Price Testing: Run Experiments to Find the Optimal Price

Price Testing: Run Experiments to Find the Optimal Price

Why Price Testing Matters More Than Gut Instinct

Most Singapore businesses set prices using one of three methods: cost-plus markup, competitor matching, or whatever number feels right. All three leave money on the table. A price testing strategy replaces guesswork with data, and the impact on profitability can be staggering — a 1% improvement in price realisation typically delivers an 8-11% increase in operating profit, depending on your margin structure.

Consider a SaaS company in Singapore selling a project management tool at SGD 29 per user per month. They assumed the price was competitive because two rivals charged SGD 25 and SGD 35. After running a Van Westendorp study with 400 respondents, they discovered their acceptable price range was SGD 32-45, with an optimal price point of SGD 38. That single insight, implemented over one billing cycle, lifted annual recurring revenue by 31% with minimal churn.

Price testing is not about squeezing every dollar from customers. It is about understanding the perceived value of your product and aligning your price to that perception. When you price too low, you signal low quality. When you price too high without the positioning to support it, you lose sales to competitors. The sweet spot exists, and testing finds it.

The Cost of Getting Price Wrong

Underpricing is the more common sin in Singapore’s competitive market. Business owners fear losing customers, so they default to being the cheapest option. But low prices attract price-sensitive buyers who churn the fastest, compress margins that could fund growth, and create a race to the bottom that benefits nobody. A structured price testing programme reveals what customers actually value and what they are willing to pay for that value.

A/B Testing Prices Online

A/B testing prices means showing different price points to different visitor segments and measuring the impact on conversion rate and revenue. It is the most direct form of price experimentation, but it comes with significant ethical and practical caveats that we will address later.

How Price A/B Tests Work

The mechanics are straightforward. You split incoming traffic into two or more groups, each seeing a different price for the same product. You then measure conversion rate, revenue per visitor, and total profit per visitor. The variant that maximises your chosen metric wins.

For an e-commerce store selling physical products, this might mean testing SGD 49.90 against SGD 54.90 for a skincare set. For a service business, it could mean testing SGD 1,500 against SGD 1,800 for a digital marketing retainer package. The key metric is not conversion rate alone — it is revenue per visitor or profit per visitor, because a higher price with slightly lower conversion often yields more total profit.

Sample Size and Duration

Price tests need larger sample sizes than typical A/B tests because the effect sizes are usually smaller. Aim for at least 1,000 conversions per variant to detect a meaningful difference. For most Singapore SMEs, that means running the test for 4-8 weeks. Shorter tests risk seasonal distortions or insufficient data.

Tools for Price A/B Testing

Google Optimize (now integrated into GA4), VWO, and Optimizely all support price testing. For e-commerce platforms like Shopify, apps such as Intelligems or Prisync allow price experiments without custom development. If you use WooCommerce, plugins like Nelio A/B Testing or custom code with Google Tag Manager can handle the split.

The Van Westendorp Price Sensitivity Meter

The Van Westendorp method uses a four-question survey to map your audience’s price expectations. Developed by Dutch economist Peter van Westendorp in 1976, it remains one of the most reliable survey-based pricing tools because it captures the psychological boundaries of price perception rather than asking a single willingness-to-pay question.

The Four Questions

Each respondent answers four questions about the product or service:

• Too cheap — At what price would you consider this product so cheap that you would question its quality?
• Cheap / good value — At what price would you consider this product a bargain — a great buy for the money?
• Expensive / getting pricey — At what price would you consider this product starting to get expensive — not out of reach, but you would have to think about it?
• Too expensive — At what price would you consider this product so expensive that you would not consider buying it?

Interpreting the Results

Plot cumulative frequency distributions for each question. The intersections of these curves give you four critical price points: the Point of Marginal Cheapness, Point of Marginal Expensiveness, Indifference Price Point, and Optimal Price Point. The acceptable price range sits between marginal cheapness and marginal expensiveness, and the optimal price point is where the “too cheap” and “too expensive” curves intersect.

