To forecast or not to forecast? That is the question. The blog below answers a different question - what's the best way to go about it?
It is becoming increasingly common for PPC clients to request a forecast of account performance, and of course this is an understandable request, particularly if budget changes are a factor. The problem with calculating accurate projections of performance is that you can’t. Whilst it’s possible to make reasonable estimations based on historical data, PPC performance still ultimately relies on consumer demand. Further to this, there are a number of factors in paid search that an account manager has no control over but that could have a substantial effect on performance such as search volume and the aggression of the competition. In this post I discuss the issues with PPC forecasting and talk through a data driven method to best predict account performance.
The Influence of… Competitors
An accurate forecast depends on certain metrics remaining static but most account managers know this is rarely the case. Even if your optimisation is and always has been spot on, paid search is by definition an auction. It is not just your own account management that has an effect but the account management of those advertisers bidding on the same keywords. It may be that historically customer acquisition and conversion has never been too difficult, particularly if you were one of a few advertisers and the only one doing a good job. But what happens if a lot more competitors enter the market or your existing competitors step up their game? It’s certainly possible to analyse the competition retrospectively using Google tools such as Auction Insights. However, successfully predicting competitor behaviour is no easy task and can really only be based on (a running theme here) historical data.
Running frequent auction insights reports can give you a good indication of which of your competitors tend to be more aggressive and when. Do they push harder in the afternoon and ease off in the evening? Getting a hold of patterns like these can help inform future PPC strategy and projections. For example, you might see that your biggest competitor’s impression share consistently falls in the final few days of the month, allowing your CPCs and daily spend to drop whilst maintaining the same, or even better visibility. It is therefore sensible to incorporate such observations in your forecasts, alongside the caveat that just because a competitor has always used a particular ad schedule or bidding strategy, it doesn’t necessarily mean they will continue to do so.
The Influence of… Seasonality
Of course, it’s not impossible to predict all future search behaviour. General trends in seasonality will often remain the same year on year. Any flip-flop merchant will surely be aware of the sharp incline in search volume between April and May and the corresponding significant drop once September rolls around, depicted here by a Google Trends analysis of the single search term ’flip flop’ in the UK.
However, notice the difference in incline from the summer of 2012 to the following summer (a truly triumphant year in the history of the flip flop). Even with a well-known, existing trend such as summer flip-flop interest combined with access to historical search data through Google Trends, the precise size of an incline or decline is unfortunately not something that can be accurately quantified in advance.
The Influence of… Campaign Expansion & Improvement
As I’ll say many times in this blog post, the most reliable way to calculate a performance forecast is to use the existing data within the account. The problem with this is the assumption that your account remains in a relatively static state. That is, the account structure, campaign settings, ad copy, keyword lists, bidding strategies, ad extension application, feature usage and general optimisation remain the same.
Ideally, your account will undergo numerous and frequent changes in the form of ad copy tests, search query additions and exclusions, new ad groups and the use of all appropriate betas and features Google has to offer. Likewise, changes on the client’s side such as website usability, product ranges and price points will likely change from year to year, especially in industries with quick range turnover such as fashion retail. All of the above will have a knock on effect on key account metrics such as clickthrough rate, conversion rate, quality score and therefore CPC and subsequently, return on ad spend.
How to Use Your Historical Data
Despite the many pitfalls of providing forecasts for PPC, many clients will still be after some hard stats. Here’s how to reach those figures using your best weapon – what you’ve already got in your account. The example below explains the calculation for forecasting performance in October 2015, using the percentage of monthly uplift from September to October the previous year.
Firstly, download top level data from the previous year for the month you wish to forecast and the month preceding it. Here, I’ve downloaded data from September and October of 2014. The second step is to calculate the percentage change from one month to the next.
The next step is to download data for the month preceding the month you wish to forecast but for the current year. Here, I’ve downloaded data for September 2015.
Finally, multiply each metric for September 2015 by the percentage change calculated from data for 2014 and use the result as your forecast for October 2015.
Not only is this method based on hard numbers, thus removing any ‘guesstimates’, but it also incorporates the effects of seasonality into your future predictions.
Client Expectations
A performance forecast based on historical data assumes that search volume, clickthrough rate, conversion rate and CPC remain the same and for the reasons above, this will almost certainly not be the case, essentially guaranteeing that any forecast data you provide will be inaccurate.
Managing the expectations of your client is perhaps the most challenging aspect of providing a performance forecast. Be clear that PPC does not create demand, but responds to it and what this means regarding potential increases or decreases in search volume. Also explain the number of influencing factors that an account manager cannot control and the potential effects on performance. It can also be useful to explain the calculation process to emphasise the point that the forecast figures are based entirely on existing data.
Despite the many caveats of providing performance forecasts, they can be useful not only for your client but also for you as an account manager when planning for budgets around peak season. When providing frequent periodical forecasts, it’s also useful to compare previous forecast calculations to actual data to aid your client in understanding the difficulty of accuracy when it comes to predicting the future.
Authored by:
