The ad dashboard shows clicks, and WhatsApp keeps lighting up. People ask, “How much?”, “Is it in stock?” and “How do I order?” Your team replies. Then the conversation goes quiet.
At the end of the day, there are plenty of enquiries but very few orders.
It is understandable to want to change the creative or the audience. Before you do, separate two possibilities: are these people a poor fit for your product, or could some of them have bought if the conversation had answered what they needed to know?
If you run Facebook or Instagram ads, or manage a brand that handles enquiries through WhatsApp, these seven checks will help you investigate both sides. Each one leads to something specific you can review or change.
1. The campaign is optimising for conversations, but the team is judging it on orders
Being willing to send a message is not the same as being willing to pay. If your campaign is mainly optimising for conversations and your team only counts incoming messages, a low cost per conversation is not enough reason to increase the budget.
That does not mean Meta can only find people who want to chat. WhatsApp also describes lead and purchase optimisation options, subject to account eligibility and setup. Check the goal your campaign actually uses rather than treating every click-to-WhatsApp campaign as the same thing. See WhatsApp’s official optimisation guide.
Then follow one cohort of new enquiries further: how many reply again, have a relevant need, receive a suitable recommendation and eventually pay?
For example, imagine one campaign produces 100 conversations and two orders, while another produces 40 conversations and four orders. Those are hypothetical numbers, not campaign results. You would still need to compare spend, order value, margin and refunds. Neither message cost nor order count tells the whole story.
Define what a qualified enquiry means for your business: a relevant need, a location you can serve and suitable purchasing conditions. If someone has only asked about price, their budget is still unknown. Record “not enough information” instead of automatically labelling them a bad lead.
2. The first reply does not continue what the ad promised
Suppose an ad helps people choose between travel-size and full-size products. They click through and receive a company introduction followed by a long product catalogue. Now they have to explain why they came.
Read the ad copy, prefilled message and first useful sales reply together. Are they dealing with the same question?
Here is an illustrative greeting, not a real customer exchange:
Are you looking at the travel set from the ad? It is designed for short trips. Are you buying it for yourself or as a gift?
It picks up the ad’s context and asks one question that could change the recommendation. That is more specific than “How can I help?” without immediately pushing a larger bundle.
Prepare an opening for each main advertising offer. If your inbox cannot identify which ad a person came from, ask them to confirm. Do not have a salesperson or AI guess.
3. The customer received a reply, but not an answer
“Thanks for contacting us. We will reply shortly” acknowledges a message. It does not answer a question about price, stock or how to buy.
Measure the automated greeting separately from the first useful response. Compare working hours, out-of-hours enquiries and different team members. Then check whether people who waited longer continued the conversation. A delay is worth investigating, but the gap between timestamps alone does not prove why someone left.
If nobody covers the inbox at night, state your service hours clearly. A tested automated flow can handle basic questions that have reliable answers. Uncertain stock, unusual commitments and complaints need a clear route to a person who can help.
AI is useful here when it answers within known boundaries. A fast answer that has not been verified is not a better customer experience.
4. The first response contains too much to deal with
Someone asks one question. The salesperson sends product details, instructions, delivery information, payment methods and three pictures. Everything may be relevant eventually, but the customer now has to decide what to read and answer first.
Try a simpler order: answer the question, add essential context, then ask one question that matters for the next step.
If someone asks whether an item is in stock, and stock has been confirmed, an illustrative reply could be:
Yes, it is in stock. Do you mean the travel set in the ad or the full-size set? I can send the matching sizes and prices.
Do not withhold the first answer just to make the customer send another message. And if they have already said they want to buy, do not put them through a fresh discovery interview. Give them clear buying instructions.
5. The quote is just a number—or the customer cannot get a price at all
Replying “$69” to “How much?” leaves the customer unsure what they are buying. But “Answer a few questions first, then I will tell you the price” can be frustrating too.
Give the verified price and what it covers, then ask a necessary follow-up question.
