What Does Searching for Documents Cost You?
Every hour a staff member spends hunting through drives, wikis and email, or interrupting a colleague to ask where something is, has a wage cost. This calculator turns your own headcount, hours and rates into an annual figure, then shows what a knowledge assistant would return if it removed the share you nominate.
Use your own estimate for hours per week. If you are unsure, ask five people in the team to track it for three days. The recoverable share is a judgement call, so the calculator leaves it to you rather than assuming one.
Document Search Time Cost
Your numbers in, an annual cost and recovered value out.
Include time spent answering other people's where-is-it questions.
Salary plus superannuation and on-costs, divided by paid hours.
Our single use case range is $25,000 to $60,000.
Annual search cost multiplied by the share you set.
Capacity estimate, not a cash saving. Recovered hours become value only if they are redeployed. The 46 working weeks allow for annual leave and public holidays. Inputs stay in your browser and are not sent to us.
Reading the Numbers
The annual search cost is a cost you are already paying, spread invisibly across payroll. The recovered value is the slice of it a working knowledge assistant would hand back.
The annual search cost is real spend
Staff multiplied by hours per week multiplied by working weeks multiplied by hourly cost. It never appears as a line item, which is why it is easy to ignore, but the hours are on the timesheet whether or not anyone counts them.
Recoverable share is the honest lever
No assistant removes all searching. Some questions need a person, some documents do not exist yet, and some of the time is thinking rather than looking. Set the share you believe is realistic for your documents and adjust it after a pilot.
Payback compares value with a build budget
If you enter a build budget, the calculator divides it by the monthly recovered value to give a payback period in months. Our published range for a single use case deployment is $25,000 to $60,000, which is where the default sits.
How the Figures Are Built
Four multiplications and one division. Every input is something you can measure in your own organisation.
Annual search hours
Staff count multiplied by hours per person per week multiplied by 46 working weeks, which allows for annual leave and public holidays.
Annual search cost
Annual search hours multiplied by the fully loaded hourly cost. Use salary plus superannuation and on-costs divided by paid hours, not the base rate.
Value recovered
Annual search cost multiplied by the recoverable share you set. The same share is applied to hours to show the time returned to the team.
Payback
Your build budget divided by the monthly recovered value, capped at 240 months. A short payback means the project funds itself from time already being spent.
Where the Search Time Actually Goes
The hours are rarely one big search. They are dozens of small ones a day, and most of them fall into a few patterns that a retrieval system handles well.
Finding the current version
The document exists, probably in three places, and the time goes into working out which copy is authoritative. A knowledge assistant that indexes one source of truth and cites it removes this class of search almost entirely, provided the source inventory was done properly first.
- Duplicate and superseded copies are the usual cause
- A citation back to the source document rebuilds trust quickly
- Freshness rules decide whether the answer stays correct
- Use the RAG data preparation checklist before indexing
Asking a colleague
The question is simple, the person who knows is busy, and both people lose time. This is the search cost that is easiest to underestimate because it lands on the expert, not the asker. Count both sides when you estimate hours per week.
- Interruptions cost the expert more than the asker
- Policy, process and where-is-it questions dominate
- A good assistant answers these with a link, not a paragraph
- Log the questions it cannot answer: they show what to document next
Reading the whole thing to find one clause
Contracts, procedures, standards and long reports are searched by scrolling. Retrieval that returns the relevant passage with its location cuts this to a check rather than a read. The gain is largest for staff who work across many long documents a day.
- Legal, compliance, engineering and claims teams see this most
- Chunking and metadata quality decide how precise the passage is
- Answers must show the source so the reader can verify
- Keep humans responsible for the decision, not the assistant
Onboarding and cover
New starters and people covering a role spend far more time searching than established staff. If your team has turnover or rotates duties, the average hours per week in this calculator is probably understated for those periods.
- Run the calculator separately for new starters if turnover is high
- An assistant shortens the time to first useful answer
- Captured questions become onboarding material
- Permissions must follow the role, not the person
Next Steps
Enterprise Knowledge Base
What a private knowledge assistant over your documents looks like in practice.
See the solution →RAG Data Preparation Checklist
The document preparation work that decides whether the recovered share is realistic.
Work through the checklist →LLM Use Case Scorecard
Score document search as a use case before committing budget to it.
Score the use case →Frequently Asked Questions
Ask a handful of people across different roles to note every search, lookup or where-is-it question for three working days, including time spent answering other people. Average it, multiply by five for a week, and use that. It is a rough figure, but it is your figure, which is worth more than any published average. Re-run the calculator once you have better data.
The fully loaded cost of an hour of staff time: base salary plus superannuation, leave loading, payroll tax and other on-costs, divided by paid hours in the year. For a mixed team, use a weighted average or run the calculator once per band. Using base salary alone understates the result.
The calculator deliberately does not suggest one, because it depends on how much of the searching is for things that are actually written down. If most questions are answered by documents that exist and are current, a higher share is defensible. If the knowledge is mostly in people's heads, set it low: an assistant cannot retrieve what was never recorded. Pilot with one team, measure, then update the share.
Fifty-two weeks less four weeks of annual leave and roughly two weeks of public holidays. If your staff are part-time or on different leave arrangements, scale the hours per week input rather than the weeks.
Not directly. It is time returned to staff, which becomes value only if that time is used for something productive or lets you avoid a hire. Treat the figure as capacity, and decide separately what the capacity is worth to you. Payback in months is a useful way to compare a build budget with that capacity without pretending it is cash.
Inputs are saved in your own browser so they persist between visits, and nothing is transmitted to us. Use the reset button to clear them.
Want to Test the Recoverable Share on Your Own Documents?
Send us the annual search cost this calculator gave you and a description of the documents your team searches most. We will scope a pilot on one document set so the recoverable share is measured, not guessed.