Exact vs Broad on the Same Term in Apple Search Ads: Running a Clean Test Without Double-Counting Spend
Most indie developers meet match types in a hurry. You add a keyword as exact because it sounds safe, you add a similar phrase as broad because it sounds like it might find more, and a few weeks later the numbers look confusing. Maybe the broad one spends more, or the exact one converts better but gets fewer taps. The honest answer is often that you were not comparing two strategies. You were letting two keywords compete for the same searches.
This post is about running a clean test of exact versus broad on the same underlying term, so you can decide which behavior you want and keep your reports readable.
Why the same term can show up twice
On Search Results, you add keywords to an ad group with a match type of exact or broad. An exact keyword only enters the auction when a search matches its phrase closely. A broad keyword enters the auction for a wider set of related queries, including ones that contain your phrase with extra words or close variants.
That means a query like "budget tracker" can be eligible for an exact keyword "budget tracker" and also for a broad keyword "budget tracker app" or "budget tracker" alone. If both live in the same campaign, you have two bids chasing one tap. Depending on how the keywords are set up, the more specific one usually wins the query, but you still need to check that in your own data rather than assume it.
It also matters where the two sit. If they are in different ad groups within one campaign, you can still end up competing with yourself, and the reports will show spend in both places without telling you the full story.
Decide what question the test answers
Before you change anything, write down the question. Good questions look like:
- For the term "habit tracker", does a broad keyword bring in searches I would never have typed into an exact keyword, and do those searches convert at a reasonable cost?
- Is exact paying a lower CPT and better CPA on the core phrase, enough that broad only makes sense for the long tail?
- Does broad mostly re-buy traffic that exact already captures?
If you cannot state the question in one sentence, you will read the output as whatever you hoped to see.
Set up the test so the two arms do not overlap
The cleanest setup is to separate the arms by ad group and make sure the overlap is intentional:
- Put the exact keyword in one ad group with its own bid.
- Put the broad version in a second ad group within the same campaign, and add the exact phrase as a negative keyword in that broad group if you want broad to handle only what exact does not already cover.
- Keep the country the same for both, because one campaign targets one country or region and mixing storefronts would make the comparison meaningless.
- Start both with a bid you can justify, not an arbitrary number. A reasonable approach is to use the same max CPT for both at first, so differences come from match behavior rather than from one arm simply being funded more aggressively.
If you do not want to use negatives, you can accept the overlap, but then you must read the results as "combined exposure to this term," not as two independent experiments.
Read the right metrics for each arm
For each arm, look at the same set of numbers over the same date range:
- Impressions and TTR. Taps divided by impressions tells you whether the query is relevant enough to earn a click from your listing. A broad arm with lots of impressions but a low TTR is often pulling in loosely related searches.
- Actual CPT. This is what you paid per tap, not your max. Compare the two arms on actual CPT, because one arm may be bidding the same ceiling but winning cheaper taps.
- Installs and conversion rate. Installs divided by taps shows how well the listing converts for that traffic. If the broad arm converts far worse, the extra reach may not be worth it at your current bid.
- CPA. Cost per install is the first cost signal Apple gives you directly. If you have revenue data, look at ROAS later, but do not judge a two-week test on purchases that have not had time to appear.
Remember that Apple's attribution resolves within about a day, and revenue tools such as RevenueCat map the install-to-purchase chain. Give the test enough time for both installs and early purchases to show up before you call a winner.
Watch for the search term report
The search term report is where the test becomes concrete. Pull the queries each arm actually received. You may find that broad is almost entirely re-buying queries exact already gets, which means it is a waste, or that broad brings in a cluster of phrases you never thought to target, which means it deserves its own keyword list. Either outcome is useful, and both are invisible if you only look at the aggregate.
An illustrative example
Suppose you run a meditation app and test "sleep meditation" exact against "sleep meditation" broad in a separate ad group, with the exact phrase negated in the broad group. Over a two-week window, the exact arm might show a higher TTR and a lower CPA, while the broad arm shows more impressions and a few new queries like "bedtime relaxation audio". The numbers here are invented for illustration, not benchmarks. The useful takeaway is the shape of the trade: exact is your efficient core, and broad is your exploration budget that needs its own review, not a silent extra layer on top of the core.
Common mistakes to avoid
- Changing bids on both arms in the same week, so you cannot tell which change moved the result.
- Calling a winner after a handful of taps. Small samples swing wildly on CPA.
- Forgetting the country. A term that behaves one way in one storefront may behave differently in another.
- Leaving the overlap unnamed. If you keep both live without a plan, the reports will keep double-counting the same searches.
Takeaway
Exact and broad are not rival philosophies. They are two ways of spending on the same demand, and they only stay readable when you separate them on purpose. Pick one question, separate the arms with negatives or ad groups, compare actual CPT, TTR, and CPA on the same dates, and read the search term report before you decide. Then make a single change at a time.
If you find this kind of side-by-side reading tedious across many keywords, AdsBuddy is built to surface the overlaps and the few changes worth making each day, while you review and approve every one of them yourself.