Spark

Say hello to Spark, your very own Fusion AI assistant! Ask it anything about your account in plain, everyday language, and it will dig through your live data and our docs to find you an answer. Think of Spark as a teammate who knows Fusion inside out. It’s great for things like:

• Keeping on top of performance. How is revenue tracking this week? Which publishers are earning the most? Why did fill dip on Tuesday?
• Checking the health of your setup. Are there any tag errors? Is anything missing from ads.txt? Are there onboarding steps still to finish?
• Finding your way around. What does eRPM mean? Where can I see revenue by demand partner? How do I set up a new ad stack?

​

Asking Spark a question

You’ll find Spark wherever you are in Fusion. Click the Spark button in the bottom-right corner of any page to open a chat, or press cmd+k (mac) or ctrl+k (windows) to open search, type your question and choose Ask Spark. If your search doesn’t match any pages, just hit enter and Spark will take it from there. Spark opens as a sidebar by default. If you’d rather it floated over the page, or took over the whole screen with your chats listed alongside, switch it with the display mode button at the top of the chat. On smaller screens, Spark always opens full screen. Not sure where to start? Each page suggests a few questions to get you going, such as Explain this page or Summarise this page. Spark always knows which page you’re on, including any filters you’ve applied, so you can simply ask “How does this compare to last month?” and it will know exactly what you mean.

​

What Spark can (and can’t) do

Spark is here to look things up and explain them, so it won’t make any changes to your account. If you ask it to create, change or delete something, it will point you to the right place in Fusion and walk you through the steps so you can do it yourself. Spark also sees exactly what you can see, and nothing more. It answers for the organisation you’re working in at the time, and can’t see data belonging to anyone else.

​

Getting the most from Spark

A few tips to help you get great answers:

• Be specific. The more Spark knows, the better. “What was eCPM for example.com last week?” will get you further than “How are we doing?”
• Keep the conversation going. Spark remembers what you’ve talked about, so follow-ups like “And the week before?” or “Break that down by ad unit” work a treat.
• Narrow down the big ones. Questions that cover every publisher in a large network take a little longer. If an answer is slow, or only covers part of what you asked, try narrowing it to a publisher, a date range or a metric.

Spark speaks your language too! It replies in British English by default, but ask in another language and it will reply in kind, or ask it to stick to a particular language from then on. Money is always shown in your organisation’s currency, and Spark will let you know whether revenue figures are gross or net.

​

Your conversations

Every chat is saved, so you can pick up where you left off at any time. Start something new with New question, and keep your list tidy with Rename chat and Archive chat. Want to share an answer with a colleague? Use Copy on a single answer, or Copy chat for the whole conversation. There’s no need to sit and wait while Spark works. Switch to another chat or page and it will carry on answering in the background. If an answer ever fails, just hit Try again.

​

Something not quite right?

We’d love to hear about it! If Spark gets something wrong, use Copy chat and send the conversation over to us via support, along with what you were expecting. Seeing the whole conversation helps us make Spark even better. Spark is in beta, so you get full access while we fine-tune it. We’ll keep you posted on what’s coming next.