Tüketici panelleri günümüzde gerek pazarlama yöneticileri gerekse marka tercih modelleri üzerinde çalışan akademisyenler tarafından yaygın olarak başvurulan bir veri kaynağı haline gelmiştir. Gelişmiş ülkelerde genellikle tarayıcı teknolojilerine dayalı olarak toplanan tüketim paneli verisi gelişmekte olan ülkelerde günlük tutma yöntemiyle elde edilmektedir. Bu yöntemle toplanan hane paneli verilerinin kapsam ve güvenilirliğinin tarayıcı teknolojisiyle elde edilen verilere göre daha sınırlı olduğu kabul edilmektedir. Bu çalışmanın amacı, alışveriş günlüğüne dayalı tüketici paneli verisiyle Türkiye’de, hızlı tüketim malları sektöründe, başarılı bir marka tercih modeli kurabilmenin mümkün olup olmadığını araştırmaktır. Bu amaçla, öncelikle, günlük tutma yöntemine dayalı panellerde doğrudan ulaşmanın mümkün olmadığı rakip markalara ait fiyat bilgilerini üretmeye yönelik dört aşamalı bir yöntem önerilmiştir. Daha sonra, Türkiye’de 2004 yılında 1632 hane tarafından yapılmış 26031 kolalı içecek alışverişinden yararlanılarak çeşitli çoklu probit ve lojit modelleri kurulmuş; modellerden elde edilen bulgular rassal olarak belirlenen 408 hanelik test örneklemine ait 6686 alışveriş gözlemi üzerinde sınanmıştır. Sonuçlar göreli fiyatlar, hane büyüklüğü, toplam harcama, bir önceki tercih ve sosyo-ekonomik statü açıklayıcı değişkenlerinin yüksek güven düzeylerinde anlamlı olduğunu ve kurulan modellerin gerek test gerekse model örneklemlerindeki hanelerin marka tercihlerini başarıyla tahmin ettiğini göstermektedir. Yapılan marjinal etki analizlerine seçilen modellerin verdiği tepkiler de teorik beklentilerle hassas bir uyum içindedir. Bu çalışma, Türkiye’de hane paneli verisi ile tüketicilerin bir sektördeki marka tercihlerini modellemeyi hedefleyen öncü çalışmalardan biri olması bakımından önem taşımaktadır.

One of the major sources of consumer level brand consumption data in Fast Moving Consumer Goods (FMCG) sector is the household consumer panels maintained by individual organizations. Consumer panels are capable of keeping histories of purchases for a sample of households. Household panel data is collected via barcode scanner technologies in the developed countries. Data collected by the optical scanning of Universal Product Code permits researchers study the effects of marketing variables on customer choice among product alternatives since the mid-1980s and many dozens of papers are published on brand and retailer choice issues since then. Employing scanner technology, panel organizations does not require its members to record all the items they purchase in a shopping trip, as their panel ID cards are scanned during the payment process at checkout. Building cooperation with the retailer companies, scanner panels are able to provide price and promotion data for all the products (in a given product category) without the necessity of being purchased. On the other hand, due to the high levels of technology costs, the household level panel information in developing markets is primarily gathered using the diary mode. It is believed that, in this mode of data collection, both the reliability and the scope of information collected from the panellists are still scarce compared to their scanner type counterparts in more developed markets. One of the important problems of diary mode data is the lack of complete price information. In diary mode panel organizations, households only record the prices of the products they purchase. As it is impossible for a shopper to list the prices of all the products in the product categories (of which the shopping basket includes an item), diary mode household panels are not capable of conducting complete price data in a given product category. As the brand choice models literature using household panel data mostly stands upon the unit prices of the products purchased and their alternatives displayed on shelves, there is a gap of modelling research in the developing markets in which consumer panel measurements are made based on diary mode. The purpose of this study is to investigate whether it is possible to establish an efficient brand choice model based on diary mode household panel data from the Turkish carbonated soft drinks market. In order to accomplish this goal, firstly, a four staged method of price data generating is suggested. Suggested method basically generates artificial unit prices for the products which are present at the retailer during the shopping trip but not purchased. Secondly, employing this price data, different multinomial probit and logit models of brand choice are built based on 26031 purchase occasions of a model sample of 1632 households. Using relevant goodness of fit and specification tests (mostly depending on the number of choices that were made correctly), the performances of these six models are tested on the 6686 purchases of a randomly selected test sample of 408 households. The results show high significance for the explanatory variables of brand loyalty (previous brand purchased), relative prices and household specific features such as total/per capita FMCG spending, household size and socio-economical status. While the coefficients have expected signs, the models predict the share of purchases remarkably well in both the model and the test samples. All of the models are capable of predicting more than three thirds of the brand choices correctly. However, Hausman tests reveal that the Independence of Irrelevant Alternatives criteria does not hold for the multinomial logit models. Therefore, theoretically, it is not possible to use the multinomial logit method in modelling the Turkish carbonated soft drinks market, although it was useful in variable selection process. Hopefully, the goodness of fit measures and specification tests show that, there is no theoretical problem for the multinomial probit models of brand choice built in this study. Moreover, the reactions of the probit models to the marginal effect analyses are consistent with the theoretical expectations and the models are capable of predicting the monthly shares of purchases with the error range of ± 3% successfully. Employing diary based consumer panel data, this study is one of the pioneering attempts to integrate the consumption level data and consumer features towards developing a brand choice model in Turkey.