The Business Case: Is the Change Worth the Cost?
A practical way to compare a proposed website improvement with a simpler alternative, using actual costs and useful outcomes.
On this page — choose a section
The idea in one sentence: invest in a more responsive website only when a useful improvement is worth its full cost.
This paper offers a way to build that case. It does not report measured customer results, quote current supplier prices or promise a return. A clearer static page may be the better investment.
Name the problem before buying the solution
“Make the website intelligent” is not a business problem. “People cannot find the right service, so staff redirect many inquiries” is one.
Start with an observable difficulty and a baseline—the situation before a change. For example:
- Visitors cannot tell which offer fits their task.
- A contact button leads to the wrong team.
- People repeat information when a conversation moves from the site to staff.
- An important answer is buried in several pages of explanation.
List the costs you actually have: content work, software, hosting, support, maintenance and staff time. Do not assume every organization has an expensive collection of disconnected tools, or that replacing those tools will automatically save money.
Also distinguish an inconvenience from a necessary conversation. A support call is not always a website failure, and sales discovery is not waste merely because the site could collect more information.
Six areas where a change might help
Scroll sideways if needed to read the whole table.
| Area | Plain-language question | Useful evidence |
|---|---|---|
| Sales inquiries | Do people reach the right team with a clear request? | Appropriate inquiries, fewer transfers, staff assessment of context. |
| Content | Can visitors find and understand an important answer? | Task completion and an uncoached explanation in their own words. |
| Acquiring customers | Does the change improve the cost of winning suitable business? | Actual sales and marketing costs, qualified outcomes and the comparison method. |
| Support | Can people resolve a task or reach appropriate help? | Correct resolution, repeat contacts and time spent by visitors and staff. |
| Understanding demand | Which unanswered questions recur? | Appropriately collected questions, interviews and task evidence—not guessed identities. |
| Ongoing relationships | Can existing customers resume useful work? | Successful task resumption, corrections, complaints and use of reset controls. |
A lead is a potential customer inquiry. Conversion is a defined action, such as submitting an inquiry or making a purchase. Neither term establishes that the visitor received a useful outcome.
Customer acquisition cost (CAC) is the sales and marketing cost assigned to gaining customers over a specified period. State what you included, how you counted customers and whether the comparison is fair. A website visit alone does not reveal the full decision that happened offline.
Attribution means assigning credit for an outcome. If customers often read a page before buying, that is an association; it does not prove that the page caused the purchase. Do not present a browsing sequence as a complete explanation of revenue.
A small, transparent business case
Consider a fictional business testing an optional, visitor-reviewed inquiry summary. Suppose it handles 40 relevant inquiries per month. If a pilot actually finds that the summary saves an average of five staff minutes per inquiry, that is 200 minutes, or about 3.3 staff hours per month.
Those numbers are assumptions for this example, not Aware Design results. Before proceeding, ask:
- Did the summary save time, or merely move work from staff to visitors?
- Was important discovery skipped?
- Did incorrect summaries create extra work?
- Is the saved time useful capacity, an actual cash saving, or neither?
- Would a clearer form produce the same benefit at lower cost?
Return on investment (ROI) compares the value gained with the cost incurred. A basic calculation is:
Net benefit = supported value of the improvement − total cost of the change.
If you express ROI as a percentage, state the period and use a positive, clearly defined cost denominator. Avoid combining speculative future revenue with measured current savings as if both were equally certain.
Include the full cost
Account for discovery, design, implementation, content maintenance, testing, accessibility, hosting, support, privacy/security review and the cost of undoing the change. If an assistant is involved, include model and tool usage, content updates, answer review and human escalation.
Compare at least three options: keep the current experience, improve the ordinary page, or add the proposed adaptation. The smaller option should be allowed to win.
What if the site includes an AI assistant?
A simple choice-and-order rule does not need an AI model. An assistant that generates answers introduces a different set of costs and risks.
A language model generates responses from its input. Inference cost is the cost of running that model to produce an answer. The total depends on actual usage, the chosen model, the amount of text processed and any tools it calls. A long conversation can be costly without being useful.
Agree on limits before launch:
- A total spending cap and a limit per conversation.
- Which questions the assistant may answer.
- When it must stop, say it does not know or route to a person.
- A useful ordinary alternative if it is unavailable or the budget is exhausted.
- What data it receives, which providers process it and what is retained.
Do not assume that most visitors will avoid the assistant or that expensive conversations will produce valuable customers. Measure those things. Paid overages and a broader scope need prior agreement, not an automatic escalation hidden from the buyer.
Could this support an ongoing service?
Possibly. A service could help a team keep content current, investigate confusing paths, maintain rules and evaluate whether changes remain useful. Those are concrete activities a buyer can understand and assess.
A subscription should state the work included, limits, review cadence, data responsibilities and how the customer can leave. Do not sell “a website that gets smarter every month” as a guaranteed outcome. Models can become outdated, visitor needs can change and a successful rule can stop helping.
More advanced proposals include forecasting likely needs or reporting patterns across visits. Treat them as separate projects. They need appropriate data, validation, access controls and a clear reason why they benefit the customer. They are not proven simply because the site can collect events.
A recurring service is justified by useful ongoing work—not by making a customer dependent on a hidden profile they cannot inspect or export.
A decision a small business or a larger team can make
Write a one-page case with five answers:
- The problem: what are people struggling to do?
- The proposed change: what will be different, and why might it help?
- The alternative: could clearer content or navigation solve it more simply?
- The evidence: what will you observe, compare and count as failure?
- The budget and stop point: what can you spend, who owns the work and when will you decide whether to continue?
A successful pilot can justify a next step. An unclear or negative result can justify stopping. Both are better than committing to a broad platform on an untested promise.
Find a suitable first problem or turn it into a build brief.