How to Establish Whether a Customer Changed a Consumable

Why this matters

Sometimes you cannot get on site. The property is seasonal, the customer is three hours out, the unit is at a rental the owner never visits, the job is a phone consultation for a shop in another region. The consumable is the most likely cause and you will never hold it. Your only instrument is a person who is not trained, may not have been present when the change happened, and has a reason to give you the answer they think you want.

Run that interview badly and you get "no, nothing changed," which is almost always technically true and completely useless. Run it well and you get a photograph of a container label that names the cause in one message. The difference is method, and the method is learnable.

What you must never ask a remote customer to do

Set this boundary before the first question, because a helpful customer will do whatever you suggest.

Never ask an untrained person to open an energized enclosure, defeat an interlock, bypass a protective device, work on a gas or fuel connection, open anything under pressure or at temperature, get on a ladder or a roof, or reach into a machine that could start. No diagnostic answer is worth it, and instructing them creates liability that follows you regardless of what they agreed to.

The safe remote request set is narrow and it is enough: look at something, photograph something, read a label out loud, listen and describe, note a time, and check whether something is present or absent from a safe standing position. If your question needs more than that, the answer is a site visit, either yours or somebody local.

If the customer describes gas odor, smoke, burning smell, water at an electrical enclosure, or anything they think is unsafe, the conversation stops being diagnostic. Tell them to leave, shut off at a safe point if they can reach one without risk, and call the appropriate emergency service. Resume diagnosis later.

Step 1: Set the window before you ask the question

The biggest failure in this interview is asking about the wrong period. Off-spec supplies produce failures long after the change, so "recently" is exactly the period in which nothing relevant happened.

Estimate the window first, from the mechanism and the duty cycle:

1. Name the likely mechanism from the symptom. Deposits and progressive behavior point to accumulation, which is long-latency. Sudden accelerated wear points to depletion, which is medium.
2. Convert the mechanism's latency from runtime hours to calendar time using the customer's actual usage.
3. Ask about that calendar window, specifically.

What breaks if you skip this: you ask an open question, the customer scans the last month, finds nothing, and you both conclude the supply is not the cause. That conclusion then hardens, and every later conversation starts from it.

Step 2: Retire the question "did you change anything"

It fails three ways at once. It scans the recent past, it is heard as an accusation, and it asks someone with no basis for the judgment to decide what is relevant. A customer who swapped a filter brand eight months ago answers no in complete honesty, because to them that was not a change, it was buying a filter.

Replace it with questions that are specific, dated, and neutral:

• "Around last spring, do you remember anyone topping this up or refilling it?"
• "Who buys the filters for this, and where do they get them?"
• "Has the delivery come from the same supplier the whole time you have had it?"
• "Is there anything in there now that came from a different place than usual?"

Each of those is answerable from memory of an ordinary event rather than requiring the customer to identify what mattered.

Step 3: Ask them to look, not to remember

Convert every question you can into a request to walk over and look at something. The three highest-yield looks, in order:

1. The container. "Is there a jug, a bag, or a box for this anywhere - garage, shed, under the sink?" Empties and part-used containers survive far longer than memories, and the label carries the grade.
2. The item itself. "Without opening anything, can you see any printing on it? Read me whatever is written there." Absence of any marking is a real finding on its own.
3. The shelf. "What else is on the shelf next to it?" A shelf holding two different-looking containers of the same nominal product is a substitution you would never have gotten by asking.

Give one instruction at a time and wait. A stacked list of three requests gets one answer.

Step 4: Get the label into a photograph

Verbal readings of a label are unreliable in both directions. People read the marketing name rather than the grade, skip the standard, and paraphrase. Ask for a photo of the whole label including the fine print and any date or lot code, plus a second photo of the container in place.

Say why, because it changes what you get: "Send the whole label rather than reading it, there is usually a grade in small print that matters more than the name on the front."

Store the photo on the job record, not on a phone. A dated photograph from the customer's own hand carries more weight in a later billing or warranty dispute than anything you could have collected on site.

Step 5: Date the change with anchors, not with a calendar

People cannot place events on a calendar and can place them against life events with reasonable accuracy. Once you have a window from step 1, translate it into anchors:

• "That would have been around the time the weather turned."
• "Was that before or after you had the other work done?"
• "Would that have been before or after the holidays?"

Then bracket. "After the holidays but before it got hot" is a window of a few months, which is plenty to test against your latency estimate. Precision beyond that rarely changes the diagnosis.

Step 6: Read the answer, not only the words

You are working without body language, so listen for structure.

