LinkedIn Ad Optimization Algorithms

Optimization algorithms ($800k+ saved)

We've developed an ad optimization algorithm which achieved +40% budget efficiency. For our pilot client spending $2m/year, that means $800k in savings or ~$2.4m in additional revenue annually. We used a small live dataset, so it's an extrapolation, but the results are promising.

Methodology

We recreated a simulated LinkedIn environment from an A/B test that our pilot client had previously ran. In their test, they spent £2406 evenly across 5 creatives in 14 days. Unfortunately we cannot share the dataset. If you email us at hi@bigged.com we may be able to discuss data sharing with our client.

‍

We then created multiple optimization algorithms based on the same technology that Meta (Facebook Ads) uses to automatically optimize spend across ads. LinkedIn doesn't have this same technology yet which is surprising.

Here are the results (higher is better). Yellow is A/B testing (what the client currently does). Brown is our algorithm. Blue is a fictitious "oracle" algorithm which can see the future:

‍

This graph can be interpreted in a number of ways. One such way is in terms of budget savings. Here is the message we sent to our client alongside this graph:

In your A/B test you spent £2406. Our algorithm achieved the same number of clicks by spending only £1466. So it saved 39% of budget and identified the winning version quicker. We're now working on the tech which finds insights and answers questions like "Why is ad A good and ad B bad?".

‍

The algorithm we've developed is based on a new technique we have implemented which far outperforms existing standard techniques and is native to LinkedIn. While we can't share details publicly due to the proprietary nature of the technology, if you email use at hi@bigged.com, we can share more details privately.

Immediate next steps

There is much more we can do to improve the ad optimization algorithms. Here are the steps we are taken in the next few weeks:

• We are increasing the size of our dataset - We have just launched a new A/B test spending £3500 across 14 days to create a new dataset for further study.
• Live optimization algorithm testing planned - We will be running our optimizer live on LinkedIn with a £3500 budget across 14 days and benchmarking it against a standard A/B test that will be launched at the same time with the exact same targeting and creatives.
• We've hired an machine learning engineer - This engineer, Denis, previously worked at Expedia Group performing parametric ad bidding optimization. He will expand the team to help us deepen our research.

Do you run ads?

If you run ads on LinkedIn or Meta, please feel free to get in touch. We're pioneering ad optimization on LinkedIn and Meta for B2B, DTC and ABM.