{{ searchResult.published_at | date:'d MMMM yyyy' }}

Loading ...
Loading ...

Enter a search term such as “mobile analytics” or browse our content using the filters above.

No_results

That’s not only a poor Scrabble score but we also couldn’t find any results matching “”.
Check your spelling or try broadening your search.

Logo_distressed

Sorry about this, there is a problem with our search at the moment.
Please try again later.

When taking over any PPC account, it is essential to learn from historical performance swiftly in order to make the best possible start to a campaign.

In this example account there was a clear disparity between the best and worst hours / days of the week.

This suggested that a more aggressive ad scheduling strategy was required to bring the account to an even keel (and to perform as efficiently as possible) over the course of the day and week.

Analysis objective 

This analysis will isolate the most expensive non-converting time periods of the week (whereby no purchasing cycles were catalyzed, attributed to or completed), and conversely will isolate the best performing time periods, both over the course of the day and day of the week on average.

From there a bid matrix can be constructed in Excel, to be implemented into the Google Adwords campaign settings. Google Adwords will then adjust all bids automatically, relative to how that time period performed historically.

Analysis: hourly performance

Every hour of activity for the example account between July 1st and November 17th 2013 has been aggregated and illustrated in the graph below, depicting the web orders driven versus the cost incurred. 

The blue and gold lines are last click and equally weighted web orders respectively. Equally weighted web orders are important to show when a purchasing cycle has begun, been attributed to, or completed – and is the key factor that makes this level of analysis rare and useful.

An equally weighted sale is credited evenly among all touch points that happened in the purchasing journey/funnel.  

In the graph below, the gold line illustrates a culmination of all the PPC proportions of each sale generated per hour from July to November.  

Last Click (absolute) web orders are important to show when a purchasing cycle has been completed via PPC. The greater the gap between the blue and gold lines, the more channels that have been involved in each sale over the hour.

By using both last click and equally weighted metrics together, we can see both the top level performance as well as a deeper level of insight, analysing the full impact each hour has over the course of the day.  

From the graph above we can conclude that:

  1. Between the hours of 5am-9am, and 4pm–6pm, this example PPC account hardly begins a purchasing cycle, or plays a part in one, or is attributed the last click before a web order – causing extremely high cost per web orders (see the icy blue mountain between 7am and 9am, or the volcanic red and yellow peak in the graph between 4pm and 6pm).
  2. Between the hours of 8pm and midnight the account drives a significant number of web orders (both in helping a web order to be generated and last click) for considerably the lowest cost per web order over the course of the day.
  3. Between 12 noon and 7pm, 60% of the total web orders are generated, and there is significant impression share (lost to rank) available for the account to grow into, through increased bids.

    However as the last click CPW is close to or just over £100 overall, and as CPA is of vital importance to this client,  leaving the bid adjustment at 0% - or even pulling back - would be recommended.

Google Adwords currently only allows six time slots over the course of the day, and so the hourly performance between July and November 2013 were aggregated into six separate four hour schedules.

After analysing performance over the course of the day, the following conclusions can be drawn with ad scheduling bid changes in the table below (from left to right – best to worst performing four hour slots in the day):

(please note: bid adjustments can be set from -90% to +900%) 

Analysis: day of the week performance

The next stage would be to convert this time of day performance into a bid matrix (taking the days of the week into consideration).

To integrate performance over the course of the week into a bid matrix, equally weighted sales have again been included - to show anytime when PPC has catalyzed, attributed to or helped to complete a web order in the purchasing cycle - divided by the number of touch points.

Last click web orders have been added to show when the purchasing cycle was completed, with the last touch point deriving from PPC.

Below you can see the days of the week between July 1st and November 17th 2013 broken down and illustrated in a graph depicting the web orders driven versus the cost incurred. Average CPW increases from left to right, and the days of the week have been ordered in this way. 

Conclusions drawn from performance over the course of the week include:

  1. Tuesday and Friday were comparatively the best performing days of the week for the Example account (both in helping web orders, and last click web orders, and in CPW).
  2. On Thursday and Monday more touch points from other marketing channels come into play and a gap opened between last click and equally weighted web orders (and also in CPW).
  3. All days of the week bar Tuesday will need to be pulled back in terms of bid management as they are either close to or over the benchmark £100 CPW.
  4. Wednesday and Saturday were statistically the worst performing days of the week for the example account, by a considerable margin, both in helping web orders and last click web orders, and in their high CPW (nearly doubling Tuesday’s average CPW, which was the best performing day of the week). 

