The best time to invite a new guest back is after their first completed purchase, while the experience is still relevant, but before the invitation feels like an immediate follow-up. Start with a short, defined window. Check that the guest can be identified and contacted through their chosen channel. Remove that guest from the campaign if they return before their message goes out.
For many restaurants, the harder decision comes first: determining whether a purchase was actually the guest’s first. A receipt alone cannot answer that. Build a dependable audience from the records you have, then compare a straightforward reminder with an offer that has a clear cost and purpose.
Define a first purchase you can recognize
Choose the event that starts the clock. A practical definition is a completed, paid purchase that can be matched to a guest record and has no earlier completed purchase in the history you can reliably review. Exclude canceled orders, unpaid orders and test transactions. Decide how to handle refunds before building the audience; a fully refunded order is usually a poor starting point for a return invitation.
Write down which sales channels are included. If your records cover direct online orders but not counter purchases, a person’s first online order may not be their first restaurant visit. Call the audience “first recorded direct-order customers” rather than claiming they are all first-time guests. That distinction matters when you interpret results.
Identity needs its own rules. A usable record should have a stable way to match the original purchase to a later purchase and enough contact information for the intended channel. An email address may let you recognize an email recipient, but a second purchase under a different address may not match. A phone number may identify an SMS recipient, but only if the number is accurate and connected to the purchase record.
Exclude records with missing or conflicting identity data from the send audience. Keep a count of those exclusions so you know how much of the new-customer picture is missing. Also avoid merging two records simply because they share a household phone or a similar name. Ask how the software handles duplicates, shared contact details and changes to an address or number.
Having contact details is not the same as having permission to market. Before a record enters an email or SMS campaign, check the guest’s preference for that channel and whether the planned contact is appropriate. Give guests a way to stop future messages, and make sure that preference is reflected in subsequent sends. Ask a qualified adviser about applicable requirements for your restaurant and channels.
If you use a restaurant ordering system, ask which customer and order fields can be used to identify a completed first purchase. Then ask whether those fields can be matched to later purchases without creating duplicate guest records.
Choose a cohort and a contact window
Build a cohort from a fixed period, such as customers whose first recorded purchase occurred during one week. A fixed start and end make the audience easier to review and give every guest a comparable opportunity to return. Record the purchase date, eligible channel, planned send date and the end of the measurement period.
Choose the delay around the occasion your restaurant serves. A weekday lunch shop may have a reason to invite someone back sooner than a special-occasion dining room. Consider how often a guest could reasonably use the restaurant, whether the next visit needs a reservation, and whether your team can handle extra orders during the proposed window.
Use the following worksheet as a hypothetical example, not a prediction of campaign results. A neighborhood lunch restaurant reviews 100 first recorded direct-order purchases from one week. Twenty records lack a dependable match or usable contact detail. Another 20 have no marketing permission for the intended channel. That leaves 60 eligible guests. Before the planned send, six return on their own, so the campaign audience is 54.
| Decision | Working choice | What to check |
|---|---|---|
| First purchase | Completed direct order in the cohort week | No earlier completed direct order in available history |
| Eligibility | 60 of 100 records after identity and permission exclusions | Record each exclusion once and honor channel preferences |
| Contact timing | Seven days after the first purchase | Remove the six guests who returned before their send date |
| Measurement | Track the 54 remaining guests for 28 days after their planned send date | Count matched, completed second purchases separately from message activity |
The table is a decision record: it identifies the starting event, audience, timing and outcome to review. The seven-day delay is a starting choice for this lunch restaurant, not a universal rule. Compare it with a shorter delay, which may feel too soon, and a longer delay, which gives guests more time to return without a prompt but may make the first experience less immediate.
Set a cutoff for data changes. The example removes guests who return before their individual send date, rather than waiting until the entire week’s cohort is ready. Also decide what happens if a guest withdraws permission, an order is canceled or two records are merged while they are waiting. The safest operational rule is to recheck eligibility immediately before sending.
Compare a reminder with an offer
A reminder is the simpler starting point. Thank the guest for their order, mention a useful reason to return and give them one clear next step. A lunch restaurant might highlight its weekday menu or direct-order page. Keep the message specific enough to be useful without assuming what the guest liked or why they ordered.
An offer adds a financial reason to act, but it can also discount a visit that would have happened anyway. Before choosing one, define the reward, the eligible purchase, the expiration date and any exclusions. Make those terms easy for the guest and the staff to understand. A code that works only on certain orders should not be presented as valid for every visit.
