The Lost Slide Problem

Slides get re-used constantly. But when they move from deck to deck, version, approval, evidence and usage can become disconnected. We reviewed the published research and audited 36 statistics commonly used to quantify this problem, tracing each one back to its source. What we found was surprising: the behavior is real and well described, but the rate and cost of rebuilding slides have never been credibly measured. Not once, in twenty-five years of the industry putting dollar values on it.

Finding the slide is only half the problem

Suppose retrieval were solved and the right slide came up first time. It still does not tell the person who found it whether this is the current version, whether it was approved and by whom, whether its figures rest on a source that is still valid, or whether it was built for this audience.

Without those answers the find is a dead end, and one rational move is to build the slide again, this time because it could not be trusted rather than because it was lost.

For a slide, provenance means four things:

• Version. Which iteration this is, and whether a newer approved one exists.
• Approval State. What approval applies, who gave it, and whether it still holds.
• Evidence. The source behind each claim and figure, so a change to the source can be traced to every slide that depends on it.
• Usage. Where the slide has been used, so affected presentations can be identified when it changes.

Each one answers a question that otherwise stops a re-use decision dead. Without Version, the safest assumption is that something newer exists somewhere else. Without Approval State, a slide approved eighteen months ago is indistinguishable from one that was never approved at all. Without Evidence, a figure cannot be defended the moment somebody asks where the number came from. Without Usage, nobody can say who else is presenting the slide on the day it turns out to be wrong.

None of those is a storage problem. A perfectly organized drive with faultless naming conventions answers none of them, which is why better filing does not fix this and the research says so.

A slide carrying all four gives a re-user what they need to make a confident decision. A slide carrying none of them is a picture. The framework is deliberately product-independent: any system, process or convention that keeps those four facts attached to the slide is doing the work, and any that lets them fall away is not.

Why this matters now

In governed environments, losing that provenance becomes more than an efficiency problem. In a 2025 FDA action involving a speaker deck, the agency identified its concerns by specific slide number and chart title, and rejected remediation by reference: safety statements elsewhere in the deck and a link to the full Prescribing Information did not mitigate the issue it found.

A slide rebuilt or edited locally may no longer inherit the approval history of the original. The full paper examines that case in detail, along with formal certification requirements in the UK.

The problem also changes as AI becomes part of presentation creation. Historically there was friction in rebuilding, because somebody had to redraw the chart or reassemble the deck. That friction is disappearing. AI can make it easier to find approved content, detect near-duplicates and connect slides to evidence. It can also make creating another version easier than finding the governed one.