This is where most of the value in an existing supplier base gets unlocked. Import gets your suppliers into Speya, but research is what actually answers a question, one you had already or one you didn't know to ask, across every supplier at once instead of one at a time. The same approach works whether you're running a quick sanity check on a handful of suppliers or a full compliance sweep across a base of thousands, and it's the same skill either way: describe what you want clearly enough that Speya can act on it, then react to what comes back.
If your suppliers aren't in a list yet, see Getting started: working with your existing supplier base for how to bring them in.
In this article
What other teams research
How it works
Define what you want
Set the boundaries
The shape of the answer
Evidence strength
Refine after results
Prompt library
FAQ
What other teams research
Existing-base work spans a wide range of topics. You can find answers to almost any question. The only limitation is whether the information is public.
The most common research areas among professionals normally include:
Ownership, board, and major investors
Sanctions and export control status
Credit rating and financial health
Sustainability certifications and science-based targets
Human rights and labor practices
Legal issues and regulatory actions
Adverse media and controversies
Geopolitical and country exposure
The rest of this article is about how to turn any topic like these into a well-defined column.
How it works
Research in Speya works the same way regardless of the topic:
Write down what you want to know and the purpose, in plain language, as a message to Speya.
Send it.
Speya interprets it and summarizes the column back to you. What you see in chat is a short summary of its plan, not the actual instruction. The full instruction it will run when conducting the research across the +250 databases is more detailed than the chat summary.
Accept or adjust the summary. Run the research directly if you are satisfied with the initial direction. Adjust the research direction by providing more context to Speya about what you want.
Speya runs the research, applying the full instruction in the background across every supplier in your list.
Open a supplier's cell to see everything together: the actual instruction that ran, Speya's reasoning, the evidence it found, and the result.
Read the results, then iterate. Sharpen the instruction, add a related column, or ask a follow-up question.
Define what you want
A good description usually settles a handful of things. You don't need all of them for every question, but knowing they exist means you can add whichever frames yours most effectively.
The topic
Name the topic. It doesn't need to be precise.
"Look at their exposure to external cost pressures." Speya works out that this usually covers raw materials, energy, labor, currency, and freight, and tells you which apply to each supplier.
Why it matters
Say what the answer is for. This gives Speya something to reason toward, beyond the topic itself.
"I want to understand financial stability before we commit to more volume."
Set the boundaries
This is really one decision with multiple faces, and they're worth treating as one thing rather than separate ideas. All answer the same underlying question, where should Speya be looking?
Decide | Say something like | What it does |
Who it's about | "For all the companies" "... and the board plus directors." "... for operational entities" | Sets the scope. Works at any level, and applies just as well to customers as to suppliers. |
What's relevant within that | "Include all the operational facilities they hold, but focus on the ones relevant to us, [your company name or category]." | Casts the net wide first, then narrows to what's actually relevant, so nothing worth seeing gets missed by narrowing too early. |
What counts as a match | "Issues" instead of "violations." "Legal issues since 2022." Naming tier vs. scope where either could apply. | Sets the threshold for what counts as a finding. See below for more on this one. |
Or ask Speya to define it | "What else should we be watching for external risk to our suppliers?" "That framing feels too generic, sharpen it." | Speya proposes the boundary itself, rather than you supplying one. |
What counts as a match. The specific words you choose set the threshold. A few patterns worth knowing:
How wide a net to cast. "Violations" returns very little, since a violation usually means something reached a court or a regulator. "Issues" casts a much wider net: allegations, NGO reports, community complaints.
What time period. "Outstanding legal issues", "any legal issues", and "legal issues since 2022" each return a different set. Name the period if it matters.
Which framework you mean. Supply chain tiers (1, 2, 3) and emission scopes (1, 2, 3) use the same numbers for unrelated ideas. Say which one you mean if there's any chance of confusion.
If you're not sure which threshold you want, run the same question two ways and compare the results directly.
The shape of the answer
Decide the shape of the answer. This is often the single biggest lever on how useful a result is.
Descriptive research, when the topic needs explaining, not just stating. "How exposed is this company to disruption in the regions where it operates? Explain your reasoning."
