Time lost searching for information: what it costs your service team.

1. The evidence: how much time searching for information costs in service

The most robust figure for service in the German-speaking countries comes from the Insight-Report Service 2022, a survey of service technicians in Germany, Austria and Switzerland run by the documentation service provider kothes together with the customer service associations of the three countries. Asked “How much time per day do you spend searching for information to complete your service tasks?”, they answered as follows:

Two things stand out. First, searching is not a fringe habit of a few technicians but the norm - the two groups that do not have to search add up to less than a tenth. Second, mechanical and plant engineering is over-represented in every long time band. There, almost four in ten search for an hour or longer, and for one in eight it is more than two hours a day - over a five-day week, more than a working day. The third survey round in 2023 confirmed this picture.

That this is not a peculiarity of the German-speaking countries is shown by two sources from the United States. In the Service Council’s Voice of the Field Service Engineer 2024, 30 percent of the technicians surveyed named the time spent finding information as the thing they dislike most about their working day - for the first time ahead of paperwork, which had held first place for five years. And according to a survey by the software vendor Salesforce (technicians in the US, 2025), 30 percent of a technician’s working hours go on information gathering and administration - more than the 28 percent spent actually delivering the service.

The formula “one working day a week”, which turns up in many vendor presentations, has a different origin, though. It usually goes back to a survey by the software vendor ABBYY from 2021: office workers in five countries, of whom 25 percent in Germany said they lose a full working day a week searching documents. Office workers are not service technicians. The sentence still holds for service, only more narrowly: it applies to the eighth of technicians in machinery manufacturing who search for more than two hours a day. For the largest group it is two and a half to five hours a week.

2. Why search time is only the visible part of the cost

Anyone who calculates the cost of searching for information as hours times hourly rate underestimates it. A search that finds nothing does not end - it turns into something else. Three knock-on costs are visible in the data.

The repeat visit. For 95 percent of the technicians surveyed in the Insight-Report, missing access to information has at some point meant that further visits were needed to complete the job; for a good quarter it happens on more than one job in five. What a repeat visit costs, we worked out in the article on the first-time fix rate. The order of magnitude: according to the benchmark by the software vendor Aquant (vendor study), a case that is not resolved on the first visit takes 2.7 visits and 13 additional days on average; in the 2026 edition, failed visits account for a quarter of total service costs at the median, and 44 percent at the weakest organisations.

The phone call. A search that fails becomes a call: to the hotline, to the colleague who knows the machine series, to the design engineer. Now it is not one person searching but two - and the second is usually the most experienced person in the company, whose time is scarcest. The Insight-Report shows why these calls are so frequent: a good 83 percent of technicians get by with their own notes, and 70 percent use those notes only for themselves. The knowledge that would shorten the search exists - it just cannot be found.

Satisfaction on both sides. At the customer, the machine stands still while the search goes on. For the technician something else adds up: searching is what they dislike most about their working day - and in a trade where successors are scarce, that is not a matter of comfort.

3. Model calculation: what searching for information costs with 100 technicians

The following is a model calculation, not a projection. It deliberately takes the lower bound of each time band from the Insight-Report, counts the technicians who find information straight away or need none as zero, and values the hours at the labour cost of the manufacturing sector per hour worked that the German Federal Statistical Office reports for 2025.

Arithmetically, almost ten in a hundred technicians in such an organisation do nothing all year but look for information. The number does not get small even if the assumptions are lowered further: with 180 field days and half an hour of search time for everyone who searches at all, around EUR 400,000 remain. And it gets considerably larger once the knock-on costs from section 2, which this table deliberately leaves out, are added.

Two caveats belong here. Not every minute of searching is avoidable - a technician who reads the safety instructions before an intervention is not searching, they are working. And the hours do not turn into money but into capacity: nobody lays off ten technicians because the search gets shorter. The capacity turns into jobs that are waiting today, hotline cases that are left lying today, or the Friday afternoon job that slips to Monday today.

4. Why full-text search in a PDF is not enough

The tool most technicians search with today is the search function in a PDF file. According to the Insight-Report, 70.1 percent of respondents receive their service documentation as a plain PDF download, 62.4 percent through an online portal, 43.8 percent still as printed instructions - although 57 percent of those receiving paper do not consider it the right medium. That PDF search is not enough is shown by the same study in the answers to the question of what gets in the way of searching:

• It is not clear where to search (59.3 percent). The answer to “error 4711 on machine series X at customer Y” is scattered: the error description in the instructions for use, the fix in a service report from three years ago, the customer’s variant in the ERP, the spare part in the parts list. A full-text search searches one file. It cannot know that the answer is in another.
• It is not clear how to search (39.2 percent). Full-text search finds strings of characters, not meanings. Whoever searches for “pressure drops” does not find the section “pressure loss in the hydraulic circuit”; whoever searches for the manufacturer’s designation of the valve does not find the in-house material number; whoever types the error code the way it appears on the display does not find it if the documentation writes it differently. The technician has to know the word that is in the document - and that is exactly what they do not know, otherwise they would not have to search. This is not a weakness of technicians but a property of language: Furnas and colleagues measured as early as 1987 that two people name the same thing with the same word with a probability of less than 20 percent.
• Documents are not available on site (36.5 percent). Almost 70 percent of respondents have no internet access at the job site. Any search that requires a connection simply does not happen there.
• The information found is incomprehensible (34.9 percent). And half of the respondents miss step-by-step troubleshooting instructions. That is not a search problem but a content problem - no search in the world finds what was never written.

