# Software selection: how to choose a vendor without regret

> Many regretted technology choices were made early: a vague need, a favourite vendor, criteria set after the answers and a price that left out half the cost. This guide gives you a fair, defensible way to choose: five stages, eight rules, a simple scoring model and the checks that turn claims into evidence.

Source: https://ciohelm.com/guides/software-selection · By the CIO Helm team · Updated 2026-11-13

## Where regretted choices are made

Many regretted technology choices were made early, long before the contract: a need described as a list of features, a favourite vendor everyone knew about, criteria written after the answers arrived, a demonstration of the vendor's best story instead of your own processes, and a price that left out half of the five-year cost. The result is a platform that fits the presentation better than the business, and a selection that cannot be defended when it is challenged.

A fair selection is not slower. It moves faster because the arguments happen once, at the start, instead of at every step.

## Five stages, five gates

Each stage answers one question, and ends with a gate where a named group decides whether to continue.

| Stage | The question | Gate decision |
| --- | --- | --- |
| **1. Frame** | What problem are we solving, and is buying the right answer? | Go to market: approve the need, scope, budget envelope and team |
| **2. Scan** | Who could meet the need, and who clearly cannot? | Shortlist: approve the shortlist and the evaluation model |
| **3. Compete** | Which offers are best, on the same evidence, judged the same way? | Finalists: choose two finalists and approve the proof of concept |
| **4. Prove** | Does it work for us, with our data, and can we trust the company behind it? | Preferred supplier: name it and the fallback; open the negotiation |
| **5. Commit** | Is this the right deal, and are we ready to start? | Award: approve the award, the contract and the budget |

Scale the method to the decision. A small, low-risk purchase in a well-known market can skip the request for information and the proof of concept and take four to eight weeks. A core platform used by many people deserves the full method, an executive sponsor at every gate and four to six months.

## Frame the need before you meet a vendor

- **Write the need as outcomes** with today's baseline: "invoice processing time from twelve days to three", not "OCR module".
- **Map how the work is done today**, with the people who do it. Pain points become your demonstration scenarios.
- **Check what you already own.** Many selections end with a platform that duplicates one already licensed.
- **Set the budget envelope** before talking to the market, so the envelope shapes the scope and not the other way round.
- **Declare interests.** Every member of the selection team declares relationships with vendors, in writing, at the start.

## Eight rules that keep the choice fair

| Rule | What it means in practice |
| --- | --- |
| **Decide how you will judge before you see the answers** | Criteria, weights and scoring scales are approved and locked before the request for proposal goes out. Any later change is logged and applied to every vendor. |
| **Keep price sealed until quality is scored** | Prices arrive in a separate file, opened only after the quality scores are agreed. |
| **A must is a must** | A vendor that cannot meet a must-have requirement today is out, however good the rest. Keep the musts few so that the rule can be applied. |
| **Score alone, then agree** | Each evaluator scores alone against anchored scales; large differences are discussed and the agreed score is recorded with its reason. |
| **Show, then prove** | Claims become evidence in two steps: a scripted demonstration of your scenarios, then a proof of concept on your data with your users. |
| **Count the whole cost** | Compare five-year total cost, not year-one price. |
| **Plan the exit before you sign** | Negotiate data return, transition help and termination rights while there is still competition. |
| **Same rules for every vendor** | One point of contact, the same information for all, questions and answers shared, and a decision log that would stand up to an audit. |

## Criteria and weights

Keep the model simple enough to explain to the board in one slide. An illustrative split for a business application:

| Criterion | Weight | What it covers |
| --- | --- | --- |
| Functional fit | 35% | How well the solution supports the scenarios and requirements, shown, not claimed |
| User experience | 15% | Ease of use for the people who will use it every day, judged by them |
| Technology and integration | 15% | Architecture, integration with your systems, security, data |
| Implementation and support | 15% | The partner, the approach, the plan, references for similar work |
| Vendor viability and fit | 10% | Financial health, roadmap, presence in your regions, contract flexibility |
| Five-year cost | 10% | Scored from the sealed price file, after quality. Many organisations weight cost at 10% to 30%, depending on the market and their procurement policy |

Anchor each score: write what a 1, a 3 and a 5 look like for each criterion before scoring starts. Without anchors, a 4 from one evaluator means a 2 from another. And test the result: if a small change of weights changes the winner, the choice is closer than the total suggests, and the decision paper should say so.

