static ranking.
Which AI systems does Obsurfable collect answers from?
OpenAI, ChatGPT, Google, Gemini, Google AI Mode, Anthropic, Claude, Perplexity, Grok, DeepSeek, Mistral, Copilot, and Meta AI.