Rehearsals Alternatives: AI Customer Intelligence With Real Purchase Data
Rehearsals builds AI replicas of real customers from 15-minute structured interviews. You can test pricing changes, ad creatives, product concepts, and UX decisions against these replicas — getting results in minutes instead of the 4-12 weeks traditional research takes.
It's a genuinely useful product for certain use cases. The team (ex-Google Gemini, ex-Wayfair, ex-DeepMind) is strong, the benchmarks are real (91% distribution accuracy, Disney+ pricing prediction within 4.3 percentage points of actual), and the approach — building from individual interviews rather than demographic stereotypes — is methodologically sound.
But Rehearsals isn't the right tool for every business making customer-facing decisions. If you're a D2C ecommerce brand, an online retailer, or any company running on Shopify and Klaviyo, there are approaches that work with your actual purchase data rather than simulated responses.
Here's how the alternatives stack up.
What Rehearsals Does Well
Credit where it's due. Rehearsals gets several things right that most AI tools miss:
Individual-level fidelity. Each replica is grounded in a real person's interview, not a demographic composite or LLM-generated persona. This matters — they correctly argue that synthetic personas built from demographic data rely on stereotypes and can't explain why someone has a preference.
Inductive modeling. Patterns emerge from individual cases rather than being deduced from general AI training data. This produces more nuanced and less "average" responses than prompting ChatGPT with a persona description.
Speed. Minutes to results vs. weeks for traditional focus groups. At a fraction of the cost.
Practical benchmarks. The Disney+ pricing study is a real proof point — they predicted consumer response to a $3 price increase with reasonable accuracy, outperforming both Gemini 3.0 and GPT-5.2.
Where Rehearsals Falls Short
The limitations become clear when you consider the full picture:
Interview dependency. Every replica requires a structured interview with a real person. Want to simulate your 10,000-customer base? You'd need to interview thousands of them. This creates a fundamental scalability constraint — you're always working with a sample, never your full customer base.
Stated preference bias. Rehearsals' own blog cites research showing stated preferences explain only 5-15% of actual buying behavior and consumers overstate purchase intent by 2-5x. The replicas are built from interviews (stated preferences), which means the very data they're encoding has a demonstrated gap with actual behavior.
No purchase validation. There's no mechanism to check whether a replica's prediction matched a real outcome. The Disney+ benchmark is the strongest proof point, and it measures stated willingness to keep a subscription, not confirmed subscription behavior.
No commerce integration. Rehearsals operates as a standalone research platform. It doesn't connect to Shopify, Klaviyo, Stripe, Skio, or any commerce tool. Insights stay in the research silo.
No feedback loop. Predictions never get measured against actual outcomes, which means the system doesn't improve over time. The 100th prediction is no better calibrated than the first. (For a deeper analysis of this stated preference problem across all digital twin platforms, see AI Digital Twins for Customer Behavior.)
The Alternatives
1. Mercana — Real Purchase Data + Enriched Personas
Best for: D2C ecommerce brands running Shopify + Klaviyo who need customer intelligence grounded in real purchase data.
Mercana takes the opposite approach to Rehearsals. Instead of building customer replicas from interviews, Mercana enriches your actual customers with 100+ data points from public sources — social media profiles, job titles, income signals, VIP status, interests, professional context — and connects that identity layer directly to real purchase and engagement data.
This creates the data foundation that interview-based platforms can't replicate:
- Enrich — Automatic identity enrichment for every customer (no interviews needed). 317,000+ profiles enriched with 94.4% VIP detection precision.
- Identify — Surface hidden VIPs (influencers, athletes, executives, retail buyers, high-net-worth) automatically across 15+ categories.
- Understand — See which customer personas, wealth brackets, and identity signals drive the most lifetime value.
- Act — Sync enriched tags and segments directly to Klaviyo for identity-informed flow routing and personalization.
The key differentiator: Mercana connects enriched customer identity to real Shopify orders. You can see that high-net-worth customers have 82% higher CLV, that athlete VIPs are worth 2.7x average, or that a specific persona cluster retains 3x better — insights grounded in actual purchases, not interview responses.
| Feature | Rehearsals | Mercana |
|---|---|---|
| Data source | 15-min interviews | Automatic enrichment (100+ public data points) |
| Scales to full customer base | No (interview-dependent) | Yes (enrich all customers) |
| Has real purchase data | No | Yes (Shopify orders + Klaviyo events) |
| Commerce integration | None | Shopify + Klaviyo + Skio |
| VIP/influencer detection | No | Yes (15+ categories, 94.4% precision) |
| Campaign execution | Export insights manually | Direct Klaviyo tag + segment sync |
| Pricing | Demo-based (enterprise) | $79-$299/mo |
When to choose Mercana over Rehearsals: You need customer intelligence connected to real purchase behavior — understanding which customer segments actually drive revenue, not simulating how interview subjects say they'd respond.
2. Simile — Enterprise Behavioral Simulation
Best for: Large enterprises (CPG, financial services, telecom) doing general-purpose behavioral research.
Simile is the most academically credentialed competitor — founded by the Stanford team that invented generative agents (10,000+ citations) and coined the term "foundation model." They recently raised $100M from Index Ventures and have partnered with Gallup to build AI simulations grounded in nationally representative panels.
