Why ecommerce tactics fail in lead generation, and the reverse

Eleven leads in a month. That was the whole data set an engineering firm handed us, along with a request to work out why their cost per lead had gone up and what to change this week.

Nothing could be changed this week. Eleven is not a number you can read. But the account had been changed every week for a year by someone who had previously run paid search for an online shoe retailer, where eleven orders is Tuesday morning.

That is the most expensive thing in paid search: advice that was true for a different business model. Ecommerce and lead generation share an interface, a bidding system and a vocabulary. They share very little else.

Where the two models actually split

In ecommerce the conversion is the transaction. Money arrives, the value is known to the cent, and the loop closes in minutes. In lead generation the conversion is a promise to talk. The real outcome happens days or weeks later, off the platform, and its value swings enormously. One form fill is worth nothing. The next is worth half a year of revenue. Google counts them one each.

Volume, data quality and delay all differ by an order of magnitude. Any tactic that depends on fast, dense, accurate feedback will work in one and flounder in the other.

Ecommerce habits that break lead generation

• Treating every conversion as equal. Fine when each one is a sale with a number attached. Ruinous when a quote request from the wrong country weighs the same as a qualified enterprise enquiry. Without values on different actions, the bidding optimises towards whatever is easiest to get, which is always the cheap lead.
• Chasing conversion volume. More orders is more revenue. More leads can be less revenue, because sales spends the week on rubbish and lets the good one go cold. There is a capacity ceiling, and above it extra leads have negative value.
• Fast iteration on short windows. A store doing hundreds of orders a week can read a seven-day test. Eleven leads a month cannot. Reading noise as signal and editing the account every Monday is the most reliable way we know to wreck a lead generation account.
• Discount creative. Twenty per cent off a product works. Twenty per cent off a consultation reads as desperate and recruits people who will negotiate everything afterwards.

Lead generation habits that break ecommerce

• Keyword-level control as the main job. Retail search is enormously long-tail and the shopping surfaces are driven by product data, not your keyword list.
• Ignoring the feed. The product data is the campaign. Google’s product title specification allows up to 150 characters, tells you to use all of them, and notes that shoppers typically read only the first 70, so the important words go at the front. It also rules out the things retailers love putting in titles: price, sale dates, shipping terms, capitals for emphasis. If you generate titles with AI you are meant to use the structured_title attribute and declare it. Most feeds we open are breaking at least two of those rules, and fixing a title costs nothing.
• One efficiency target for the whole account. A store selling €20 accessories alongside €2,000 equipment cannot run to a single target. In lead generation, a single cost per lead across one narrow service line is often perfectly reasonable.
• Optimising to conversion count rather than value. Ten cheap items is not one expensive one, and only value-based bidding can tell the difference.

The discipline that actually transfers

Both models are ruined by measuring what is convenient instead of what matters. In retail that means counting revenue and ignoring margin and returns. In services it means counting leads and ignoring which ones closed. Whenever we are shown an account with excellent platform metrics and a business that is not growing, this is nearly always it: the automation is optimising hard towards a number that does not represent money.

Remember those eleven leads. The firm had treated it as a volume problem for a year. It was not. One of the eleven had become their second-largest contract, and nothing in the account knew that. Google had been handed eleven identical events and had optimised, obediently, for the kind of lead it could get most of, which was the kind that never answered the phone. Eleven was never too little data. It was eleven pieces of the wrong data.

Feeding the right data back

For lead generation, the mechanism for this is enhanced conversions for leads, which matches hashed customer data from your form against the click that produced it, so you can upload what happened later. Google’s documentation on enhanced conversions for leads is also where the current deadline sits: from 15 June 2026 these uploads moved to the Data Manager API and are blocked in the Google Ads API. Google has also said the web and leads settings will be combined into one, so check which of the two yours is using. If your CRM integration was built before last spring, somebody needs to check it still runs.

The rest of the lead generation setup is unglamorous. Write down what a qualified lead is before you build anything, and measure that rather than raw form fills. Give a booked call and a brochure download different values, even rough ones. Judge performance over 60 to 90 days and accept being uncomfortable for the first six weeks. Then track how long it takes you to ring a new lead back, because in our experience that moves outcomes more than most things you can change in the account.

For retail: fix the product data, segment by price band or margin rather than product category, bid to revenue or ideally to profit, exclude what you cannot fulfil, and watch the new versus returning customer mix. A return on ad spend built entirely on repeat buyers is not growth, it is a loyalty programme with a media budget. Some of that overlaps with the measurement work we do before touching a campaign.

One we have not solved

Some businesses are genuinely both. A manufacturer selling €40 spare parts online and quoting €200,000 installations through a form. The textbook answer is separate accounts with separate goals, and we have built it that way, and it still goes wrong, because the spare parts buyer and the installation buyer are often the same company at different moments and the attribution never knows which conversation it is in.

We do not have a clean answer to that one. If you run that kind of business, we would genuinely like to hear how you have split it.