Two slides lie on the desk: the first says the market is worth PLN 380 million, the second – PLN 620 million. Both were prepared reliably, both set out the methodology, and both concern the same category and the same year. This situation does not necessarily indicate an error on the part of the contractor. It may be a natural consequence of the fact that bottom-up and top-down market sizing rely on two different approaches to arriving at a figure – and the difference between them may itself carry information that no single estimate can provide.<\/p>\n
The fundamental difference lies in the direction of reasoning. Top-down modeling<\/strong> starts with the size of a broad whole – an overarching market, a statistical category, or household expenditure within a given group – and narrows it down through successive share-based filters until it reaches the segment of interest. Bottom-up modeling<\/strong> works in the opposite direction: it starts with an elementary unit, such as a customer, point of sale, transaction, or device, and multiplies it by the number of such units in the population, as well as purchase frequency and value.<\/p>\n In practice, the choice of method is rarely entirely flexible. It is determined by data availability. The top-down approach requires a reliable starting point at a high level of aggregation: data from GUS (Statistics Poland), Eurostat, industry reports, regulator reports, customs data, or compilations from chambers of commerce. If such a starting point exists and is defined in a way that closely matches the category under study, top-down produces a result quickly while remaining consistent with official statistics.<\/p>\n The bottom-up approach works well where a category is new, hybrid, or spread across several statistical classifications – that is, where no single overarching figure can be identified. In such cases, market sizing<\/strong> must be built from scratch or based on a combination of secondary and primary data: interviews with supply chain participants, surveys conducted on a sample of users, retail audits, or partners’ transaction data.<\/p>\n Typical configurations in which one method has an operational advantage are as follows:<\/p>\n It is worth noting that both approaches are part of a broader set of techniques used in market sizing methods<\/strong>. These also include analogous modeling, which transfers a market structure from a reference country while adjusting for purchasing power and category penetration, and the sum-of-shares method, in which market value is reconstructed from the financial data of identified players, supplemented by an estimate of the long tail of fragmented competitors.<\/p>\n Robust bottom-up and top-down market sizing often involves building two as-independent-as-possible models and comparing their results. In this context, independence primarily means that the models should rely on different calculation paths and, where possible, independent sources for key assumptions. If both rely on the same estimate of category penetration, their agreement has limited confirmatory value.<\/p>\n The process of building a bottom-up model usually includes the following steps:<\/p>\n A top-down model is built in the reverse order, and its quality depends on the justification for each successive narrowing coefficient. The key question is: how is it known that the segment under study represents this particular share of the overarching category rather than another? If the answer is “based on expert estimation,” the model requires triangulation – confirmation of the coefficient through an independent source, such as sales data from selected players or desk research<\/a> covering financial statements.<\/p>\n Two independent estimation methods rarely produce identical results, and the range between them may be more informative than the figure itself. A narrow range may indicate that the market definition and assumptions are stable, while a wide range points to variables where knowledge is weakest and which require additional measurement.<\/p>\n A practical procedure for reconciling discrepancies involves decomposing the difference. Rather than asking “which figure is correct,” the discrepancy is broken down into factors: how much results from different definitions of market boundaries, how much from differences in the assumed average price, how much from penetration assumptions, and how much from including or excluding gray-market channels, parallel imports, or intra-group sales. After such a decomposition, it often becomes clear that most of the difference is driven by one or two variables – and these are the areas where further fieldwork<\/a> should be focused.<\/p>\n Most discrepancies in bottom-up and top-down market sizing do not result from arithmetic errors, but from unrecognized assumptions built into the model design. Below are the pitfalls most frequently observed in project work:<\/p>\n A separate limitation concerns the time horizon. Top-down modeling often relies on historical data published with a delay, while bottom-up modeling based on fieldwork may incorporate more recent information. Comparing them without adjusting for category dynamics may generate an apparent discrepancy that in fact results from differences in the timing of measurement.<\/p>\n A model whose results cannot be verified has limited usefulness, regardless of how precise the final figure may appear. The following list makes it possible to assess the completeness of documentation prepared by a contractor or internal team:<\/p>\n Models accompanied by sensitivity analysis are usually easier to update – changing one parameter does not need to require rebuilding the entire structure, and the recipient can see which assumptions actually determine the result.<\/p>\n Bottom-up modeling involves reconstructing market value by aggregating transactions at the elementary-unit level – an individual customer, point of sale, or device. It requires determining the size of the population, category penetration, purchase frequency, and average transaction value, preferably separately for each segment with a different profile. Data may come from primary research, transaction data, and sufficiently detailed secondary sources.<\/p>\n It is sufficient when a reliable source describing the overarching category exists and the segment under study represents a clearly defined and documented part of it – for example, in markets covered by mandatory reporting to a regulator. It also works as a quick order-of-magnitude test before deciding whether to launch full fieldwork. It loses credibility when the path from the overarching category to the segment requires several coefficients based solely on expert judgment.<\/p>\n The discrepancy should be broken down into factors: first, check whether both models use the same market definition, price level, and period, then determine which variables account for the largest share of the difference. Often, this comes down to one or two items – for example, average price or category penetration. Once these variables have been identified, additional measurement should focus specifically on them, and the result should be reported as a range with scenario descriptions rather than as a single figure.<\/p>\n\n
How can both models be built, and how should discrepancies between them be interpreted?<\/h2>\n
\n
What errors most often distort market sizing?<\/h2>\n
\n
What should a well-documented market sizing model include?<\/h2>\n
\n
Frequently asked questions<\/h2>\n
What does bottom-up modeling involve?<\/h3>\n
When is the top-down approach sufficient?<\/h3>\n
What should be done when the two methods produce different results?<\/h3>\n