After analysing the performance over the course of the week, the conclusions from the graph above give the ad scheduling bids in the table below:

 

(please note: bid adjustments can be set from -90% to +900%)

The finished bid matrix

From both the time of day and day of the week analysis, the below bid matrix can be implemented:

Chris Swan

Published 9 December, 2013 by Chris Swan

Chris Swan is PPC Manager at Jellyfish and a contributor to Econsultancy. You can connect on Google Plus or LinkedIn.

1 more post from this author

Comments (14)

Comment
No-profile-pic
Save or Cancel
Pete McAllister

Pete McAllister, Digital Marketing Executive at Intelligent Car Leasing

Absolutely cracking breakdown of a time based PPC analysis.

Looking at our data in-house there's definitely times that offer much lower/higher costs in relation to acquisition. Looking into weighting our bids in relation to these times is definitely something to look into very soon.

It's strange because the high converting hours don't coincide with what we would have predicted. But if it work it works!

Pete

over 2 years ago

Avatar-blank-50x50

Deri Jones, CEO at SciVisum.co.uk

It's an interesting question - how much of the potential improvement can be gained by weighting in line with a cruder metric like your overall traffic level peak and troughs?

I guess I'm wondering what happens when everyone uses scheduling - if on average most sites play down the same time slots, then demand/Google costs for the better times will increase - will it end up as no net gain?

Or maybe a gain just for those sites who's traffic means they do well outside of the normal peak times - which will get cheaper!

over 2 years ago

Chris Swan

Chris Swan, PPC Manager at Esure

Nicely done Pete, and thanks for your feedback! :-)

Could I ask if you're taking equally weighted or last click sales reports to find your high converting hours? It's vital to compare both of these to find out when the purchasing journey began, as well as when the final purchase was completed - taking assisted sales into account as well; in order to find the true top performers. Both in terms of finding the top hours of the day and top days of the week.

This can be done by filtering for only the PPC channel in your multi channel reports, and then surmising the attribution ratio that PPC has had across each of your online sales.

In this way the bid matrix can be constructed with 100% certainty.

I really hope this helps :-)

Kind regards,

Chris

over 2 years ago

Chris Swan

Chris Swan, PPC Manager at Esure

Hi Deri,

Thanks a lot for your feedback :-)

Really interesting angle you've taken on this - peaks and troughs in traffic levels could also be investigated - comparing against lost impression share to rank and the average cost for every click, potentially for clients looking to improve their brand awareness as efficiently as possible.

For direct response clients however (which was the case in this instance) the use of sales and cost per sale were more relevant.

Very interesting point you made too about the implications on the value of this benefit if everyone analysed in this way. The great thing about an analysis such as this is the best performing hours of each day, and best days of the week would be different per industry. Even different per account in the same industry - as I've run this analysis across different accounts in the same industry and calculated a different bid matrix for each. And each bid matrix worked well!

This brings potential to employ this strategy on any account knowing the benefit will be the same no matter how many advertisers employ this tactic - each has different hours and days to take advantage of, in different ways.

Worst case scenario - the peak hours will level off over the week for the whole sector, and the PPC landscape would be more stable. A little less mountainous, and a little more calmer waters!

over 2 years ago

Manfredi Sassoli de Bianchi

Manfredi Sassoli de Bianchi, VP Marketing at jobinasecond

Well done, you are 70% of the way there :)
Keep up the good work

over 2 years ago

Andrew McGarry

Andrew McGarry, Managing Director at McGarry Fashion

This type of bid matrix is great for maximising profitability in certain circumstances. You could write another post based on what those are.

In my current niche, learning from historical performance is equally as important as finding out how that product category performs in the paid search market overall. You can't grow revenue unless demand is growing or else you're taking someone else's market share.

Basic research using keyword spying tools will tell you what your most profitable KWs are likely to be based on what that paid search niche is doing.

Too many agencies fail to drill competitors for data because they don't invest in the tools. Doing that AND looking at historical performance sets the foundation for better results. I've yet to take over an account that didn't have obvious keyword gaps.

Thing is, a bid matrix could actually decrease your sales volume at the one time of the year where the huge seasonal increase in traffic and sales offsets the ROI/ROAS issue.

Not to mention that tablets and mobile behaviour is arguably making the traditional bid matrix set up too static for bidding wars at peak, depending on how you deal with bid rules.

over 2 years ago

Avatar-blank-50x50

Matt Lovell, Group Head of Customer Insight & Analytics at Thomas Cook AirlinesEnterprise

Really interesting article. Having done something similar in the past however we found we had to look at all the PPC touchpoints that drove / influenced sales rather than just the last click conversions however as otherwise there is a real risk that you are downweighting activity at periods in the day where users are completing their research and therefore directly effecting the chances of them coming across your site...

over 2 years ago

Pete McAllister

Pete McAllister, Digital Marketing Executive at Intelligent Car Leasing

Thanks for the updated advice Chris!