This is a fictional writing example. The price, contents and shipping terms do not describe an actual product:
The set is $69 and includes a cleanser and moisturiser. Shipping is extra. Are you looking for both, or just replacing one item? I can send the individual prices too.
If the price depends on the option, explain the range and how it is calculated. Once you have the required information, provide an exact quote. The point is to give the price context, not delay revealing it.
For categories such as skincare, do not turn a product recommendation into a diagnosis or promise a guaranteed result to justify a purchase.
6. Ready buyers are still hearing the introduction, while uncertain buyers are being pushed to order
“I will have a look” could mean the price is wrong, the options are unclear or the person is simply busy. Do not decide which without evidence.
Instead of relying only on high-, medium- or low-intent labels, record two things: what has been confirmed and what remains unanswered.
- They need the travel set but do not know the delivery time: explain delivery conditions next.
- They asked how to pay: provide the correct buying steps and involve a person if needed.
- They are unsure about suitability: check verified information and escalate anything you cannot confirm.
- They have only asked for a price: their budget and intent remain uncertain.
These notes help the next team member continue without starting over. They also reduce the risk of AI taking every customer through the same long script.
7. Follow-up has become “Are you still interested?”
If the conversation stopped at delivery, the next useful message may be delivery information. If it stopped at choosing a bundle, clarify the difference instead of resending the entire catalogue.
Start with what the customer actually said, then decide whether there is something useful to add. Silence does not tell you they are waiting for payday, worried about results or seeking a family member’s approval.
Define stopping rules too. Stop when someone declines. Do not let an automated sequence keep pushing for an order while a person needs to resolve the issue. If there is nothing new to say, repeated reminders are not a substitute.
When using the WhatsApp Business Platform, follow the applicable permission, template and customer-service-window rules. Outside the 24-hour window after the customer’s last message, ordinary free-form messages are not an unrestricted follow-up channel. Use approved templates where required and respect opt-out requests. See the WhatsApp Business Messaging Policy.
Where these checks fit in SalesNumberOne
SalesNumberOne focuses on what happens after an enquiry arrives: understanding needs, recommending products, addressing concerns and involving a person when needed. The examples below use the actual admin interface observed on 12 September 2026. They show configuration and simulation screens—not customer sales results or proof that every account has the same capabilities enabled.
Rather than starting with “Reply like a great salesperson,” I would start with these three tasks.
First, make the product information accurate enough to quote from
Prepare more than a product description. Include sizes, prices, bundle differences, shipping conditions, after-sales policies and common questions. Any customer proof needs a genuine source and permission to use it. Outdated prices should not remain in the reference material.
This connects directly to the pricing problem above. A natural tone cannot repair a factual gap. If the system cannot verify something, it should say that confirmation is needed instead of filling in the blank.
Open Knowledge Centre in the left navigation. The observed interface has sections for Products and Bundles, Business Knowledge, Sales Knowledge and Media Library. In Products and Bundles, check whether an item is published and whether AI may recommend it. The page states that AI uses published content only: saving a draft and publishing are different actions.
For example, you may want AI to discuss an individual product while that item is not enabled for individual recommendations. Check your selling rules first rather than simply telling AI to be more flexible. Organise policies, sales questions and verified media separately instead of squeezing them into a greeting.
Figure 1. The actual Products and Bundles screen, shown in its original Chinese interface. Product thumbnails are obscured and customer information is not shown. This was a read-only inspection; no recommendation settings or knowledge publications were changed.
Second, connect the advertising promise to clear conversation rules
For a travel-set campaign, prepare rules such as: explain the size when someone asks about the travel set; quote the price when asked; do not push a bundle on someone who only wants one item; ask a person to confirm uncertain stock.
The observed route is AI Business Planning → I Already Have a Sales Strategy. Its four steps cover choosing an industry template, describing the sales approach, setting the AI identity and simulating a customer conversation. The page says changes saved in the first three steps take effect immediately. Review the scope before changing a live account; do not treat those steps as an isolated draft area.
To inspect replies, use the fourth step, Simulate Customer Conversation. It is marked simulation-only and says no real messages are sent. Successful simulation rounds consume credits, so read the on-screen terms before testing.