For a Singapore-based B2B consultancy, a Van Westendorp study might reveal that their hourly rate of SGD 150 falls below the “cheap” threshold, meaning prospects perceive the rate as suspiciously low. Moving to SGD 220 — within the acceptable range — could actually increase close rates by removing the quality concern.

Limitations

Van Westendorp works best for products that respondents already understand. It struggles with genuinely novel products where people have no mental anchor for pricing. It also does not account for competitive alternatives — respondents answer in isolation. Supplement it with competitive analysis and, where possible, real-world transaction data.

Gabor-Granger Direct Pricing Method

The Gabor-Granger technique measures purchase intent at specific price points. Unlike Van Westendorp, which asks respondents to name prices, Gabor-Granger presents prices and asks whether the respondent would buy at that price. This makes it better suited to products sold in markets with established price expectations.

How It Works

Present respondents with a random starting price from your predetermined range. Ask a simple purchase intent question on a 5-point scale. If the answer indicates willingness to buy, increase the price in the next round. If not, decrease it. Continue for 4-6 rounds. This sequential approach reveals each respondent’s maximum willingness to pay without directly asking them to name a number — which people are notoriously bad at.

Building a Demand Curve

Aggregate the results across all respondents and you get a demand curve showing the percentage of people willing to buy at each price point. Multiply the percentage by the price to find revenue-maximising points. Factor in your cost structure and you can identify the profit-maximising price.

A Singapore meal-kit delivery service used Gabor-Granger with 600 respondents to test prices between SGD 9.90 and SGD 16.90 per serving. The demand curve showed a steep drop-off above SGD 13.90 but relatively flat demand between SGD 11.90 and SGD 13.90. They priced at SGD 13.50 — capturing maximum revenue in the flat zone while staying below the drop-off point.

When to Use Gabor-Granger Over Van Westendorp

Choose Gabor-Granger when you have a defined price range in mind and want to optimise within it. Choose Van Westendorp when you are exploring and want to understand the full psychological price landscape. Both methods require a minimum of 200 respondents for reliable results, and 400+ is preferable for segmented analysis.

Conjoint Analysis for Feature-Price Trade-Offs

Conjoint analysis does not test price in isolation. It tests how customers trade off features, brand, service level, and price simultaneously. This makes it the gold standard for complex products and services where price is one of several decision factors.

Choice-Based Conjoint (CBC)

In a choice-based conjoint study, respondents see sets of product configurations (each combining different feature levels and prices) and choose which they would buy, or choose “none.” By analysing these choices across hundreds of respondents and thousands of decisions, statistical models calculate the relative importance of each attribute and the part-worth utility of each level — including each price point.

For a web design agency, a conjoint study might test combinations of turnaround time (2 weeks, 4 weeks, 8 weeks), number of revision rounds (1, 3, unlimited), included pages (5, 10, 20), ongoing support (none, 3 months, 12 months), and price (SGD 3,000 to SGD 15,000). The results reveal not just the optimal price, but which features drive willingness to pay and which features customers do not value enough to justify their cost.

Running Conjoint Analysis

Conjoint analysis requires specialised software. Sawtooth Software, Conjointly, and SurveyAnalytics offer dedicated conjoint tools. Qualtrics includes conjoint modules in its higher-tier plans. Sample sizes of 300-500 are typical, with each respondent completing 12-20 choice tasks. The analysis uses hierarchical Bayesian estimation or multinomial logit models to extract utilities.

Practical Applications in Singapore

A Singapore-based co-working space operator used conjoint analysis to design their pricing tiers. They discovered that 24/7 access and meeting room credits had disproportionately high value — far more than the premium gym access they had been including. By restructuring their tiers to emphasise access and meeting rooms while making gym access an add-on, they increased average plan value by 22% while reducing costs.