Add to this an obstacle the survey does not ask about, because technicians take it for granted: in machinery manufacturing, the answer is rarely in running text. It is in the exploded-view drawing with item numbers, in the torque table, in the wiring diagram, in the scanned instructions of the machine from 2009. To full-text search, those are pictures.

5. What semantic search over technical documents changes - and what it does not

Semantic search means the system compares the meaning of a question with the meaning of passages, not the letters. “Pressure drops” finds “pressure loss in the hydraulic circuit”, the material number finds the manufacturer’s designation, the error code finds its entry regardless of spelling. That solves the second obstacle from section 4. It solves the first if the index does not cover one file but all sources - documentation, service reports, parts lists, ticket system - and puts the question in the context of the specific machine: serial number, variant, modifications.

Three properties decide whether such a system is any good in service. Every answer must show its source - document, page, section -, because a plausible wrong answer costs more than any search time; which other criteria count is set out in the checklist for AI assistants in field service. Drawings, tables and scans must be processed as content, not skipped as images. And the system must work offline or hand over its answers before departure, otherwise it does not apply to seven in ten technicians on site. The obvious first place to use it is therefore not the technician in the field but the hotline they call - that is where there is internet, and that is where the cases that failed on site end up.

What semantic search does not change: the fourth obstacle. Incomprehensible documents stay incomprehensible, missing step-by-step instructions stay missing, and a paper archive that was never scanned is searchable by no system in the world. Nor does it change anything if nobody maintains the answers: a system the technicians have distrusted twice, they do not use a third time.

6. When the effort is not worth it

The model calculation above applies to a large service organisation. It scales down, but the effort for a knowledge system does not scale down to the same degree. In our view, semantic search is the wrong first investment if one of the following conditions applies:

1. The service team is small and the cases rarely repeat. For five technicians looking after one machine series, a well-kept folder with twenty fault patterns is the better answer. The benefit of a search scales with the number of technicians asking the same questions about the same installed base.
2. The documentation is thin or exists only on paper. Then the problem is not the search but the stock. Digitising and structuring come first; the Machinery Regulation from January 2027 provides a reason for that anyway.
3. Nobody has measured the search time. Anyone who does not know how long their own technicians search, and for what, can neither select a system nor judge later whether it achieved anything. The measurement costs two weeks and no budget - and it belongs at the beginning, not at the end.

7. First steps: measure search time before you buy software

Getting started needs no software. It needs four things that can be done in a month:

1. Two weeks of tally sheets. For every job, each technician notes: What did I search for? Where? How long? Found? The result is your own distribution - and usually the surprise of how often the answer is “from a colleague”.
2. The twenty most frequent questions. From the hotline and from the tally sheets. They show which sources a system has to cover first - and later they are the acceptance test for every vendor.
3. A source inventory. Where are the instructions for use, service reports, drawings, parts lists, old tickets - and in what format? How much of it is scans, how much paper? That determines the effort more than any licence.
4. A baseline. First-time fix rate and hotline volume before the start, so that the effect can later be demonstrated rather than claimed.

Only then is the conversation with vendors worthwhile - with your twenty questions, your documents and your number.

8. Frequently asked questions

9. Sources

All figures quoted in the text, with their origin. Studies by vendors with a commercial interest are marked as vendor studies. The core source is German; the links lead to the originals.

• kothes, KVD, KVA and SKDV, Insight-Report Service 2022 (PDF, November 2022; survey of service technicians in Germany, Austria and Switzerland, April to July 2022, number of participants not stated, 43.6 percent from mechanical and plant engineering; kothes is a technical documentation service provider): search time per day (p. 9), repeat visits (p. 10), obstacles to searching (p. 18), delivery media (p. 19), internet access on site (p. 21). In German.
• kothes, presentation slides on the Insight-Report Service 2023 (PDF hosted by the University of Applied Sciences Upper Austria, October 2023; third survey round April to July 2023): confirmation of the search-time distribution. In German.
• Service Council, Voice of the Field Service Engineer 2024 (symposium slides, October 2024; benchmark survey, more than 2,500 technicians, mostly US; research and membership network): time spent finding information as the top dislike of the working day.
• Salesforce, news release of 9 April 2025 (vendor study; 350 technicians and tradespeople in the US, February to March 2025): share of working hours spent on information gathering and administration.
• ABBYY, press release of 10 November 2021 (vendor study; Sapio Research, 5,025 office workers in the US, UK, France, Germany and Japan, September to October 2021; German sub-sample figures as reported by it-daily.net, in German): origin of the formula “one working day a week”.
• Furnas, Landauer, Gomez and Dumais, The vocabulary problem in human-system communication (Communications of the ACM, November 1987): probability that two people choose the same term.
• German Federal Statistical Office (Destatis), press release No. 148 of 29 April 2026 (annual estimate of labour costs 2025): EUR 49.50 per hour worked in the manufacturing sector. In German.
• Aquant, Field Service Benchmark Report 2024 (vendor study, PDF; 145 service organisations, 24 million work orders): visits and duration after a failed first visit (p. 9). 2026 edition (19 February 2026; 161 service organisations): share of failed visits in service costs.