> ## Documentation Index > Fetch the complete documentation index at: https://docs.bitscale.ai/llms.txt > Use this file to discover all available pages before exploring further. # Reading Your Data > Confidence scores, freshness windows, catch-all domains, and cascade enrichment. Know when to trust a result and when to verify. **F-4 ยท GTM Foundation ยท 100 XP ยท \~12 min** Complete the data audit challenge at the bottom to earn 100 XP. # Reading Your Data You ran your first enrichment. You got results. Now the question nobody tells you to ask: **how much of this can you actually trust?** Not all enrichment data is equal. A company headcount at 90% confidence is different from one at 40% confidence. An email address on a catch-all domain is different from one that's verified deliverable. A funding date from last week is different from one that's 18 months old. This module teaches you to read the signals behind the values, not just the values themselves. ๐ŸŽฌ **Video coming soon.** *** ## Confidence Scores: What They Actually Mean Most Bitscale enrichments return a confidence score alongside the data. Here's how to read it: | Confidence | What It Means | Action | | ---------- | ---------------------------------------- | --------------------------------- | | 90โ€“100% | Multiple sources agree. High confidence. | Use it. | | 70โ€“89% | Probable match. One primary source. | Use it โ€” verify if high-stakes. | | 50โ€“69% | Possible match. Conflicting signals. | Flag for manual review. | | Below 50% | Unreliable. The provider is guessing. | Discard or run a second provider. | **The 94% trap:** A 94% confidence score doesn't mean the data is 94% accurate. It means the algorithm is 94% confident in *its guess*. The guess can still be completely wrong. For email addresses especially, always run a separate validation step โ€” confidence score โ‰  deliverability. *** ## Freshness: When Was This Data Last Seen? Some enrichment providers show a `last_seen` or `updated_at` timestamp. When they don't, use these rules: | Data Type | Safe Freshness Window | Notes | | --------------- | --------------------- | ------------------------------------- | | Funding data | Up to 12 months | Older than that โ€” re-check manually | | Headcount | Up to 6 months | Accurate to ยฑ20% within this window | | Job titles | Up to 6 months | \~15% error rate at 6+ months | | Email addresses | Up to 3 months | Format-valid โ‰  deliverable after this | **Practical rule:** For active outreach, use data under 6 months old. For list building and research, up to 12 months is usable. For anything older โ€” re-enrich before sending. *** ## The Catch-All Domain Problem A catch-all domain is one where *every* email address returns as "valid" during verification โ€” regardless of whether the mailbox actually exists. This is the #1 undetected deliverability trap. **How it works:** * Normal domain: `john.smith@acme.com` โ†’ Verification checks if this specific mailbox exists โ†’ Returns TRUE or FALSE * Catch-all domain: The server accepts *any* email to this domain as valid, including completely made-up addresses โ†’ Verification always returns TRUE **Why it matters:** If you send to catch-all addresses and half the mailboxes don't exist, your bounce rate spikes. High bounce rate โ†’ damaged sender reputation โ†’ everything you send starts landing in spam. **How to handle catch-all domains in Bitscale:** 1. Add an email verification column to your grid 2. For rows flagged as catch-all, either skip them entirely or use a tool with better catch-all detection (MillionVerifier handles this better than most) 3. Add a filter: `Export only IF email_verified = TRUE AND catch_all = FALSE` *** ## Cascade Enrichment: Don't Stop at One Provider When a single provider returns empty or low-confidence data, don't accept it as a dead end. Run a cascade. **The pattern:** ``` Column A: Email (Apollo) โ†’ Primary provider Column B: Email (Hunter.io) โ†’ Fallback if Column A is empty Column C: Email (Best) โ†’ AI column: use A if not empty, else B Column D: Email (Verified) โ†’ Verification on Column C result ``` Enterprise Bitscale users get 85%+ email coverage using cascade enrichment vs. 60โ€“65% with a single provider. **AI column logic for cascade (paste this directly):** ``` If {{email_apollo}} is not empty and not null, return {{email_apollo}}. If {{email_hunter}} is not empty and not null, return {{email_hunter}}. Otherwise return "not_found". ``` Set this AI column to "Logic only" mode โ€” no LLM call needed, it's just conditional logic. *** ## Your Data Quality Checklist Before exporting any grid for outreach, run through this: * [ ] Confidence scores reviewed โ€” anything below 70% flagged or removed * [ ] Freshness checked โ€” no data older than 6 months in active use * [ ] Catch-all domains identified and handled * [ ] Email verification column run (separate from enrichment) * [ ] Empty contact rows reviewed โ€” worth manual lookup or remove? * [ ] Headcount filter applied โ€” company size still matches your ICP? **Rule of thumb:** A clean list of 200 rows beats an unverified list of 1,000 rows every time. Your deliverability, reply rates, and SDR time all improve when the list is smaller and better. *** ## ๐ŸŽฏ Challenge ยท 100 XP Run a data quality audit on the grid you built in F-3. 1. Add a **confidence score review** โ€” identify any rows where confidence is below 70% 2. Add an **email verification column** and run it 3. Flag any **catch-all domains** (your email verification tool should do this) 4. In your submission, report: * How many rows passed quality checks * How many were flagged or removed * What percentage of emails came back verified and deliverable [Submit your F-4 audit โ†’](https://bitscale.fillout.com/academy-challenge-f4) *** One more Foundation module. Then you're ready for the Knowledge Check and 200 bonus XP.