First off welcome to Quantopian! You may not need a custom factor. It depends upon how you want to use the factor. One could simply use the built-in factor for DailyReturns (see the docs) but then supply a mask to only get those stocks associated with companies which had earnings released 1 day ago. The mask can be made using another built in factor BusinessDaysSincePreviousEvent (see the docs).

Here is the pipeline code for something which may do what you want.

def make_pipeline():  
    """  
    Create a pipeline for days since earnings announcement  
    Also fetch returns for any security which made an announcement within 1 day  
    """  
    pipe = Pipeline()  

    # Create our slice for the last actual earnings data  
    fq0_eps_act = fe.Actuals.slice('EPS', 'qf', 0)  
    # Factor for the asof_date  
    fq0_eps_asof_date = fq0_eps_act.asof_date.latest

    # Factor for business days since the asof_date  
    days_since_release = BusinessDaysSincePreviousEvent(inputs=[fq0_eps_act.asof_date])  
    # Filter for stocks with earnings released 1 day ago.  
    made_public_1_day_ago  =  days_since_release <= 1 

    # Daily returns for just those stocks with earnings released 1 day ago (others will be nan)  
    returns_for_securities_with_recent_earnings_announcement = DailyReturns(mask=made_public_1_day_ago)

    return Pipeline(  
        columns={  
            'fq0_eps_asof_date': fq0_eps_asof_date,  
            'days_since_release': days_since_release,  
            'made_public_within_3_days_ago' : made_public_within_3_days_ago,  
            'returns': returns_for_securities_with_recent_earnings_announcement,  
        },  
        screen=made_public_1_day_ago  
    )

You also might get some ideas from these posts which discuss earnings and have some notebooks and algos to copy.

https://www.quantopian.com/posts/count-the-number-of-business-days-since-earnings
https://www.quantopian.com/posts/new-data-factset-estimates
https://www.quantopian.com/posts/factset-estimates-example-algorithm

Good luck.

thanks Dan! this looks great. you are right, it looks like I might not need a CustomFactor. I am trying to put together a multi-factor model. one of the factors would be reaction to earnings, following earnings announcement, and how it compares to the benchmark.

I am guessing I can put together a pipeline which would show me the earnings, returns one day after earnings and return of the benchmark one day after the earnings were announced using your approach. Is it also possible to add factors such as daily eps in the same pipeline?