A strict verification, when you need a clean gate. "Is this company on a sanctions list? Answer Yes, No, or Unknown." For anything that needs to hold up as a verified fact, such as a certificate, ask for the specifics that make it checkable: the certificate or reference number, the issue and expiry dates, and the current status, or Unknown if none of that can be confirmed.
A scale, when you want to compare suppliers at a glance. "Rate environmental risk as Low, Medium, or High, with a short justification." Adverse media works well either as a strict Yes/No or as a scale, depending on how much nuance you want: a straight "any adverse media in the last three years, Yes or No", or a fuller "overall sentiment: Positive, Mixed, or Negative".
A trend, when the direction matters as much as the current state. "Show the credit rating for each of the last five years, and state whether the trend is Improving, Stable, Declining, or Volatile."
Format | Example prompt | Use it when |
A strict verification | "Is this company on a sanctions list? Answer Yes, No, or Unknown." | You need a clean gate. For anything that must hold up as a verified fact, such as a certificate, ask for the specifics that make it checkable: reference number, issue and expiry dates, status. Reach for a Registry when the fact needs to come directly from the source. |
Descriptive research | "How exposed is this company to disruption in the regions where it operates?" | The topic needs explaining, not just stating. |
A scale | "Rate environmental risk as Low, Medium, or High, with a short justification." | You want to compare suppliers at a glance. Adverse media works well either as a strict Yes/No or as a scale (Positive/Mixed/Negative), depending on how much nuance you want. |
A trend | "Show the credit rating for each of the last five years, and state the direction of the trend as Improving, Stable, Declining, or Volatile." | The direction matters as much as the current state. Separate current state and trend with two distinct columns. |
Evidence strength
Speya surfaces conflicting sources and thin evidence rather than smoothing them into a confident-sounding answer. That's deliberate: qualification and compliance work is really a question of where to look first, and evidence strength answers that before you open a single source.
Weaker evidence shows up directly in the worksheet, as a red dotted line under the text in the cell, so you can spot where to look closer just by scanning the list, without opening a single cell first.
Alongside every answer Speya returns, Confidence scores attaches an evidence strength. Strong answers can be accepted as they are. Thin ones become your review list. The practical outcome is less time spent across the whole process, since effort goes where the evidence is weak instead of being spread evenly across everything.
Direct means direct, current, authoritative evidence supports this item. This is the strongest level.
Strong means the sources agree, and the answer is well supported.
Mixed means the sources say different things. There's evidence, it just doesn't line up cleanly.
Limited means there's little or no supporting evidence behind the answer, and it needs human review. This is a statement about the state of the research, not about the supplier: read it as not established either way, not as a hidden No.
Refine after results
Here, you've already got a result and you're reacting to it: something is right but hard to read, too broad, built on a concept that doesn't quite fit, or missing a dimension you now realize you need.
Not every reaction should change the existing column. If what you actually want is a different framing of the same topic, a number where you have prose, a date where you have a level, a trend where you have a snapshot, the better move is often a new column alongside it, not a rewrite of the one you have.
A single column that tries to hold a description, a score, a date, and a trend all at once usually ends up worse at all four than four columns would be at one each.
When you need it | Example prompt | What it does |
Splitting into a companion column | "Keep this column as the written explanation, and add a separate column that just gives the certificate expiry date so I can sort by it." | Two views of the same fact, each doing one job well, instead of one column trying to do everything. |
Adding a trend companion | "Add a second column showing how this risk level has moved over the last three years." | Keeps the original column intact for current state and adds the change-over-time view alongside it. |
Tightening scope | "Narrow this to the production sites relevant to what we buy from them, not the company's full global footprint." | Refocuses the research without changing the underlying topic. |
Rejecting a framing | "Market entry barriers feels too generic here, sharpen it to something we can actually assess for [category]." | Speya proposes the replacement rather than you having to supply it, pushing it to find the specific dimensions that are relevant for your category or business model. |
Reordering | "Put the overall level first, then the detail." | Restructures the column so the conclusion leads and the reasoning follows. |
Debugging a disagreement | "Why do the country risk and regulatory exposure columns disagree for [supplier]? And, what's the difference between them?" | Usually reveals that the two columns are answering slightly different questions. Speya suggests altering one of them or both, to support your research goal better. |
Prompt library
Every prompt here is something you type to Speya, grouped by what it's for.
