To drive campaign performance it is just as important to give the system time as it is enough budget. For a first campaign we would recommend a 4 - 6 week flight and at least a £5k budget. You should also be prepared to allow for the entire campaign flight plus your conversion pixel loopback windows (if you are conversion tracking) before you will see the true performance of the campaign.
It’s easy to get nervous in the first week or two of a new campaign if performance is not where you want it to be but making too many changes, too often, or even stopping and starting the campaign can be the worst thing you could do for the machine learning optimisation. The performance at the end of week one is not usually anywhere near where it will end up once the flight and look-back windows have taken place. It is really worth seeing the entire campaign investment through before making a decision about whether programmatic works for you. Our team of experts are also here to help guide you through this process.
Optimising a Live Campaign
Optimisation is a balance between driving performance and maintaining volume. You need your campaign to pace as well as perform.
Our AI machine learning and predictive modelling takes care of a lot of campaign optimisation for you, and analyses thousands of combinations of data points to drive performance based on your set optimisation strategy.
The very basic principles of manual optimisation are:
Try not to restrict your initial set up too much with tight targeting
Trim - Analyse and remove things that are not working as the campaign runs
Upweight - or reallocate budgets to strategies or targeting that work especially well
Pricing - Make sure you are paying the right prices
Try not to make too many changes too frequently or you won’t understand the true impact of your optimisations (or what you might need to revert to get back to your previous baseline)
When analysing your activity for trimming, ensure you set a meaningful impression and spend threshold. Can you really make a decision about whether a site is performing for you if it has only delivered 1000 impressions? Your efforts will feel pretty fruitless if you only remove activity with very low spend levels - this will likely not have much impact on your buying pattern.
Where to look first:
Are you happy that you are counting the campaign metrics campaign accurately (this could be clicks - macro based, or conversions - pixel based). Check your implementations (and the numbers you are seeing on your pixels with your internal numbers) if you have any doubt.
Take a data type to review - Look at your spend and impression levels across these - ensure the allocation makes sense to you based on performance and inventory quality. Machine learning is very clever but can sometimes latch on to the wrong thing, so a human eye is invaluable for monitoring around the inventory that’s right for you.
Devices: make sure all your delivery is not going to mobile, especially if you are not seeing conversions here.
Suppliers & Domains: check for delivery against supply and domains and remove any that have significant delivery, but are not performing, as well as any you don’t like the look of.
Other Levers for optimisation:
Budgets: Is your budget allocated across your strategies in the best way?
CPM: Is your CPM too low or too high, try adjusting this gradually to see impact. It’s better to start high to get traction and then start to bring this down. Keep an eye on your win rates (the percentage of your bids that you win and serve as impressions) as a rough guide. A good win rate is 40-50%. If you are coming in below this you may want to raise your CPM, but bear in mind this can also be affected by block lists on the publisher side and inventory availability.
Frequency Cap: We find best practice for frequency capping is 1-3 per day for prospecting and 3-12 per day for retargeting, but you can also try testing this at different levels.
CTR / CVR / CVR+CTR / CCVV: Have you got the optimum optimisation type set for your strategy? See campaign strategy 101 for more details on this.
Other Data: Look at as much data as you can (geo, context, creative etc). Can you see any specific targeting parameters that are working very well - if so, think about upweighting these. You can do this by creating a new Line Item. On the flip side, can you see anything that is being allocated more spend than you think makes sense - if so, remove them. It can also be worth trying to look at different targeting types together, for example, sliced by device, category and supplier, but this can become quite granular so again make sure you are setting a meaningful impression and spend threshold to make decisions about.