Three useful test questions are “How much?”, “I only want one item” and “Can you guarantee this suits me?” Check whether the response gives prices transparently, respects the request and avoids unsupported promises.
Figure 2. The original Chinese simulation interface. Conversation text, timestamps, the account identity, merchant-specific identity and usage record are obscured. It shows where to inspect a response, not the result of a new test or evidence of conversion performance. No simulation was sent for this article.
Third, write follow-up as a bounded strategy, not just a reminder
The observed Automatic Follow-up page separates Automatic Replies, AI Strategies and AI Follow-up Plans. Event-triggered immediate replies and follow-up after silence need different treatment. Do not put the first greeting and “what to say if they stop replying” into one undifferentiated setting.
The AI Strategies detail view contains a goal, expression logic, prohibited wording and allowed media. Those four fields are a useful way to review the seventh gap: does the message add information, or merely demand a reply? What must it not say? Are images allowed, and are they genuine and authorised?
The strategy shown below asks about ordering directly but without pressure. Its restrictions prohibit guarantees, pressure, medical claims, price changes and unauthorised links. Those are clearer boundaries to review than “try harder to close the sale.”
Figure 3. A cropped view of Automatic Follow-up → AI Strategies, in the original Chinese interface. It contains no customer list or account identity. The page inspected also reported test mode with real sending prohibited. No settings or sends were activated, saved or executed; this screenshot does not establish that the account is sending automated follow-ups.
Use what you learn from conversations to improve one rule at a time. Do not change the ad, price and response flow together. Stop when the customer declines, and never let a sales strategy override channel permissions, service windows or template requirements.
This can make your sales process clearer. It cannot guarantee an increase in sales. Nor does this article claim that advertising attribution, a ROAS dashboard or automatic purchase-event feedback to Meta is an available SalesNumberOne capability.
Review a batch of conversations before deciding what to change
Start with 30–50 new enquiries from the same period, including both buyers and non-buyers. That is a practical starting workload, not a statistically representative sample requirement.
Keep returning customers separate when assessing new-customer ads. Give each cohort a consistent observation period: an enquiry from yesterday has not had the same opportunity to convert as one you have followed for two weeks.
- Many ad clicks but few messages: check the WhatsApp entry point, prefilled message and advertising promise. This alone does not establish a sales-response problem.
- Few replies after the first message: check useful response time, relevance and information overload. Do not immediately conclude the leads are poor.
- Clearly mismatched needs: review the creative, service area, audience and price expectations. More sales questions cannot make every lead a fit.
- Silence after a recommendation or quote: review the reasoning, fees and purchasing obstacles. Do not assume price is the only concern.
- Real orders with no matching ad report: investigate event feedback, attribution windows and reporting definitions. Missing reporting is not proof that the ad produced no orders.
Choose one recurring issue that you can change. For example, replace a price-only response with “price + what is included + one necessary question.” Do not change the budget, creative, bundle and entire sales process on the same day.
If WhatsApp is already receiving a steady stream of enquiries, increasing the budget is not necessarily the next step. First establish how those enquiries are handled and where conversations stop. That tells you where the next hour of effort—and the next part of the budget—belongs.
If you have not checked the entry point yet, start with our Facebook Page and WhatsApp setup guide and click-to-WhatsApp advertising guide, then use the checks above to review what happens after the click.
Frequently asked questions
Should I pause an ad if WhatsApp messages are high but orders are low?
Not based on message count alone. Check spend, orders and profit, then investigate the entry point, lead fit and conversation process. If losses exceed what you can afford, control spending first. Do not keep increasing the budget simply to collect more chat samples.
Should I avoid giving the price if people disappear after asking?
No. State the correct price, specification and necessary fees, then ask a question that genuinely affects the recommendation. Do not make answering extra questions a condition for learning the price.
Does faster AI response automatically mean more sales?
No. Accuracy, product fit, clear buying instructions and the competitiveness of the product and price all matter. Test response quality first, then assess the outcome using actual order data.