• A fast, flat "no" to a specific dated question is usually honest, because a fabricated denial tends to arrive with justification attached.
• An answer that explains before it answers ("well, the guy at the store said it was the same thing, so...") has already told you what happened. Do not pounce. Say "that is completely reasonable, they usually are close" and keep going.
• A shift from "I" to passive ("it got changed at some point") means someone else did it and they may not know who. Follow that thread rather than the original question.
• "I don't know" repeated means the interview has hit its limit. Move to physical corroboration rather than pushing, because pushing converts a cooperative witness into a defensive one.

Step 7: Interview the household or the staff, not just the caller

The person who calls the shop is frequently not the person who buys the supplies or does the topping up. Ask directly and early: "who usually handles this one?" On a commercial site the answer is often a maintenance person, a cleaner, or a delivery driver, none of whom will be in the room.

On a rental or managed property, add the tenant and the previous management company. The most common source of an undocumented change is a handover nobody wrote down.

Step 8: Corroborate with something that is not testimony

Testimony alone should not carry a diagnosis you are going to bill against. Look for one non-testimonial confirmation:

• The photographed label itself, checked against the specification.
• A receipt, a delivery note, or an order history the customer can search.
• A different residue color, level, or consistency visible at a fill point in a photo.
• A prior job record from your own shop showing what was installed and when.
• Runtime hours or a cycle count that fits your latency estimate.

One of these plus a consistent interview is a defensible conclusion. An interview alone is a probable cause, and it should be written up with that word.

Worked example: a window that changed the answer

Remote-only situation. Equipment at a property the owner visits a few times a year. The complaint is progressive: works fine at first, degrades through a long run, recovers overnight. The mechanism reads as accumulation, and on this equipment class accumulation of the kind described typically needs several hundred running hours to become symptomatic. Call it 400 running hours as the working estimate.

Duty: the owner says it runs about 2 hours per day. So 400 running hours is about 200 days, which is roughly 6 to 7 months back. That is the window.

The first call had already produced "no, nothing has changed" in response to an open question. The second call asks a dated question instead: "Sometime around six or seven months ago, so before it got warm, did anybody deliver, refill, or top this up while you were not there?"

The answer changes immediately. There was a delivery while the property was empty, and it came from a different supplier because the usual one could not schedule it.

Requested look: is there a container, a delivery note, or anything with printing on it? There is. Photo of the label arrives. The grade on the label does not match the specification for that equipment.

The timing now closes in both directions. The delivery sits inside the estimated window, and the estimated window came from a latency calculation made before anyone knew a delivery had happened. That independence is what makes it convincing rather than circular, and it is worth saying so in the write-up: the window was predicted from the symptom, then the event was found inside it.

Conclusion recorded as confirmed, with the label photo, the delivery date from the note, and the latency reasoning attached. No site visit, no component replaced, and the corrective action is a drain, a refill to specification, and a written specification sent to both the owner and the supplier.

What would have kept this at probable: if the container had been gone and the only support were the owner's recollection of a delivery, the write-up says probable, and the recommendation includes a sample at the next physical access.

Verify you got this right

• Does the dated event sit inside the window you calculated before you found it? Calculating the window after learning the date is not evidence, it is fitting.
• Do you have one non-testimonial item? Photo, receipt, delivery note, or your own prior record.
• Would the customer describe your questions as curious rather than accusing? If the interview ended tense, you will not get the follow-up cooperation the fix requires.

What changes the answer

• The equipment is new or recently serviced by you. Your own records replace most of the interview, and the question shifts to what happened between your visits.
• The customer is a managed property or commercial site. Interview the person who touches it, not the person who called, and expect documentation to exist somewhere in a system nobody has checked.
• The customer has an incentive in the outcome. A pending warranty or insurance claim changes how you weigh testimony. Lean harder on non-testimonial corroboration and say in the write-up which items are which.

The failure modes

Asking the open question first. Once a customer has said nothing changed, they are committed to it, and every later question sounds like you are calling them a liar.

Stacking instructions. Three requests in one message reliably produces one answer and no way to know which of the three you got.

Accepting a spoken label. The front of the container is a name, not a grade. Get the photo.

Diagnosing the person. Deciding early that the customer is hiding something makes the interview adversarial, and the usual truth is far more boring: somebody bought a thing that looked the same.

Writing testimony up as confirmation. If your evidence is one conversation, the record should say so. Overstated confidence is what loses the dispute months later.

References

• Manufacturer documentation for consumable specifications and supply requirements
• Trade-standard practice on limits of remote guidance to untrained persons on energized, pressurized, or fuel-carrying equipment
• OSHA general guidance on hazard recognition for occupants and non-qualified persons
• See related: Why Off-Spec Supplies Fail Slowly Instead of Immediately; The Consumable Verification SOP; How to Explain a Supply-Caused Failure Without Blaming the Customer