AI Pricing Analysis
Get AI-powered recommendations on your pricing strategy and margins.
AI Pricing Analysis (Smart Pricing)
Smart Pricing reviews your own estimate history and gives you actionable recommendations. Open AI > Smart Pricing (/ai/pricing) to run an analysis.
Recommendations are suggestions, not automatic changes.
Viewing a saved analysis requires the view_financial_reports permission; running a new one requires manage_pricing. Without manage_pricing, the page shows the latest saved analysis (if any) but the Run Analysis button is disabled.
If OpenAI is not configured for the workspace, or a provider call or its response fails, the page still returns a statistical summary and fallback recommendations instead of showing a dead-end error — see "The AI badge" below.
What Smart Pricing Analyzes
Each run builds a deterministic statistical summary from your organization's own last 500 estimates (all locations), grouped by service type (residential, commercial, holiday lights, generic):
| Data point | What it tells the analysis |
|---|---|
| Estimate and win counts | How many estimates and accepted (sold) estimates exist per service type |
| Win rate | Accepted estimates divided by total estimates, per service type and org-wide |
| Average estimate / sold totals | The average estimate amount, and the average among only accepted ones |
| Margin | Gross margin from the same cost-snapshot data as the Margins report, shown only when every accepted estimate in that service type has cost data — otherwise "--" |
| 90-day trend | Estimate count and average total among estimates from the last 90 days |
| Cross-service clients | Pairs of service types bought by the same client, when at least 2 clients bought both — the only input to bundle recommendations |
V2 does not have a regional pricing-benchmark source. The analysis reports only your own historical data — no market or competitor comparison is fabricated. This shows up on the page as a note under the summary.
Running an Analysis
- Go to AI > Smart Pricing.
- Click Run Analysis (or Run Analysis Again if one already exists).
- Wait a few seconds while CE Pro builds the summary and generates recommendations.
- Review the results on the page.
When OpenAI is configured, a standard analysis reserves 5 credit cents against the org's monthly AI credit before calling the provider (see AI Features Overview). If OpenAI is not configured, the statistical fallback does not reserve or consume AI credits. Provider errors or invalid responses also release the reservation.
The saved result and its credit charge commit together. If publishing the result fails, the charge rolls back and the reservation is released. After provider work finishes, CE Pro rechecks the member's current Manage pricing access and location scope before it saves the result or settles credits. If the member was deactivated, lost permission, or changed location scope during the provider call, the analysis is not published and the reservation is released.
The AI badge and the 24-hour cache
Every analysis is labeled AI (generated by the configured OpenAI model) or Statistical fallback (the deterministic rules below, no model call) next to its generated timestamp. Fallback is used whenever no OpenAI key is configured for the workspace, or the model call or its response fails validation — Smart Pricing always returns a usable result rather than an error.
A generated analysis is cached per org for 24 hours. Opening the page shows the most recently saved analysis, however old, with its Generated timestamp — CE Pro does not automatically refresh a stale one for you. Click Run Analysis Again to force a new analysis regardless of how fresh the cached one is.
Understanding the Results
The analysis produces three sections:
Pricing Summary
A short written summary, plus the total estimates analyzed, the org-wide win rate, and the "no regional benchmarks" note described above.
Service-Level Recommendations
One card per service type (plus a card for each qualifying bundle pair), each labeled with one of four kinds:
- Increase price — This service type's win rate is meaningfully above the org average and has enough volume to support it.
- Hold price — Win rate is close to the org average, or there isn't yet enough volume (fewer than 5 estimates) to trust a signal.
- Decrease price — This service type's win rate is meaningfully below the org average.
- Bundle opportunity — Two service types are bought together by enough clients that a bundle discount could increase average deal size. Bundle cards list both service types instead of one.
Every card shows its current rate, suggested range, reasoning, and a confidence level (low, medium, or high). A Statistical fallback analysis is always low confidence and never includes a projected win-rate or revenue change — those two fields only ever come from the AI model, and CE Pro omits them rather than showing a fabricated number when they aren't available.
Margin by Service Type
A table showing estimate count, win rate, average estimate, average sold amount, and margin per service type. Margin reads "--" for a service type where cost data isn't available for every accepted estimate yet.
Acting on Recommendations
Recommendations are suggestions, not automatic changes. Smart Pricing never writes to your price book, redline pricing, or the pricing engine. To adjust your pricing:
- Review each recommendation carefully, including its confidence level.
- Apply your own market knowledge — Smart Pricing has no regional benchmark data, only your own history.
- Go to Admin > Pricing to update your rates. See Setting Up Your Default Pricing.
- Monitor your win rate after making changes.
Running Follow-Up Analyses
Run a new analysis after making pricing changes to see the impact — Run Analysis Again always generates a fresh one, even if a cached analysis from within the last 24 hours exists. Give it a few weeks after a pricing change before comparing results, so enough new estimates accumulate to move the numbers.
Compare the new analysis to the previous one. Look for improvements in margins and stable or improving win rates.
Tips
- Do not change everything at once. Adjust one or two service types at a time so you can measure the effect of each change.
- Watch the confidence level. A low confidence card (including every Statistical fallback card) is a weaker signal than a medium or high one from the AI model.
- Review regularly. Run a new analysis periodically to stay current as your estimate mix changes.
- Factor in your costs. Smart Pricing looks at revenue, win rate, and margin where cost data exists. Make sure you also account for changes in your supply costs, labor rates, and fuel expenses.
Related articles
Was this article helpful?
Still need help? Contact support