> ## Content Index > Fetch the complete content index at: https://blog.heimdex.co/llms.txt > Use this file to discover other available public pages before exploring further. # Video Search Vendor Checklist: 7 Questions to Ask Before You Sign - URL: https://blog.heimdex.co/en/video-search-vendor-checklist-7-questions-en/ - Published: 2026-08-24T04:24:35.000Z - Updated: 2026-08-24T04:24:34.000Z - Description: Video search demos always impress. Here are 7 questions that reveal what really happens after you sign — before it costs you months and budget. - Author: Heimdex - Tags: Insights Every vendor demo looks the same. Query goes in, right clip comes out, everyone nods. The differences show up later. In the security review. In a quote line nobody expected. In month three, when indexing the real archive takes eleven days instead of the two hours the demo implied. Here are the seven questions that separate vendors after the demo ends. --- ## 1\. Where does the original footage get processed? If your archive has unreleased material or likeness-rights complications, this filters candidates before you compare anything else. **Ask:** On-premise, cloud, or both? Does on-premise operate in a closed network? What requires outbound connections (license checks, model updates, logs)? Where do vectors get stored? **Judge by:** "We support on-premise" isn't enough. **Get closed-network operation and every outbound connection documented in writing.** That's what your security review will catch. **Heimdex:** Both models. On-premise keeps originals in existing storage with vectors stored internally. The distinction is made explicit before contract. --- ## 2\. Does it run on your existing storage, or replace it? Search budget and storage budget come from different approval lines. Some solutions assume a specific platform, quietly turning software procurement into an infrastructure project. **Ask:** Does it integrate with our current NAS as-is? Is separate storage required? What server and GPU specs, and can we use existing hardware? **Judge by:** Hand over your equipment list and ask **"does this work on this configuration?"** Watch for hardware appearing as a separate quote line. **Heimdex:** Deploys on existing local, NAS, or internal storage. GPU requirements calculated against archive size and target indexing time. --- ## 3\. How long does indexing take, and what does it cost? Demos run on short samples. The real load hits when you index the full existing archive. **Teams that skip this find out after deployment.** **Ask:** How many days for our archive (10 TB / 5,000 hours)? On what hardware? One-time cost or ongoing? What happens when new footage arrives daily? **Judge by:** Ask for the number against *your* archive, on hardware you'd actually run. A vendor who can't produce that estimate hasn't done it at your scale. **Heimdex:** Indexing time estimated during evaluation, not after. --- ## 4\. Does search work on footage with no metadata? Every team has the folder nobody tagged. That's usually where search would help most, and where tag-dependent tools return nothing. **Ask:** Does search work on folders with no filenames or tags? Does it search existing tags in parallel? Can the pilot run on our worst-metadata archive? **Judge by:** Semantic search is table stakes now. **The question isn't whether it works, it's how well.** Confirm with your footage, not the vendor's demo reel. **Heimdex:** No dependency on tags or filenames. Pilots run on the archive the customer specifies, including untagged material. --- ## 5\. Is this a finished application, or an API you'll build on? Two product types exist in this market. One gives your dev team models and APIs. The other gives editors an application. **Without internal engineering capacity, the first means a development project follows the purchase.** **Ask:** Is there an editor-facing interface, or do we build it? Are permissions, accounts, and audit logs included? Can results hand off to editing tools? Are adjacent features (summarization, auto-editing, blur) in the same product? **Judge by:** If the quote has a separate "integration development" line, **pin down its scope and owner.** That number sometimes exceeds the license. **Heimdex:** Finished application. Natural language search, segment summarization, short-form repurposing, and likeness/PII blur in one product. --- ## 6\. Who handles deployment and incident response? On-premise solutions run far longer than they take to deploy. In broadcast environments tied to air schedules, **response time matters as much as features.** **Ask:** Local deployment team, or routed through overseas HQ? What's the incident SLA? Do model updates require downtime? On termination, in what format can we export our index? **Judge by:** **That last question is the one to insist on.** Without a data export clause, switching vendors later becomes practically impossible. **Heimdex:** Deployment and support handled domestically, in-house. SLA negotiated by contract scale. --- ## 7\. What happens as the archive grows? A tool that performs well on 500 hours can degrade at 50,000\. Query latency, ranking quality, and rebuild time all scale differently. **Ask:** What's the largest archive in production with this tool? How does query time change as the index grows? Does new footage trigger re-indexing of existing material? **Judge by:** Ask for a reference customer at or above your archive size. **If the largest deployment is much smaller than your library, you're the scale test.** **Heimdex:** Incremental indexing, so new footage doesn't trigger full re-indexing. Scale references available on request. --- Run these past any vendor you're seriously considering, including us. The answers narrow the field faster than another round of demos. [**Talk to us about your archive →**](https://heimdex.co/?ref=blog.heimdex.co) [![](https://storage.ghost.io/c/1b/90/1b90bca5-caef-43b3-90d6-4b197d90c045/content/images/2026/08/2-------.png)](https://www.heimdex.co/en/contact?ref=blog.heimdex.co)