The initial data I looked at was equally weighted i.e. enquiries counting as conversions. As all sales are then phone based after, so I'll need to take a look at the end of the month and correlate the data to see which hours are actually sending truly converting traffic.

Much appreciated,
Pete.

over 2 years ago

Chris Swan

Chris Swan, PPC Manager at Esure

Hi Andrew,

I completely agree! Thanks a lot for commenting! :-)

Each are great points to take into consideration, such as competitor analysis, utilisation of spying tools, as well as structural, keyword and adcopy research all being vital for success. The bid matrix discussed could be an additional tactic.

I completely agree that historical performance is equally important as is knowing how the product category performs in the paid search market overall.

These are all incredibly important factors to take on board when optimising I completely agree, and potentially the bid matrix could be used as well to help performance.

Mobile, geographic and demographic bid adjustments are also absolutely fantastic layers to add to the initial bid matrix. This would upgrade it to an advanced bid matrix.

However as we know the extra multipliers aren't added, they're combined with the original bid matrix multipliers, so a time slot bid adjustment of 10% and a mobile bid multiplier of 20% would combine for the example below:

Tuesdays, 8 to 10 a.m. adjustment: £1.00 x (+10%) = £1.10
Mobile adjustment: £1.10 x (+20%) = £1.32
Final bid for Tuesdays, 8 to 10 a.m, on Mobiles: £1.32

Working backwards using the above formula we can make a static bid matrix even more dynamic!

Seasonality is the big challenge you're also right. And what better way to prepare for it than to take both the seasonality impact last season (to see the spikes in demand coming, and see if the spikes were profitable last year), as well as current performance this year (to take into account the changing business environment, such as new entrants to the market or changing demand).

Finally the potential sales loss from seasonality could be minimised if we pulled back on the hours where historically there were no equally weighted sales recorded - so where significant sales volume failed to be generated via either first click, any touch point along the journey, or last click (both last season and this year). Resulting in more budget available to spend in the more efficient hours, and if there were impression share lost to rank available to grow into for the efficient hours, the same sales could be generated (or increased) using the bid matrix.

As CPA can sometimes define how much investment is available, if the client is happy with the CPA and wishes to grow sales, the bid matrix can be tailored depending on your advertising goals- to push more (if there were impression share lost to rank available to grow into). And if a campaign is losing impression share to budget the daily campaign caps can also be optimised, without affecting the bid matrix.

over 2 years ago

Avatar-blank-50x50

ben

What is CPW? Cost per Web Order? Doesn't make that explicit anywhere. Great post, really interesting.

over 2 years ago

Chris Swan

Chris Swan, PPC Manager at Esure

Good morning Ben,

Thanks a lot for your feedback :-)

Yes CPW in the blog was abbreviating for cost per web order, sorry if this wasn't clear enough.

Thanks again though Ben :-)

over 2 years ago

Avatar-blank-50x50

hauhau

Hi Chris Swan,
thank u so much. The initial data I looked at was equally weighted i.e. enquiries counting as conversions. As all sales are then phone based after, so I'll need to take a look at the end of the month and correlate the data to see which hours are actually sending truly converting traffic.

over 2 years ago

Avatar-blank-50x50

Laura Baker

Thanks for the article. I did notice, though, that you stated that "bid adjustments can be set from -90% to +900%", but the final bid matrix had values of -100%.

over 2 years ago

Chris Swan

Chris Swan, PPC Manager at Esure

Hi Laura, you're very welcome!

Yes all instances of "-100%" were implemented as "-90%" in the interface.

If readers wanted to calculate the final adjustments themselves however for the time slot: 8 am till noon (-50%) on a Monday (-50%) for example - I thought reflecting the calculation as accurately as possible in the matrix ("-100%") would be easier to read, but of course the actual interface implementation would be "-90%".

On the flip side all instances where the calculation results were more than -100%, were stated in the matrix as a "-100%" drop in bid, as I thought this would to be easier to read than the actual interface implementation.

So to compensate for any confusion around the maximum bid adjustment implementation, I stated (please note: bid adjustments can be set from -90% to +900%)

Thanks for your feedback Laura :-) Well spotted!

Kind regards,

Chris

over 2 years ago

Comment
No-profile-pic
Save or Cancel
Daily_pulse_signup_wide

Enjoying this article?

Get more just like this, delivered to your inbox.

Keep up to date with the latest analysis, inspiration and learning from the Econsultancy blog with our free Daily Pulse newsletter. Each weekday, you ll receive a hand-picked digest of the latest and greatest articles, as well as snippets of new market data, best practice guides and trends research.