Vendor selection

You help the owner choose a supplier they will not regret in a year. That means agreeing what is needed before looking at anyone's sales pitch, comparing vendors on the same evidence, counting the full cost over the life of the contract, checking that the vendor is sound and safe to trust with the business's data and customers, and handing the owner a clear recommendation with the trade-offs named. You never sign, accept terms, place orders or commit spend; the owner does, after a qualified person reviews the contract where it matters.

When to use

• "We need a new accounting / CRM / payroll / logistics provider."
• "Find us three cleaning companies and compare them."
• "Is there a better deal than our current supplier?"
• "Which of these two quotes should we take?"
• A contract renewal is coming up and the owner wants to test the market.

What you need

• The business need, the problem with the current situation, budget, timeline, and who will use or be affected by the service.
• Requirements from the people who will use it (ask_teammate, brief_team).
• Current costs and contract (if replacing), and data on current volumes and usage.
• Research access: web_search and web_extract for the market; your own browser for vendor sites and trial accounts if the owner wants trials (sign-ups use the owner's details only with their go-ahead).
• Who will review the contract (owner, solicitor, accountant, security adviser) and who approves the spend.

Method

1. Write requirements before meeting vendors. Group as Must (fail any, the vendor is out), Should, Could, Won't (MoSCoW). Include volumes, integrations, service levels, security and data protection needs, contract length preferences, budget range. Get the owner to approve them. Template in references/selection-templates.md.
2. Long list (6 to 10) from web_search, industry directories, review sites, and the owner's network. Screen against the musts using public information. Note how you found each one.
3. Short list (3 to 4). Send a request for information or quote (RFI/RFQ) built from the requirements, with a deadline and a pricing template so quotes are comparable. You draft it; the owner approves and sends it, or you send it from the agreed mailbox only with their go-ahead.
4. Scripted demos or trials. Give every vendor the same 3 to 5 real scenarios from the business ("enter this order with these exceptions"). Score what you see, not the slides. Note who attended and what was promised in writing.
5. Fix the weights before scoring. Criteria and weights (summing to 100) are agreed with the owner before any vendor is scored. Score 1 to 5 against written anchors (5 = fully meets with evidence; 3 = meets with workarounds; 1 = does not meet). Where possible two people score independently, then discuss differences above one point.
6. Total cost of ownership over three years (or the contract term): subscription or unit prices with the stated annual increases, implementation, migration, training, internal staff time, add-ons you will realistically need, usage overages at your forecast volumes, and the cost of leaving (data export, termination fees, parallel running). Price the same scope for every vendor.
7. Due diligence on the leader and runner-up:
   • References: two or three current customers of similar size, called by the owner with your question list.
   • Financial stability: public filings where available (for example the national company register), years trading, ownership.
   • Security and data protection, where they will hold the business's or customers' data: independent assurance reports or certifications (such as SOC 2 or ISO 27001) and a data processing agreement. Have a qualified person review anything sensitive.
   • Insurance certificates where relevant (liability, professional indemnity).
8. Contract points. List what to negotiate and what needs a qualified review: term and renewal, notice period, price increase caps, service levels and credits, liability caps, data ownership and export, termination for convenience, assignment, governing law. You flag; a solicitor or the owner decides. Do not interpret legal clauses beyond summarising what they say, with clause numbers.
9. Build the comparison workbook with execute_code (scoring with SUMPRODUCT, TCO by year, summary with cost per score point) and tie it out against a Python recompute (code in the reference). Check weights total 100.
10. Recommend in a short memo: the recommended vendor and why, the runner-up and when it would be the better choice, the score and cost table, the key risks and how the contract should address them, negotiation asks, and the decision needed. The owner decides and signs.

Worked example: choosing an order system

Requirements approved by the owner: 5 musts (600 orders a month, 14 order types, export to the accounting software, a customer portal, data export on exit), 4 shoulds, 2 coulds. Long list of 8 from web_search, two directories and two recommendations; 3 failed a must on public information. The short list of 3 received the same RFQ with a pricing template, and all three ran the same three scripted scenarios in their demos. Weights fixed before any scoring: must-have fit 30, ease of use 15, integration 15, support 10, security 10, stability and references 10, contract flexibility and exit 10.

	Vendor A	Vendor B	Vendor C
Weighted score (of 5)	3.95	3.90	4.05
3-year TCO	57,330	42,180	57,700
Cost per score point	14,514	10,815	14,247
Key risk	Weak exit terms	Limited integration	Smaller support team
Due diligence on the leader and the runner-up:

• Two references each for B and C, called by the owner with the question list in the reference.
• Company register checked for both; ownership and years trading noted.
• C holds an independent security assurance report; B does not yet. Recommendation: C, because integration and exit terms matter more to this business than 15,000 over three years. B is the runner-up if cost becomes the deciding factor. Negotiate with C: cap annual increases at 3%, a fixed-fee implementation, data export in CSV at no charge. For qualified review: the liability cap and the data processing agreement. The workbook and the memo outline are in references/selection-templates.md.

Scripting a demo

Give every vendor the same three to five scenarios from the business's real work, with the data to use: "Enter order 1182 (14 lines, two on back-order), amend the delivery address after dispatch, and produce the invoice." Score what you see happen, not what is promised; note anything that needed a workaround; ask for every promise in writing after the session. A vendor that will not run your scenario is telling you something.

Total cost of ownership: the lines people miss

• Annual price increases written into the contract.
• Add-ons the business will realistically need in year 2.
• Overage charges at the forecast volume rather than today's.
• Internal staff time for implementation and testing.
• Parallel running of the old system.
• Training for joiners, not just the launch team.
• The cost of leaving: export, termination fees, re-implementation. Price the same scope for every vendor over the same term, and show year 1 and the 3-year total side by side.

Output

• vendor-comparison-<topic>.xlsx: Requirements, Long list, Scoring, TCO, Summary, Due diligence, Contract points.
• Recommendation memo (1 to 2 pages).
• Drafts: RFQ, demo script, reference-call questions, regret letters for unsuccessful vendors (sent only with the owner's go-ahead).
• A show_card with the summary table: vendor, weighted score, 3-year TCO, key risk.

Checks before you finish

• Requirements and weights were approved before scoring; weights total 100.
• Every vendor was scored on the same scenarios and the same pricing scope.
• TCO includes implementation, internal time, increases and exit costs.
• The workbook formulas match the Python recompute.
• Every claim a vendor made that matters to the decision is in writing.
• Contract points needing qualified review are flagged, not resolved by you.

Pitfalls

• Choosing on the demo. Polished demos of features you will not use. Script them.
• Changing weights after scoring. It quietly makes the favourite win.
• Licence price as the cost. Implementation, internal time and exit often double it.
• Ignoring lock-in. Ask how you get your data out, in what format, and at what cost.
• Auto-renewal and price escalators buried in the terms.
• One reference, chosen by the vendor. Ask for customers of your size and talk to them.
• Letting a vendor's email set the bank details or the process. Supplier onboarding and bank details follow the owner's verification process.

See also

• contract-renewal-tracking, purchase-requests, supplier-payment-run (finance).