How do you collect site readings from customers without retyping?

This method covers process readings that a customer's staff take by hand, such as coolant concentration, pH or conductivity, for a supplier's technical service. It does not replace the record format a regulator sets for GMP laboratories, environmental compliance reports or legal-for-trade metering, or data that online sensors log on their own.

What do you need before you request site readings from a customer?

Before you request site readings from a customer, you need an agreed list of what is measured, where, how and by whom, written down once for each system you support. Without it, readings arrive in different shapes and get retyped.

The US OSHA metalworking fluids manual describes the same groundwork for a fluid testing procedure. It says the written procedure should cover where and when to collect the samples, which tests to perform, a specific protocol for each test, and the name of the person responsible for performing and recording the results (US OSHA, accessed 2026). The same manual leaves the choice of parameters to the site: the fluid manager decides which factors to record and track, prioritized for that facility (US OSHA, accessed 2026).

Agree five items with the customer:

• the systems and sampling points, each with a fixed name, such as "Line 3 central sump, return side";

• the parameters for each point, with one unit per parameter;

• the test method or instrument for each parameter, such as a refractometer reading converted with the product's factor;

• the operating range for each parameter from the product's technical data sheet or the program you agreed;

• the person at the customer site who takes and enters the readings.

Add the last accepted value of each reading as the baseline for the next.

Why does retyping customer site readings cause errors?

Retyping customer site readings causes errors because each manual copy of a value is a new chance to change a digit, drop a unit or attach the wrong date, and most of those errors look like plausible numbers. A value taken from an email, a phone call or a photo is copied at least twice, at the site and again at the supplier.

Research on data entry shows how large and how hidden these errors are. In the Barchard and Pace study, single entry gave error rates of around 1%, and most errors were within the allowable range of the variables, so histograms and frequency tables would not reveal them (Barchard, Stephens and Ochoa, 2017). Checking by eye did not help: visual checking was no more accurate than single entry, and both left roughly 30 times as many errors as double entry (Barchard, Stephens and Ochoa, 2017). The study used psychology data sheets, not plant readings, so the rates are an indication, not a value for your site.

Medicines regulators reach the same conclusion. MHRA guidance asks that records be accessible at the point of activity "so that informal data recording and later transcription to official records does not occur" (MHRA, 2018). The practical lesson for site readings is to remove the copy rather than check it harder: the person who reads the instrument enters the value once, into a field that already holds the unit and the sampling point.

How do you collect customer site readings step by step?

Customer site readings are collected without retyping in six actions that carry out the four steps (request, enter, review, approve), so that the value the customer enters is the value that reaches the record.

1. Build the request from the agreed list: one field per parameter per sampling point, with the unit fixed in the field label and the operating range shown next to it.

2. Send the request to the named person at the customer site and ask for the sampling date and time, with the time zone, for each value. MHRA guidance names synchronized clocks and a known time zone as aids to traceability when data are used across sites (MHRA, 2018).

3. The customer enters each value directly from the instrument, plus any observation such as foam, odor or appearance, and any additions made since the last reading.

4. Check each value against the acceptance table in the next sections before the next reading is due, while the customer still remembers the sample.

5. Approve the values that pass into the record with your name and the date, and return the others to the customer with a specific question instead of changing them yourself.

6. Compare the approved value with the trend and act on real changes.

In Lubinpla, a field data request follows this pattern: values the customer enters become project records once the team approves them.

Which fields should a site readings request include?

A site readings request should include fields for the system, the sampling point, the sampling date and time, the parameter, the value, the unit, the test method, the person who took the reading and any observations. Each field exists because a reading without it cannot be compared with the last one or traced back to its source.

MHRA guidance explains the reason in one line about metadata: "Without the context provided by metadata the data has no meaning" (MHRA, 2018). Its examples attach a sample reference, a time and an analyst's name to a bare value of 3.5. The US EPA guidance on data verification lists the same kind of information on sample custody forms, including the sample number, the date and time each sample was collected and the analyses requested (US EPA QA/G-8, 2002).

The table below shows what each field prevents, with example entries that are illustrative only.

The key row is the fixed unit: when the unit sits in the field label, the customer types only the number, and a unit mismatch cannot enter the record.

How do you check customer-entered site readings before accepting them?

Customer-entered site readings are checked before acceptance by confirming the unit, the sampling point, the date and time, the test method, completeness and the size of the change from the last accepted reading. The table below decides whether a customer-entered reading can be accepted into the record.

The change from the last reading is the check that catches plausible-looking errors, the in-range kind that histograms miss in data entry studies (Barchard, Stephens and Ochoa, 2017). NIST describes the moving range as the absolute difference between two consecutive readings and uses its average to set the limits of an individuals chart (NIST/SEMATECH, accessed 2026). A reading whose jump from the last value falls far outside the usual moving range is more likely an entry or sampling problem than a sudden process change, and asking first costs one message. EPA data verification guidance names the same targets, listing transcription errors and conversion factors among the factors checked (US EPA QA/G-8, 2002).

What should you do with a site reading that looks wrong?

A customer site reading that looks wrong should be returned to the person who entered it with one specific question, and kept as entered until they answer, because a silent correction leaves a value no one can trace.

The US EPA data verification guidance shows how to treat such a value. In its example, 15 of 16 sample records list a volume of 1,000 milliliters and the sixteenth lists 1,000 grams, and the verifier must find out whether the unit is wrong, the entry sits in the wrong field or the sample really was weighed (US EPA QA/G-8, 2002). The guidance adds that any change to reported results should carry a note of explanation (US EPA QA/G-8, 2002).

For site readings: ask whether the value, unit and sampling point are as entered. If it was an entry error, record the correction with the reason and keep the original. If the value is confirmed, ask for a re-test or a new sample with the same method. If the re-test agrees, treat the reading as a real process signal and act on it under the agreed program.

Bring in a product or application specialist when a confirmed reading stays outside the operating range after the normal correction, or when it comes with symptoms. The OSHA metalworking fluids manual links foul odor to uncontrolled microbial growth and skin irritation to causes such as too high a concentration or high alkalinity (US OSHA, accessed 2026).

How often should you request site readings from a customer?

Request customer site readings at the frequency the customer's own testing procedure sets for each parameter, and ask for extra readings after any change to the system. Parameters drift at different speeds in different systems, so no general interval fits every site.

The OSHA metalworking fluids manual says a procedure should be in place to test the fluids periodically and that actual concentration in machines "must be checked frequently and adjusted as needed" to maintain the recommended range (US OSHA, accessed 2026). It leaves the exact interval and the priority of each parameter to the site's fluid manager, who customizes the list for the facility (US OSHA, accessed 2026). A supplier's request should follow that schedule rather than invent its own.

Fix the day and time of the routine request so that readings form an even series. Raise the frequency after a new charge, a top-up with a different batch, a cleaning or an upset, and return to the routine interval once readings settle. Keep enough consecutive readings to set limits for the change check: the NIST individuals chart example computes its limits from the first 10 batches of a new process (NIST/SEMATECH, accessed 2026). The OSHA manual also asks that data be tabulated so that relationships and trends can be spotted (US OSHA, accessed 2026).