Their digital twins are built from deep qualitative interviews and have demonstrated 85% accuracy on General Social Survey replication. Named customers include CVS Health, Wealthfront, and Banco Itau.
| Feature | Rehearsals | Simile |
|---|---|---|
| Interview depth | 15 minutes, structured | Deep qualitative (longer) |
| Academic backing | Ex-Google/DeepMind team | Stanford professors, invented the field |
| Published benchmark | 91% distribution accuracy | 85% GSS accuracy |
| Gallup partnership | No | Yes (in development) |
| Enterprise customers | F50 claims | CVS Health, Wealthfront, Banco Itau |
| Pricing | Demo-based | ~$100K+/year |
When to choose Simile over Rehearsals: You're an enterprise with a six-figure research budget and need general behavioral simulation (not just ad creative or pricing testing). The Gallup partnership gives them a representativeness claim nobody else has.
Limitation both share: Neither Simile nor Rehearsals can validate predictions against actual purchases. Both are fundamentally stated-preference systems.
3. Aaru — Accenture-Distributed Cohort Simulation
Best for: Enterprises already working with Accenture who want demographic cohort simulations.
Aaru operates at the cohort level (not individual), using demographic and public data to simulate how consumer segments will respond. Their Accenture partnership gives them distribution into large enterprises. They've raised $50M+ and claim 90% correlation accuracy, though the methodology behind that number is less transparent than Simile's or Rehearsals' benchmarks.
When to choose Aaru: You're an Accenture client and want simulation capabilities integrated into your existing consulting relationship.
Limitation: Cohort-level simulation misses individual variation. No commerce integration. Biggest public proof point is predicting one election.
4. Traditional Research (Surveys, Focus Groups, UserTesting)
Best for: Teams that need qualitative depth from real human conversations and are willing to wait.
Traditional research still has an irreplaceable advantage: actual humans telling you things that no AI model can infer. The nuance of body language, the unexpected insight from an off-script question, the emotional resonance that only comes from watching a real person interact with your product.
The tradeoffs are well-known: $20K-$600K+ per study, 4-12 weeks turnaround, small samples (typically 8-30 people), and structural biases (social desirability, response bias).
When to choose traditional research: You're exploring a completely new market, need regulatory-grade consumer evidence, or want the depth that only real human interaction provides.
5. Frontier LLM Prompting (ChatGPT, Claude, Gemini)
Best for: Quick directional gut-checks with zero cost and zero setup.
You can prompt any frontier LLM with a customer persona and ask how they'd respond to your pricing change or ad creative. Rehearsals correctly argues this approach defaults to mean/average responses and misses individual quirks. They're right — this is the baseline that every specialized tool should beat.
When to use LLM prompting: Brainstorming sessions, very early concept validation, or when you need a directional answer in 30 seconds and don't need statistical rigor.
Decision Matrix: Which Tool For Which Job
| Decision Type | Best Tool | Why |
|---|---|---|
| Understanding who your D2C customers are | Mercana | 100+ enriched data points + Klaviyo/Shopify integration |
| VIP/influencer identification | Mercana | 15+ VIP categories, 94.4% precision, connected to purchase data |
| Identity-based Klaviyo flow routing | Mercana | Direct tag/segment sync + persona-based personalization |
| Knowing which segments drive revenue | Mercana | Enrichment connected to real Shopify orders |
| Ad creative pre-testing (brand) | Rehearsals | Strong benchmarks for creative prediction |
| Pricing simulation (no existing customers) | Rehearsals or Simile | Can simulate from interview data |
| Enterprise CPG research | Simile | Gallup partnership + enterprise pedigree |
| Quick directional check | ChatGPT/Claude | Free, instant, good enough for brainstorming |
| Deep qualitative understanding | Traditional research | Nothing replaces real human conversation |
| Enterprise via Accenture | Aaru | Distribution channel advantage |
The Bigger Picture
The synthetic consumer research space is real and growing. Simile, Rehearsals, and Aaru are building genuine technology with strong teams and real enterprise customers.
But there's a structural gap in the market: none of these platforms connect prediction to real purchase outcomes. They all validate against what customers say (surveys, interviews, stated intent) rather than what customers do (purchase, subscribe, churn, convert).
For D2C ecommerce brands — where every decision eventually shows up as revenue in Shopify and engagement in Klaviyo — the most valuable customer intelligence isn't built from interviews with a sample. It's built from enriched identity data connected to real commerce behavior across your entire customer base.
That's the data position Mercana occupies. Enriched personas connected to real wallets — the foundation that makes purchase-validated prediction possible. No other platform in the synthetic consumer research space has both sides of that equation.
Related articles
- AI Digital Twins for Customer Behavior — Deep dive into why simulations without purchase data are just guessing
- Customer Intelligence for Ecommerce: The Complete Guide — What customer intelligence is and how enrichment works
- VIP Customer Detection for Ecommerce — How AI-powered VIP detection finds hidden value in your customer base
Ready to find the VIPs in your customer base?
Mercana enriches your Shopify customers with 100+ data points. Setup takes 2 minutes. First 1,000 enrichments free.