Paid social for parts: where it works, where it quietly burns margin

Paid social arrived in automotive retail on a promise borrowed from apparel: cheap reach, fast creative iteration, a direct line into a buyer's phone. Four years of that experiment later, the honest finding is narrower and more useful. Social is very good at starting a purchase conversation and mediocre at finishing one, and in a category where the first question is whether the part fits a 2019 F-150 with the 3.5L, finishing is most of the work.

That does not make it a bad channel. It makes it a specific instrument with a specific job, and the teams getting value from it are the ones who stopped asking it to behave like search.

What the benchmarks actually say

Pulled from the 2025–2026 ecommerce medians rather than from a case study someone screenshotted:

The Meta figures are Triple Whale's ecommerce medians for the twelve months to July 2026; the TikTok benchmark is Billo's April 2025 ecommerce cut; the automotive CPC comes from LocaliQ's 2026 benchmarks. Read them as order-of-magnitude, not gospel.

Two things follow immediately. First, a $38.99 acquisition cost against a $73 average order leaves roughly $34 of gross margin before cost of goods, payment fees and shipping. On a 30% margin accessory that is a break-even first order at best. Second, the click is cheap while the intent is thin, which is the exact inverse of a part-number query on Google. You are buying attention from people who have not yet admitted they need a part.

Why the economics break on hard parts

• Fitment risk lands on your P&L. A wrong rotor is not a small refund. It is return freight, an open-box unit, a one-star review and a support thread.
• Carts are single-item and low value. A $30 pad set or a $19 sensor cannot absorb a $39 acquisition cost. Apparel bundles; parts usually do not.
• No bay attached. Nothing in a social checkout upsells installation, inspection or the second axle, which is where fixed ops actually makes money.
• The attribution is generous. View-through windows, modeled conversions and consent gaps all inflate platform-reported ROAS. If your analytics lost a fifth of sessions to consent, your denominator is already wrong. The report format we hold ourselves to starts from reach for exactly that reason.

Where paid social earns its place

There are five jobs in this category where social is genuinely the right tool.

1. Accessories, lighting and appearance. SEMA's 2025 research puts half of specialty-equipment buyers under 40, and Amazon's own data shows automotive lighting growing 26% year over year. These are wants with photography, not needs with fitment anxiety. Short-form video sells a light bar far better than a text ad can.

2. Demand with a date. Winter tire changeover, truck month, detailing season, towing season. Creative can carry urgency in a way a product feed cannot, and the offer does not need a catalog rebuild to run.

3. Service and fixed-ops offers inside a radius. A geo-fenced appointment campaign for brake inspection or a 60k service converts on a booking, not a part. That is a real, measurable outcome, and it compounds into bay utilization instead of a one-off order.

4. Retargeting the fitment drop-off. People who selected a vehicle, viewed a part and left are the cheapest incrementality available. This only works when the catalog feed behind the dynamic ad is accurate. A retargeting ad for a part the shopper already discovered does not fit is worse than no ad.

5. Creator and install content as a demand asset. The same footage that runs as a paid placement also becomes the reference material a buyer checks before purchase, and increasingly the material an assistant summarises. It has a second life that paid search does not.

Creative is the targeting now

Broad targeting and automated placements moved the work out of the ad account and into the asset. In this category that means a few non-negotiables:

• Put the vehicle in the first two seconds. Year, make, model, engine. Not "premium quality parts."
• Show the install. Torque spec, connector type, bracket, the thing that goes wrong.
• Say the part number out loud and put it on screen.
• Keep the claim identical to the product page. Every mismatch between creative and PDP becomes a return.
• Film in a real bay. Shops trust other shops; nobody trusts a stock handshake.

The catalog is the campaign

Dynamic product ads inherit every flaw in the feed: missing GTINs, absent fitment attributes, out-of-stock variants, price mismatches, suppressed SKUs. A catalog with 30% incomplete fitment data gets 30% of the reach and 100% of the returns. Before the next creative test, run the boring audit.

Measurement that survives 2027

Platform ROAS is a directional number, not a decision. The decision comes from four things: a geo holdout or incrementality test, contribution margin per order after returns, new-customer share versus repeat, and offline conversion tracking for service bookings. If a channel cannot be turned off in one market for four weeks and compared to another, you are not measuring it, you are describing it.

A 60-day test that answers the question

1. Pick one category with photographic appeal and complete fitment coverage: lighting or accessories, not braking.
2. Build one vehicle-specific creative set, three angles, two formats, filmed in a bay.
3. Run two matched markets, one dark, same spend window, same offer.
4. Judge on contribution margin after returns, plus the incremental branded-search and direct-traffic lift in the test market.
5. Kill it or scale it. Do not carry a permanently ambiguous line item.

The 2027 wrinkle

Social starts the question. Answer engines finish it. Someone watches a brake kit install on a Reel, then asks an assistant which kit fits their truck and whether it includes hardware. The assistant reads your product page, your schema and your fitment data. Not your ad.

That is the argument for sequencing rather than for silence. Paid social on a catalog that machines cannot read buys attention you have no way to convert. Paid social on a clean one compounds: the creative drives the query, the product data answers it, and the same attributes that won the impression win the recommendation.

Paid social is not a parts channel. It is a demand-start channel that happens to sit next to a parts business. Treat it that way and the budget becomes defensible.