Sales show what people bought. Market demand also includes what they wanted and did not buy.
What is the difference between demand and market demand?
Demand can describe one buyer, while market demand adds up all buyers. Economics textbooks call market demand the horizontal sum of individual demands. In practice, it answers one question: how much would this whole market buy at this price?
Your sales are what your company sold. Market sales are what the whole category sold.
Market demand is what buyers wanted, including what they could not get or did not find.
Market size is usually a value for a period. For that calculation, read the guide to total addressable market.
The gap between sales and demand is real. When a shop runs out of stock, sales stop while demand continues. Forecasters call this censored demand: the shop never sees how much more it could have sold.
How do you measure market demand?
You measure market demand by combining sources that see different moments in the purchase. Each method below answers a different question.
| Method | What it measures | Lag | Main weakness |
|---|---|---|---|
| Sales data | Your own completed transactions | After the purchase | Misses unmet demand and the rest of the market |
| Surveys and stated intent | What people say they will do | Before the purchase | Stated intent can differ from behaviour |
| Panel and syndicated data | Category purchases across retailers or households | After the purchase, plus a reporting delay | Depends on coverage; panels are self-reported |
| Search and orientation signals | What people look into | Before the purchase, in some tests by weeks | Orientation is not a sale; often relative, not absolute |
| Social | What people post and follow | Around or before the purchase | Users differ from the general population |
| Test markets | Reaction to one offer, real or simulated | Before launch, but the test takes time | Can be costly and visible to rivals |
Sales and syndicated data are your anchor because they record real purchases. A 2022 industry explainer describes point-of-sale data as transaction-level data collected at the store. It describes panel data as self-reported purchases by households.
How do orientation, forecast and realised sales differ?
They describe different things, and mixing them up leads to the wrong decision. Orientation is measured. A forecast is expected. Realised sales are what happened. Keep a label on each one.
| What you read | What it is | What it is not |
|---|---|---|
| Orientation moments | Moments when people look into products, needs and brands, explore alternatives and prepare choices | Individual people or purchases |
| Demand development | How orientation changes over time, by market, need or brand | Growth in your own sales |
| Demand forecast | The expected development of demand | Exact revenue, sales or stock per SKU |
| Realised sales | What was actually sold, by channel or product | Total demand, because unmet demand stays hidden |
Levels matter as well. A market is too wide for a product choice, and an attribute is too narrow for a budget. Five levels keep the reading precise, always in the same order. This example is illustrative, not a market finding.
| Level | Illustrative example |
|---|---|
| Market | Home cleaning appliances, Germany |
| Category | Robot vacuums |
| Type | Robot vacuum with mop |
| Specification | 3 litre water tank |
| Attribute | Quiet operation |
Orientation can be read at all five levels. In Trendata, demand forecasts up to 12 months ahead apply at category, type and specification level.
What is the difference between leading and lagging signals?
A leading signal moves before the outcome you care about, and a lagging signal confirms it afterwards. Sales are lagging. Orientation can be leading. Neither is better, because one gives you time and the other gives you proof.
The test is timing. Forecasting: Principles and Practice separates forecasts made with information available in advance from forecasts that use later information. Only the first kind is a genuine forecast.
Take the signal and your sales over several years. Shift the signal by one, two and three months, then compare the fit with sales at each shift.
Then hide the most recent period. Forecast it using only what you would have known at the time.
A signal that only fits the past is a description. A signal that works on hidden data is a forecast.
Can search data measure consumer demand?
Yes, search data can indicate consumer demand, but how much it adds varies by case. Choi and Varian showed that search data can improve estimates of current activity, such as car, home and retail sales. They framed this as predicting the present, because official figures arrive late.
Goel and colleagues went further in PNAS in 2010. They forecast film, video game and song outcomes days or weeks ahead. Search improved baseline models by anything from modest to dramatic, depending on the application.
Read the limits too. One free trend tool scales a sample of searches from 0 to 100. The scale is relative to total searches in that place and time. Low-interest terms can show as zero. Its own help page says it is not a scientific poll. It advises treating the tool as one data point among others.
How reliable are surveys, social and test markets?
