How do you find a root cause when the customer gives no data?

This method covers a complaint about a delivered chemical product where the customer cannot or will not share process data, samples or photos. It does not cover a complaint involving injury or a regulatory report, where the investigation rules of that jurisdiction set what must be collected.

What do you need before looking for a root cause with almost no customer data?

Before looking for a root cause with almost no customer data, you need your own records for the complained lot, a dated list of hypotheses and a written list of the facts the customer has not given. Your side usually holds more than the complaint does: the certificate of analysis (CoA), the batch record, a retained sample, the technical data sheet (TDS), the SDS and the shipping record. Write the missing facts as a list of questions, one per line: application temperature, mix or dilution ratio, substrate, storage conditions, open date, and when the problem first appeared. Step 1 of the method is this list, because a gap you have not written down cannot be filled or reported. Step 2 is asking the customer once, with the shortest list that would split your hypotheses. Rank the questions by how many hypotheses each answer would rule out, and send the top few.

Which public sources can stand in for missing customer data in a chemical complaint?

The public sources that can stand in for missing customer data in a chemical complaint are substance databases, the supplier's own SDS and TDS, and public records about the conditions the customer worked in, each used only for what it can decide. Step 3 of the method uses them with labels. A substitute value never becomes a fact about the customer's line; it narrows the hypothesis list or sets a bound, and the label travels with it into the verdict. When two public sources give different values for the same substance property, record both with their sources rather than picking the one that fits the favored cause. The table below shows which public or typical values may stand in for absent customer data and which verdicts they cannot support.

The key row in the substitute source table is the TDS row: a typical value tells you what the product normally does, and only the CoA or a retest of the retained sample tells you what this lot did.

How do you use typical values in a root cause analysis without treating them as this product's values?

Use typical values in a root cause analysis without treating them as this product's values by writing the word "typical" next to every one, keeping the document and revision it came from, and never using it to rule out a lot defect. A TDS describes the product in general; the CoA reports results measured on one lot against its specification. When the two are confused, an investigation can wrongly conclude that a lot was in specification because the TDS says so. Read the TDS's own note on its values: if it says they are typical and not specifications, they cannot stand in for the lot. Substance values need the same care. An ECHA CHEM entry describes a registered substance, not the mixture you sold, and ECHA says of registration data that it "does not verify the information before dissemination" (ECHA, 2024). PubChem's Substance collection "archives chemical descriptions submitted by individual data depositors" (Kim et al., 2025), so name the depositor next to the value. A substance value can still be useful: a boiling point far above the use temperature makes loss by evaporation unlikely, which narrows the list without confirming any cause.

How do you record a root cause value you could not confirm?

Record a root cause value you could not confirm by writing four things next to it: the value, its label, the source and date, and the record that would settle it. Step 4 of the method is this record. Use a short, fixed set of labels so a reader can tell at once how strong each input is: measured on the complained lot, measured on the retained sample, supplier typical value, public substance value, estimate with a stated basis, and unconfirmed. An estimate with a stated basis names its reasoning in the same line, for example "line temperature estimated at 12 °C from the regional weather record for that date" (illustrative). An unconfirmed value is left blank with the reason, such as "customer did not answer the storage question on the date asked" (illustrative). Regulated complaint files already work this way for missing facts. For US drug products, 21 CFR 211.198 lists what the record must include "where known" and, where no investigation is conducted, requires "the reason that an investigation was found not to be necessary" (eCFR, 2026). Never upgrade a label later without a new record, and never delete a blank: the blank is what tells the next reader which data would change the verdict.

When is unconfirmed the right root cause verdict?

Unconfirmed is the right root cause verdict when the only check that could test a hypothesis needs a record you cannot get, and the substitute sources can bound the hypothesis but not decide it. Three cases are common in chemical complaints with little customer data. First, the deciding fact sits with the customer, such as storage temperature or mix ratio, and the one question went unanswered. Second, the only evidence is a typical or public value, which describes the product or substance in general, not this lot on that line. Third, two hypotheses predict the same observation and no available record splits them. HSE guidance on investigations notes that "there may be a lack of information and many uncertainties" and asks investigators to keep an open mind (HSE, 2004). An unconfirmed verdict is not a failed analysis. Unconfirmed differs from ruled out: a hypothesis is ruled out only when a test that could have shown it true came back against it. Write each unconfirmed verdict with the record that would settle it and who holds it, so the customer reply can say exactly what data would reopen the case.

How do you check a root cause verdict built on little customer data?

Check a root cause verdict built on little customer data by confirming that no verdict is stronger than its weakest labelled input, and that every unconfirmed verdict names the record that would settle it. Run five checks before the reply goes out.

1. Each confirmed verdict rests on at least one value measured on the complained lot or its retained sample, not on a typical or public value.

2. Each ruled-out verdict came from a check that could have shown the hypothesis true, which is what makes it a test rather than an assumption; the Stanford Encyclopedia of Philosophy summarizes Popper's view that support counts only from a prediction "which might conceivably have been false" (Stanford Encyclopedia of Philosophy, 2026).

3. Each public value carries its source and label in the record.

4. Each estimate states its basis in the same line.

5. A second reader can trace every verdict to a cited record without asking the analyst.

If check 1 fails, downgrade the verdict to unconfirmed. Send the reply with the open questions in it, tied to the verdict each answer would change.

When should you call a specialist about a root cause with little customer data?

Call a specialist about a root cause with little customer data when two hypotheses stay unconfirmed and only a test on the customer's material could split them, when the complaint involves injury or a possible regulatory report, or when public substance values conflict. An analytical lab can compare a sample of the customer's remaining material with your retained sample under a method agreed before testing, which often replaces the missing process data. When ECHA CHEM and PubChem give different values for the same property, a product stewardship or toxicology specialist can judge which applies to the grade you sold. Where the complaint may carry a reporting duty, the regulatory owner for that jurisdiction decides what must be collected and by when. Lubinpla works root cause cases in this order: it checks your own records first, then public sources on the web, asks you once for what is still missing, and if a fact is still absent completes the verdict with a stated-basis estimate or marks it unconfirmed, with your approval at each step. Bring the labelled value record and the list of unanswered questions to any specialist.