Geotagging, the process of tagging of documents with geographical information, is an important issue in geographical information retrieval systems. In this paper, we propose a novel approach for geotagging of textual web documents. Existing techniques disambiguate place names occurring in a document individually first and then find its focus. As opposed to this, the proposed approach, considers all the place names together without disambiguating them individually. It is a heuristic-based three step process. The first step is hierarchical, the second one is geometric technique, and last step reports the final focus. The approach requires only a gazetteer and a named entity recognition tool for processing. It reports the geographical focus of a web page as a tuple (latitude and longitude). We have tested our algorithm on two data sets, namely, the Wikipedia and Open Directory Project (ODP) data sets. The proposed approach reports the focus correctly up to continent level for 97.07% of the Wikipedia pages and correctly up to country level for 95.57% of them. Moreover, for more than 60% of the documents, the error in the distance between the actual focus, and the reported focus is within 15 km, i.e., less than 10 miles for this data set. In addition, the median error is 8.17 km. The technique also performs very well on huge data sets like ODP with sample size of around 600000 web pages. The proposed system can be used in any geographic information retrieval system where geographic meta-data of documents needs to be obtained for indexing and searching.