Each one is useful for a specific job, and none measures total demand alone. Surveys explain motives and test ideas, but people do not always do what they say. A 2026 industry commentary names price changes, promotions, availability and habit as reasons for that gap.
Armstrong, Morwitz and Kumar tested four intentions-based methods on car and wireless data. Each beat a simple extrapolation of past sales.
Combining the methods gave error rates about one-third lower than extrapolation. Their data covers those products only, so treat it as evidence and not as a rule.
Social shows what active users care about, but active users are a specific group. Mellon and Prosser found that British Twitter and Facebook users differ from the general population in age, gender and education.
Test markets come closest to real behaviour for a single offer. A 1995 Quirk's article sums up the criticism: real test markets are expensive and slow, and they can tip off competitors. It presents simulated test markets, with a mock store and a model, as cheaper, quicker and more private. Their output depends on that model.
How do I calculate market demand?
Multiply the number of likely buyers by how much each buys in a period. Multiply that by the average price if you need a value. This is arithmetic, not measurement, so its quality depends entirely on the inputs.
- Define the market. Name the need, the region and the period.
- Estimate buyers. Use panels, surveys or public population data.
- Estimate purchase rate. Use panel data or your own repeat rates.
- Add the price. Use the price paid, not the list price.
- Check the direction. Compare your estimate with orientation signals over time.
This gives you a level. A level says little about movement. The last step is therefore the one that tells you if demand is growing or shrinking.
Which method should you use?
Use the method that fits your decision, then add one source from the other side of the purchase. Forecasting: Principles and Practice draws the first line. Quantitative methods need relevant past data, while a new product with no history needs structured judgment.
| Decision | Start with | Add |
|---|---|---|
| Next year's budget | Sales and syndicated data | Search and orientation signals |
| New product idea | Search and orientation signals | Surveys, then a test market |
| Market entry | Syndicated data and market sizing | Orientation signals per country |
| Message or positioning | Surveys | Social |
| Stock and supply planning | Your own sales, corrected for stock-outs | Leading signals for turning points |
Keep the sources labelled. A dashboard that blends purchases and orientation into one score hides what you actually know. For the method after measurement, read the guide to demand forecasting.
Trendata publishes this guide. It belongs to the search and orientation row of the table, so it has the weakness of that row: orientation is not a sale.
Trendata reads aggregated orientation moments in Search, Social and AI environments, 18.5 billion a day. These are not individual people or purchases. The data is updated monthly, with 48 months of history.
It offers demand forecasts up to 12 months ahead, at category, type and specification level. They describe expected demand, not exact sales per SKU.
For more on the approach, read what Demand Intelligence is and these market intelligence examples.
Frequently asked questions
What is the most accurate way to measure market demand?
No single method wins. Sales and syndicated data are the firmest record of purchases. Leading signals arrive earlier. Test each signal against your own results.
How do you measure the size of a market?
Multiply buyers by purchase rate and price, or use a syndicated market estimate. Market size gives a level. It does not show direction.
How can I check market demand for free?
Start with a free trend tool and your own sales history. A trend tool shows relative interest, not sales or absolute counts, so treat it as one data point among others.
Why are sales not the same as demand?
Sales only count completed purchases. They miss buyers who found the product out of stock, chose a substitute or never found it.
Sources
- Individual and market functions (Curtis and Irvine, Principles of Microeconomics)
- Censored demand (Tessa Wilkie, STOR-i, Lancaster University, 2020)
- Sales forecasts for existing consumer products and services (Armstrong, Morwitz and Kumar, 2000)
- Why consumers say one thing and do another (NIQ, 2026)
- POS data vs panel data: key differences (NIQ, 2022)
- Predicting the Present with Google Trends (Choi and Varian, Google Research, 2009)
- Predicting consumer behavior with Web search (Goel et al., PNAS, 2010)
- FAQ about Google Trends data (Google)
- Twitter and Facebook are not representative of the general population (Mellon and Prosser, Research and Politics, 2017)
- Trade Talk: Improving the odds with STMs (Quirk's, 1995)
- Forecasting data and methods (Hyndman and Athanasopoulos, Forecasting: Principles and Practice, 3rd ed.)
- Forecasting with regression (Forecasting: Principles and Practice, 3rd ed.)