The situation: a plant that does not stop production
The case concerns a producer of insulation materials: a continuous plant with furnaces and production lines, where external contractors work in parallel with production. About 20 contractor companies are used every day for overhauls, welding, work at height and machine service. The plant has no shutdowns that would allow work to be concentrated in a single time window.
Management asked three questions. Where and when do external companies work, and which types of work dominate? How can the location accuracy of permits be improved? How can contractor entry be controlled at the gatehouse? This description covers the first two, because they required data analysis.
The problem: the permit points to a building, not a machine
In the permit system the work location was often described in general terms: a building name, an entry such as "Other" or "Entire line". Such a location does not say where the service team returns, and security cannot see work on the right line. It is also impossible to see whether the same machine needs work every week.
Measurement confirmed the scale of the problem: 44 of 106 permits had a general location. That is over 40% of records that fell into one bucket in spatial analysis. A continuous plant pays twice for this: weaker risk control and a lack of data for planning overhauls.
Distribution of 106 permits: time, work type and location quality
The solution: data analysis and a heat map on a 3D model
The analysis covered data from the plant's instance over a period of 9 weeks: the number of permits per week, issue hours and days, types of especially dangerous work, contractor companies and plant areas. The results were plotted as a heat map of work concentration on a 3D model of the plant, which made it possible to see the spatial distribution without reading tables.
In the second step the plant structure was improved in the 3D map editor: locations, lines and floors. A contractor now selects a specific point at the machine, for example a filter, a section, a hopper or a chamber, instead of a building name. As a result, later analyses will show recurring failures at the same devices.
Scope of the analysis
Permit data, distribution over the week and over the day, types of work, contractor share and plant areas. Data on the QR gate, training and lone working did not yet exist in the analysis period, so they were not included as system data.
The effect: numbers from 9 weeks
The main patterns were clear. Permits were concentrated in time: 82% were issued between 6:00 and 10:00, and 62 of 106 fell on Tuesday and Wednesday. None were issued at the weekend. Hot work accounted for 43 permits and work at height for 40.
The concentration also applied to companies and places. The four largest contractor companies accounted for 63% of permits. Two halls, the warehouse and the production hall, collected 76 of 106 permits. In the warehouse hall 24 of 39 permits concerned work at height, and in the production hall 24 of 37 concerned hot work.
The first week of the period had 24 permits, probably because of shutdown work. This assumption needs to be confirmed with the plant, so it was treated with caution in the trend analysis.
- 82% of permits issued in the 6:00–10:00 window
- 62 of 106 permits on Tuesday and Wednesday, 0 at the weekend
- 63% of permits from the four largest companies
- 76 of 106 permits in two halls
- 44 of 106 permits with a general location
What the plant changed on the basis of the data
The first decision concerned shift cover. Since most permits are created on Tuesday and Wednesday mornings, approvers and security should be available at those hours, and inspections and acceptance are planned outside the peak. The change requires no new resources, only a schedule adjustment.
The second decision concerned companies. Four companies generate most of the traffic, so they receive training and work qualification matrices first. The third decision is a data quality indicator: the share of permits with a general location will be measured every quarter and should fall as the structure in the map editor is refined.
The fourth change is organisational. Since 43 permits concerned hot work and 24 of them were concentrated in one hall, that hall was added to the recurring fire protection review.
Conclusions for other plants
First, permit analysis requires no additional hardware: the data the system already collects is enough. Second, location quality is a condition of analysis quality, so it is worth measuring its level from the start. Third, a heat map changes the conversation with the board: instead of a table of numbers, you see where risk is concentrated.
The pattern is universal for plants where external work takes place alongside production. It is worth checking similar distributions in a chemical plant, a steelworks or a heating plant, remembering that the values concern one plant and one period and are not an industry norm.
Video
Frequently asked questions
Why is a general location a problem?
An entry that points to a building instead of a machine makes it impossible to count recurring work at the same device or to see it on a map. In the studied plant this concerned 44 of 106 permits.
Does such an analysis require additional devices?
No. The described analysis used data the permit system was already collecting. Only the location structure in the 3D map editor was refined.
Can these values be transferred to another plant?
The analysis pattern can, the numbers cannot. The distribution of hours, days and companies depends on how overhauls are organised, so each plant should measure its own period.
Let us talk about the data from your plant
We will show which reports can be built from the permits, entries and reports you already have. Book a conversation with the InnerWeb team.
Let's talk