The experience of Altmetric LLP, an altmetric tool developer, reveals common issues that demand attention when designing alternative metrics for response to scholarly writings. Identifying what can and should be measured for different user groups is fundamental. A default is to count all relevant mentions in a set of online sources, permitting drill down for more qualitative information. Data source selection varies by need, ranging from government documents to social media comment sites. Since the topic of discussion can be elusive, a tracking method must point backward to original articles or data. Text mining helps for text documents, but audio and video are less workable. Multiple versions of a single article and subsections of books and datasets add ambiguity and redundancy. Valid interpretation depends on context and the relevance and timeliness of data and sources, requiring continual reassessment.