Begin with the problem, not a vendor list
The most expensive way to select a development partner is to start with a spreadsheet of agencies and work backwards to a justification. Start instead with the performance problem your learning is meant to solve, and write that down in one paragraph. What behaviour should change, for which population, over what period of time, and what business measure will move if it does?
This paragraph becomes the spine of every later conversation, because it is the only part of the process that is genuinely under your control. Vendors will change.
Tooling will change. What will not change is the gap you are trying to close, and a supplier who cannot connect their proposed solution back to that paragraph is describing a catalogue, not a plan.
If the gap is genuinely unclear, an assessment or front-end needs analysis is usually a better first purchase than a full development cycle. It is also a useful litmus test: agencies that are willing to scope carefully before quoting tend to be more disciplined about delivery than those that price optimistically and negotiate upward later.
Define what custom actually means in your context
Custom is a spectrum, not a binary. One supplier may be authoring a genuinely bespoke course from a blank brief, while another is assembling an off-the-shelf content library into a branded shell with a new assessment and a localised scenario. Both can legitimately be described as custom, and the price difference between them is substantial.
Decide up front where you want custom depth in each deliverable. Custom instructional design and original scenarios are expensive.
Custom visuals, branding, and a refreshed look and feel are moderate. A custom deployment layer for your LMS, SSO, and reporting is a technical cost that is often worth paying separately rather than folding into a content rate.
Being explicit here protects the budget. When a supplier later says a requested feature is out of scope, the conversation is about a decision you already made in writing, rather than a surprise that erodes trust.
Separate capability claims from demonstrated work
Every supplier website claims expertise in learning design, accessibility, and technology standards. The evidence that matters is narrower: a completed project, in a comparable subject area, delivered under constraints similar to yours, and identifiable by a client who will speak to you without a script.
Ask for two or three specific references and actually call them. Ask what the project scope was, what went wrong, how the supplier responded, what the vendor was like to work with weekly, and whether they would use them again. Reference calls that only confirm what the pitch deck promised are a sign the pitch deck is all you will get.
Portfolio pieces also need a closer look. Ask whether a shown project was developed in house, co-developed, or was a largely off-the-shelf product rebranded. A supplier that answers this question plainly is more credible than one that presents every artefact as a bespoke original.
Judge the team, not just the company
Custom eLearning is delivered by individuals, and the individuals named on the proposal are not always the individuals who do the work. Ask who the instructional designer, project manager, media producer, and technical developer will be, and whether they will remain on the project for its duration.
Sample work is the most reliable way to test a candidate team. Ask to see a short sample lesson built to your subject area, and critique it together. A weak team produces a weak sample; a strong team produces a strong sample and a good conversation about why they made each choice.
For a guide to assessing the quality of the instructional design work itself rather than the agency, see our guide on evaluating instructional design quality.
Compare proposals on structure, not headline price
Headline prices on proposals are rarely comparable because scope definitions differ. One supplier may include two rounds of revisions, licensed stock footage, and translation of one course, while another excludes both. Ask each to quote against a written scope sheet that you own, so every line maps to the same deliverable.
Look for the cost drivers rather than just the totals. The variables that most often move a custom eLearning price are course count and length, the number of media assets, the level of original scripting versus adapting existing material, interactivity complexity, review cycle depth, language coverage, and the number of learning management system integrations required. Our guide on what affects the cost of custom eLearning development covers each of these in detail.
A supplier who shows you a cost breakdown by these drivers, and is willing to explain which ones you could reduce if the budget is tight, is giving you negotiating room. A single lump sum does the opposite.
Pressure-test fit on logistics and scale
The procurement risk in custom development is rarely creative. It is logistical: unclear review cycles, an unexpected number of stakeholder rounds, content owners who do not respond, and a single launch date that assumes everyone else behaves. Ask how the supplier handles stalled reviews, what the escalation path looks like, and who on their side can make decisions quickly.
Scale matters too. If the plan involves a curriculum rather than a single course, the real question is whether the supplier can produce consistently across many units and keep visual and technical quality from drifting. Ask how they maintain style consistency and how they handle a second cohort of subject matter experts who were not involved in the first.
If your project includes a live or facilitator-led element, a heavily scripted digital course may be the wrong medium. The honest supplier will tell you when a blended approach serves the outcome better.
Commonly overlooked pitfalls
- Shortlisting on brand recognition and awards rather than on work that resembles your subject matter, audience, and delivery constraints.
- Comparing headline totals from proposals written against different scope definitions, which makes the numbers meaningless.
- Accepting named-team commitments that are not contractually binding, and losing your named designer mid-project.
- Ignoring review-cycle governance, so the project is delayed by well-meaning but uncoordinated stakeholders.
- Treating accessibility, analytics, and language versions as extras rather than scoping them from the start.
Next steps
- Write a one-page problem statement, target audience profile, and the business measure you intend to move.
- Build a scope sheet that defines deliverables line by line, then require every shortlisted supplier to quote against it.
- Shortlist three to five suppliers using comparable work and reference-call evidence rather than marketing claims.
- Run a short paid sample lesson with each of the final two or three candidates and score them against written criteria.
- Require staffing commitments, review-round limits, and IP and accessibility terms in the contract itself.
Related guides
Put this into practice
The vendor evaluation checklist and the RFP template turn a guide into a process you can run this week.