## Demonstrations on your script, with your data

Vendors demonstrate their best story unless you give them yours. Send every shortlisted vendor the same script two weeks ahead: five to eight scenarios drawn from your pain points, with sample data, a fixed time per scenario, and your users in the room scoring as they watch. Ask to see the configuration behind a screen, not just the screen. Then take two finalists further: deeper demonstrations and, for a major platform, a proof of concept on your own data, with pass or fail tests agreed beforehand.

## The five-year cost

Year-one price is the most misleading number in a selection. Compare every finalist on the same five-year view:

- Subscriptions or licences, with the annual increases in the contract, and the add-ons needed to meet your musts.
- Implementation, integration and data migration, from the partner and from your own teams.
- Training, change, and the internal time of the people seconded to the project.
- Running costs: support, hosting, administration, upgrades.
- Exit costs: data extraction and transition at the end of the contract.

## Due diligence, procurement and legal

- **Bring procurement and legal in at stage 1**, not at the contract. Public bodies and some regulated companies must follow procurement rules that fix parts of the process.
- **Check the evidence, not the brochure**: security certifications and test reports, where data is held and which sub-processors touch it, how AI is used inside the product, resilience and recovery commitments, and the vendor's financial health.
- **Cap what can change**: renewal price increases, notice periods, and what happens to your data and the service if the vendor is sold.
- **Choose the implementation partner as carefully as the software.** They are often separate choices, and the partner decides more of the outcome than the product.

## Reference calls that reveal problems

Choose the references yourself where you can: a customer of similar size, industry and scope, ideally one that went live in the last two years. Then ask questions that cannot be answered with a slogan: *What surprised you after go-live? What would you negotiate differently? How many of the people who sold it to you are still on your account? What did the implementation really cost against the plan? Would you choose them again?*

### Run your next selection with the Solution Selection Kit

A 22-sheet selection workbook (requirements, evaluation model with anchored scales, knock-outs, panel scoring with calibration, demonstrations, proof of concept, references, due diligence and five-year cost), the vendor response and sealed pricing workbooks, four decks (framework overview, vendor briefing, finalists and award recommendation), eleven templates from the charter to the award memo, and a guide. One-time price , no subscription.

[Download the kit](https://ciohelm.com/download/solution-selection-kit)[See what is inside](https://ciohelm.com/toolkits/solution-selection-kit)

## Questions

**How long should a software selection take?**

Four to eight weeks for a small, low-risk purchase; around three months for a standard selection; four to six months for a core platform with a proof of concept. Most of the delay in real selections comes from criteria agreed late and decision makers who were not involved early.

**How many vendors should be on the shortlist?**

Three to five for the request for proposal, and two finalists for demonstrations in depth or a proof of concept. A longer shortlist costs vendors and evaluators time without adding information.

**Should we always run an RFP?**

No. For small, well-understood purchases, a short quote request and a demonstration are enough. Keep the principles (criteria before answers, the same rules for every vendor, five-year cost) whatever the size.

**What if the business already has a favourite vendor?**

Include it, and run the process fairly. A favourite that wins a fair selection is a stronger choice, and one that loses has told you something important before you signed.

## Toolkits on this topic

- [Build vs Buy Kit](https://ciohelm.com/toolkits/build-vs-buy-kit): a build or buy decision that holds
- [Vendor Negotiation Playbook](https://ciohelm.com/toolkits/vendor-negotiation-playbook): negotiate with the vendors you already have
- [GCC Partner RFP Pack](https://ciohelm.com/toolkits/gcc-partner-rfp-pack): choose a build, operate, transfer partner for your capability centre

## Related

- [Technology business case: a structure finance will recognise](https://ciohelm.com/guides/technology-business-case)
- [IT steering committee: terms of reference that make it decide](https://ciohelm.com/guides/it-steering-committee)
- [IT strategy on a page: template and example](https://ciohelm.com/guides/it-strategy-on-a-page)
- [Project Delivery Framework Kit: from approval to benefits](https://ciohelm.com/toolkits/project-delivery-framework-kit)