Deciding what to look at
"I work in [role]. What should we be tracking about our suppliers that we haven't thought of?" Useful when you know the general area you're responsible for but haven't yet broken it into specific questions. Speya proposes a starting framework based on your role, which you can then accept, trim, or extend.
"Is [concept] a standard thing to look at for this kind of review?" Useful when you're not sure if an idea is recognized practice or something specific to your situation.
Starter columns by topic
Topic | Prompt | Why it's useful |
Ownership | "Who owns the company, who's on the board, and who are the major investors?" | Useful before onboarding, and worth re-running periodically since ownership changes can quietly change the nature of a relationship you already have. |
Sanctions | "Screen each supplier against all relevant sanctions lists. Answer Yes, No or Unknown." | A hard gate rather than a judgment call, usually the first column any compliance-driven base gets, since a Yes here typically blocks the relationship outright. |
Credit rating | "Reconstruct each supplier's credit rating over the last five years and state whether the trend is Improving, Stable, Declining, or Volatile." | Financial health is rarely useful as a single snapshot, since one bad year says less than a clear multi-year direction. |
Sustainability disclosure | "Does each supplier publish a sustainability or ESG report? Answer Yes or No, with the report link if available." | A simple disclosure check that's often the entry point into a wider sustainability review, since the report itself becomes a source for follow-up questions. |
Certification validity | "Is each supplier's [certification] still valid? Give the certificate number, issue and expiry dates, and status, or Unknown if it can't be confirmed." | Certifications lapse quietly, so asking for the specifics rather than a plain Yes or No means the answer is checkable later, and ready to pair with a monitor. |
Legal issues | "Research any legal issues for each supplier, including investigations and allegations, not only convictions." | Deliberately wider than "violations," this catches what's been raised even if nothing's been formally decided, which matters most for early risk signals. |
Adverse media | "Check each supplier for adverse media in the last three years and give an overall sentiment: Positive, Mixed, or Negative." | Surfaces controversies you'd otherwise only catch by chance. Ask for a plain Yes or No instead if you only need to know whether anything exists at all. |
Geopolitical exposure | "Which suppliers are exposed to geopolitical or economic disruption that could affect supply?" | Broad by design, usually a first pass you narrow afterwards to the specific countries or routes relevant to what you actually buy. |
Making it trackable
"Give me a level so we can monitor how this changes over time." Signals that you want a consistent, comparable category rather than a one-off paragraph.
"Add a second column showing the direction this has moved in the last few years." A trend companion to an existing snapshot column.
"Use the same basis for every supplier so I can compare them." Prioritises consistency across the list over the richest possible answer for any single company.
FAQ
Do I need to specify which sources Speya should check?
No. Say what you need the answer to do, and Speya works out where to look.
My first result came back too vague. Did I do something wrong?
No. That's a normal first pass. Read what came back, say what's missing, and ask again.
I asked for violations and got almost nothing. Does that mean the supplier is clean?
Not necessarily. Ask for issues, allegations, or controversies as well to see what's been raised without reaching a formal outcome.
Can I ask about one supplier, one category, or a customer the same way?
Yes. The method doesn't change with scope or with the type of company.
How do I know if an answer is a verified fact or Speya's reasoning?
Research columns reason from public sources and can read slightly differently between runs. For anything that needs to be exact, use a connected registry instead.
I want to compare suppliers on the same measure. How do I make sure the answers line up? Ask for it directly: tell Speya to use the same basis for every supplier and return a fixed set of values rather than free text.
Does "tier 1, 2, 3" mean the same as "scope 1, 2, 3"?
No. Tiers describe how far a sub-supplier sits from you. Scopes describe where a company's emissions come from. Name which one you mean.
What's the difference between Limited evidence strength and a No data found result?
They're two separate outcomes, not two points on one scale. No data found is a system state, it means Speya couldn't produce an answer at all. Limited is an evidence strength attached to an answer Speya did produce, meaning what's behind it is thin and it needs human review, not that the answer is negative.
Why does a result say "No data found"?
It means Speya found no data addressing the question for that supplier, not that the answer is negative. Use Research missing to try again, or ask Speya to reason about likely answers where nothing is published.