Ethical and Legal Considerations

Price testing, particularly A/B testing with real transactions, sits in an ethical grey area that requires careful handling. Singapore’s Consumer Protection (Fair Trading) Act does not specifically prohibit price testing, but the practice must not be misleading or deceptive.

The Fairness Problem

If two customers buy the same product at the same time and pay different prices, you have a fairness problem. Amazon faced backlash in 2000 when customers discovered they were being shown different DVD prices based on browsing history. The reputational risk is real, especially in Singapore’s tight-knit market where word travels fast through forums, Reddit, and Telegram groups.

Ethical Price Testing Approaches

The safest approaches include testing prices sequentially rather than simultaneously (show one price this week, another next week), testing on new products where customers have no price anchor, using survey methods like Van Westendorp and Gabor-Granger that do not involve real transactions, and offering the lower price to anyone who purchased at the higher test price if discovered. Some businesses test on landing pages where the price is shown but the transaction happens later, allowing them to honour whatever price the customer saw.

PDPA Considerations

Under Singapore’s Personal Data Protection Act, if your price test uses personal data (browsing history, location, purchase history) to determine which price a user sees, you need a legal basis for that data processing. Consent or legitimate interest may apply, but segment-based pricing that uses personal data requires careful legal review. Random assignment based on session ID, with no personal data used in the allocation, is the simplest compliant approach.

How to Run a Price Test Step by Step

Whether you choose A/B testing, surveys, or a hybrid approach, the process follows a consistent structure. Getting this right matters as much as choosing the right method.

Step 1: Define Your Objective

Are you optimising for revenue, profit, market share, or customer lifetime value? These objectives can produce different optimal prices. A market-share strategy might favour a lower price that maximises volume, while a profit-maximisation strategy might favour a higher price with lower volume. Be explicit about your goal before you test.

Step 2: Select Your Method

For quick directional insights, use Van Westendorp (survey-based, low cost, fast). For demand curve estimation within a price range, use Gabor-Granger. For complex products with multiple attributes, use conjoint analysis. For definitive real-world validation, use A/B testing with live transactions. Many businesses combine methods — a survey to narrow the range, then an A/B test to validate the final price.

Step 3: Design the Test

For A/B tests, keep variants to 2-3 prices to manage sample size requirements. For surveys, recruit respondents who match your target customer profile — not a general panel that includes people who would never buy your product. For conjoint, limit attributes to 4-6 to avoid respondent fatigue.

Step 4: Run and Monitor

Monitor your test daily for anomalies but do not stop early unless something is clearly broken. Statistical significance calculators for revenue metrics (which are not binary like conversion) require specialised tools. Bayesian approaches work well here because they give you probability distributions rather than binary significant/not-significant results.

Step 5: Analyse and Implement

Look beyond the headline number. Segment results by customer type, acquisition channel, and geography. A price that works for organic search traffic may not work for paid traffic from Google Ads, because the intent and price sensitivity differ. Implement the winning price, but plan to re-test in 6-12 months as market conditions change.

Price Testing in the Singapore Market

Singapore’s market has characteristics that affect how price tests should be designed and interpreted. Understanding these factors prevents you from applying generic advice that does not fit the local context.

Small Market, High Stakes

Singapore’s total addressable market for most B2C products is 5-6 million people. For B2B, the pool is even smaller. This means your price test sample might represent a significant portion of your total market. Sequential testing (changing prices over time and comparing periods) is often more practical than simultaneous A/B testing because you cannot afford to alienate half your market with a price they perceive as unfair.

Price-Quality Perception

Singapore consumers are sophisticated and price-aware, but they are not purely price-driven. Research consistently shows that Singaporeans use price as a quality signal, especially for services. A branding consultancy that prices below SGD 5,000 for a brand identity project may struggle to win quality clients not because the deliverable is poor, but because the price signals that it might be.