In a separate hypothetical example, suppose a second order has $18 in sales and the restaurant considers a $3 discount. The discounted sale would be $15 before any other adjustments. If 12 guests use the discount, the discounts total $36 and those orders bring in $180 after discounts, rather than $216 at the undiscounted amount. Those figures exclude food, labor, payment processing and any campaign costs; they are sales and discount amounts, not profit.
Compare that cost with the reminder alternative. The reminder has no discount cost, though message delivery and staff time may still cost money. Ask vendors for a written quote that distinguishes platform charges, message charges and any setup costs. Do not count the $180 in discounted sales as campaign profit or assume all 12 visits were caused by the offer.
For a fairer comparison, consider assigning eligible guests to reminder and offer groups before sending. Use the same eligibility rules and measurement period for both. If the groups are small, a few orders can change the apparent result substantially. Treat an early difference as a reason to learn more, not proof that an offer will always outperform a reminder.
Ask whether the system can schedule messages, suppress guests who have already returned, respect channel preferences, limit code use and export the results. Fleksa’s restaurant marketing automation page describes email and SMS campaigns, segmentation and promo codes. Confirm the specific workflow and any setup needed before making it part of your plan.
Measure completed second visits, then check the workflow
Message opens and clicks can tell you whether people interacted with a message. They do not establish that a guest came back. Your primary measure should be the number of eligible guests with a matched, completed second purchase within the chosen measurement period. Divide that number by the number of guests in the group you are measuring, and report both the count and the denominator.
In the hypothetical cohort, if nine of the 54 guests make a matched second purchase within 28 days of their planned send date, the observed second-purchase rate is 9 divided by 54, or about 16.7%. Keep the six guests who returned before their planned send in a separate pre-send return count. They should not be credited to a message they were never meant to receive.
That 16.7% is an observed rate, not the campaign’s lift. Some guests might have returned without a message. To estimate the difference, retain a comparable eligible group that receives no campaign message, if your audience is large enough to make the comparison useful. Measure both groups over the same period and show their counts as well as their rates. An unmatched counter purchase can still make your reported rate too low, so describe which channels are captured.
Review a few supporting measures alongside the return count: messages sent, delivery failures, opt-outs, code redemptions and total discounts. Opens are useful for diagnosing a message, but do not substitute them for purchases. If an offer generates redemptions, compare the discount total and other quoted campaign costs with the sales observed; neither figure alone tells you the profit or the number of visits caused by the campaign.
Before launch, run a simulation with sample records in a safe test environment. Do not send a real guest message, take a card payment or dispatch a courier. Your acceptance check should cover these cases:
- A completed first purchase with a dependable identity and the relevant channel permission enters the waiting audience.
- A canceled purchase, a record with missing identity data and a record without the relevant permission do not enter the send audience.
- A guest who makes a matched second purchase before the planned send is removed before any message is sent.
- A guest who withdraws permission before the planned send is suppressed.
- A matched, completed second purchase inside the measurement window appears in the return count; an open without a purchase does not.
Record the expected and actual result for each simulated case, plus who will fix any mismatch. Once those checks pass, a live service pilot is a separate decision: use a small eligible audience, follow your normal permission and send controls, and monitor the actual campaign. A live campaign involves real guest messages, unlike the simulation. Pause it if suppression, preferences or reporting do not work as expected.
For broader questions about connecting customer records and daily operations, review restaurant management software alongside your ordering workflow. You can also use the related buying guide to prepare questions about data and campaigns.
Next step: Write down your first-purchase definition, one contact window and the five acceptance checks above. Then compare your marketing requirements with Fleksa’s available workflows and ask how each requirement would be configured for your restaurant.
Frequently asked questions
When should a restaurant invite a first-time guest back?
Choose a delay that fits the reason guests visit. A week can be a useful starting hypothesis for a frequent lunch occasion, while a restaurant used less often may need a longer window. Remove guests who return before their scheduled message.
Should every new guest receive a discount?
No. A useful reminder may be enough. Compare it with an offer, account for the discount and campaign costs, and avoid assuming every discounted purchase was caused by the promotion.
What if a guest returns through a different channel?
Count the purchase only if you can reliably match it to the original guest and it falls within your defined measurement period. Report which channels your measurement covers so an unmatched return is not mistaken for a guest who never came back.
