We tend to break keywords into themes. Its a good practice. However, an optimum level exists, where further breaking may result into decrease in conversion rate, which might result into negative ROI.
We often break Adgroups into Themes and microthemes. You can break adroups into as many themes as you can. People tend to think, breaking Adgroups into themes increases Quality Score and hence decreases cost per conversion. However, it is not always true. It may increase your quality score but may also decrease your conversion rate. Further, Google's system does not have any Adgroup level Quality Score. System recognises only Keywords and adtexts. However, a pseudo quality score does come into existence at adgroup level, because of this very set up and assuming its existence is beneficial, but tighter theme does not enhance it. Consider a case, where:
ADG= Theme1 + Theme2 + Theme3. ; Now breaking it into, ADG1, ADG2, ADG3 for individual themes, would increase the frequency of ad delivery. However, it would also decrease the rate of conversion. It is my observation, though can be proven mathematically too, Conversion rate for ADG > conv rate of (ADG1 + ADG2 + ADG3). If you are concerned about the cost and have budget limitations, which is practically a barrier for every advertising campaign, breaking keywords into too many themes is not a good idea. Advertisers who manage dedicated campaigns for Display network alone, should pay more attention to conversion data, because you never get keyword level data for these campaigns, and adgroups are structured according to themes alone. If you have too many adgroups in your campaign, try consolidating them by removing unnecessary grouping. A good campaign does have keywords grouped around a theme into adgroups, I do not mean to propose anything against this idea. However, it might become too much and may require consolidation to improve your conversion rate, and this does improve it.