GST and Pricing Psychology

With GST at 9%, Singapore businesses must decide whether to show GST-inclusive or exclusive prices. For consumer-facing price tests, always test with the final GST-inclusive price, because that is what customers actually pay. For B2B, test with GST-exclusive prices but show the inclusive total clearly. Your test results will be skewed if you test one way but sell another.

Multi-Currency Considerations

If you sell regionally — to Malaysia, Indonesia, or other ASEAN markets — your Singapore price test may not translate directly. Purchasing power parity, competitive dynamics, and price expectations differ significantly across the region. Consider running separate tests for each market rather than applying a single price with currency conversion.

For businesses looking to integrate pricing insights into their broader marketing efforts, a cohesive content marketing strategy can communicate your value proposition and support premium pricing positions through educational content and thought leadership.

Frequently Asked Questions

How many respondents do I need for a price sensitivity survey?

For Van Westendorp or Gabor-Granger studies, aim for a minimum of 200 respondents from your target customer profile. If you plan to segment results (by age, income, or usage frequency), you need 200+ per segment. For conjoint analysis, 300-500 respondents is standard. Fewer than 150 respondents for any method produces unreliable results.

Is it legal to A/B test prices in Singapore?

Singapore law does not specifically prohibit price A/B testing. However, practices that are misleading or deceptive could fall foul of the Consumer Protection (Fair Trading) Act. The safest approach is to test sequentially (different prices at different times), use survey methods, or ensure that any customer who discovers a price discrepancy is offered the lower price.

How often should I re-test my pricing?

Re-test pricing every 6-12 months, or whenever there is a significant change in your product, market, or cost structure. Major events like a new competitor entering the market, a change in GST rates, or a significant product upgrade all warrant fresh price testing.

Can I use price testing for services, not just products?

Yes. Service pricing is often better suited to survey methods (Van Westendorp, Gabor-Granger, conjoint) because service purchases tend to be less frequent, making A/B tests impractical due to small sample sizes. Present respondents with clear service descriptions and deliverables before asking pricing questions.

What is the difference between price testing and dynamic pricing?

Price testing is a research method to find the right price. Dynamic pricing is a strategy of changing prices in real time based on demand, inventory, or customer segment. Price testing informs your baseline price; dynamic pricing adjusts that price situationally. Airlines and ride-hailing apps use dynamic pricing; most Singapore SMEs benefit more from price testing to find a stable optimal price.

How do I test prices for a product that does not exist yet?

Use concept testing combined with pricing surveys. Present respondents with a detailed product description, feature list, and mockups, then run Van Westendorp or Gabor-Granger questions. Conjoint analysis works particularly well here because it tests feature-price trade-offs simultaneously, helping you design both the product and its price.

Should I test prices on my existing customers or new prospects?

Test on both, but separately. Existing customers have a price anchor (your current price) and their responses will cluster around it. New prospects have no anchor and give you a cleaner read on market willingness to pay. If you plan to raise prices, test on new prospects first — they will tell you what the market will bear without the bias of an existing relationship.

What sample size do I need for a price A/B test?

Price A/B tests need at least 1,000 conversions per variant for reliable results. Because you are measuring revenue per visitor (a continuous variable) rather than a binary conversion, the statistical tests are different and require more data. Use a revenue-based sample size calculator, not a standard conversion-rate calculator.

How do I handle price testing for subscription products?

For subscriptions, lifetime value matters more than initial conversion. A lower price might convert more users but attract higher-churn segments. Run your test for at least 2-3 renewal cycles before drawing conclusions. Many Singapore subscription businesses find that a higher initial price with generous cancellation terms outperforms low introductory pricing that increases later.

Can price testing damage my brand?

Poorly executed price testing can damage trust if customers discover they were shown different prices. Survey-based methods carry no brand risk because they are explicitly research. Sequential A/B tests (different prices at different times) are low risk. Simultaneous A/B tests with different prices for different visitors carry the highest risk and should be handled carefully with a plan for addressing customer complaints.